Network News

X My Profile
View More Activity

Curves

According to Clive Crook, the Congressional Budget Office has found that "the Democratic plans ... would 'bend the curve' in the wrong direction." It hasn't found that. In fact, the CBO is specifically avoiding that question. This blog post from Doug Elmendorf didn't get a ton of attention when it appeared, but it's actually rather important for understanding how the CBO is conducting its analysis:

Although we think we can provide a rough indication of the level of federal health spending or the budget deficit 20 years ahead, we are not confident that we have an analytic basis for projecting their growth rates at that point, much less for evaluating whether those growth rates will continue in future years. Therefore, we are more confident talking about whether proposals would “lower” or “raise” the curve of the federal budget deficit or budgetary commitment to health care 10 to 20 years from now than we are discussing the shape of the curve in that time period or the level or slope of the curve beyond that period.

All this talk of curves is getting a bit baroque, but in essence, you have to understand that there are two separate issues in health-care spending. The first is how much we're spending in any given year. The second is how much that spending changes from year to year. To put it very simply, there's spending and there's growth rates, and the two are different.

When Elmendorf says that the CBO will focus on "raising" or "lowering" the curve, he's saying that the CBO feels able to predict spending, but not growth. And one of the key changes in the health-care debate over the past few weeks is that the CBO's estimates of short-term spending have wrested the conversation away from the administration, which has preferred to talk about long-term growth rates.

(Via Brad DeLong.)

By Ezra Klein  |  August 3, 2009; 3:24 PM ET
Categories:  Health Economics  
Save & Share:  Send E-mail   Facebook   Twitter   Digg   Yahoo Buzz   Del.icio.us   StumbleUpon   Technorati   Google Buzz   Previous: Should We Envy Texas?
Next: More on Red State Models

Comments

And in fact there's a further disconnect between what's going on on the federal balance sheet and what's going on in the economy as a whole -- the plan is that short to medium-term increase in federal spending will lead to lowered costs and lowered growth rates in health care costs more generally, some of which will benefit federal programs but most of which will go to the private sector.

To the extent that CBO isn't comfortable talking about growth rates, and is only looking at things on the federal ledger, it's got a double structural bias away from quantifying the full benefits of reform policies (which I think they've mainly been up front about, but hasn't stopped the misuse of their analysis).

Posted by: Mike_Russo | August 3, 2009 3:56 PM | Report abuse

"When Elmendorf says that the CBO will focus on "raising" or "lowering" the curve, he's saying that the CBO feels able to predict spending, but not growth."

No.

Elmendorf is saying that the he feels comfortable giving a qualitative, high-level view on whether growth rates will increase or decrease (compared to the baseline budget) beyond a ten year period. His statement IS about growth rates, NOT spending.

Read your own excerpt:

"we are not confident that we have an analytic basis for projecting their growth rates at that point.. Therefore, we are more confident talking about whether proposals would “lower” or “raise” the curve."

PS Yeah the Elmendorf post didn't garner a lot of attention, but you're also one of the few people who gave the Orszag smackdown a lot of attention. It would have been good to give Elmendorf some visibility, as this post was a direct response to Orszag.

Posted by: wisewon | August 3, 2009 4:05 PM | Report abuse

... and more importantly a reasonable response to Orszag. On the whole, its nice to have a non-partisan, professional view on spending projections, and doesn't seem unreasonable for CBO to provide qualitative guidance on the financial impact of health care, as a ten-year period is known to be too short by all parties.

Posted by: wisewon | August 3, 2009 4:08 PM | Report abuse

"All this talk of curves is getting a bit baroque,"

You're the best writer on the Interwebs.

Posted by: eRobin1 | August 3, 2009 4:20 PM | Report abuse

Wisewon -- I don't think that's right. If you click through, the paragraph from which that excerpt comes says, more fully:

If the projected budget deficit is lower 20 years from now under a reform proposal than it would be without any policy changes, then that curve is clearly being bent downward, on average, during the next twenty years; if the projected deficit is higher, then that curve is being bent upward. Would those downward or upward trajectories continue indefinitely? That sort of extrapolation might seem natural, but we simply cannot tell whether it is appropriate. Although we think we can provide a rough indication of the level of federal health spending or the budget deficit 20 years ahead, we are not confident that we have an analytic basis for projecting their growth rates at that point, much less for evaluating whether those growth rates will continue in future years.

Which is to say, they have a pretty good grasp on the spending levels and growth rates over the next ten years, they can project those growth rates out another ten years or so to get a rough sense of federal spending 20 years out, but at that point Elmendorf explicitly disclaims having any support in his modeling for estimating what those growth rates will be.

Then he says "raise" and "lower" in the last sentence, rather than "bent upward" or "bent downard" as he does earlier in the same paragraph.

Posted by: Mike_Russo | August 3, 2009 4:35 PM | Report abuse

Mike,

Read the whole post from Elmendorf. Another paragraph on the same point reads:

"After we have developed an estimate of the growth rate of the costs or savings in each broad category, we assess how those costs or savings would evolve from the end of the 10-year budget window through the following decade. The result is a very approximate sense of whether a piece of legislation is increasing or decreasing the federal budgetary commitment to health care or the federal budget deficit during the second decade and at the end of that decade."

The point is: The CBO's analysis on growth rates for the first decade is quantitative (growth rates determines the level of budget spend, that's how their model works, he's saying) and after the first decade he's giving a qualitative sense of whether the growth rate will be higher or lower than what's currently projected.

Ezra could ask him himself, but that is what he's clearly saying the post.

Posted by: wisewon | August 3, 2009 4:48 PM | Report abuse

is there any thought give to the fact that when medicare was first adopted they expected costs in 1991 of $9 Billion and ended up with costs of $61 Billion?

Everyone that talks about these costs needs to realize they're very broad estimates and unlike to remain consistent with reality when all is said and done. They're getting better at estimating these things but they're definitely not fullproof.

Posted by: visionbrkr | August 3, 2009 5:04 PM | Report abuse

Yeah, I did read the whole thing before I replied, and saw that paragraph -- I think what he's actually saying there, though, is what I said above. The key sentence is "[a]fter we have developed an estimate of the growth rate of the costs or savings in each broad category, we assess how those costs or savings would evolve from the end of the 10-year budget window through the following decade."

Since there's no reference to growth rates in the latter part of the sentence, I believe what that's saying is that 1) they model costs/savings and growth rates at the end of 10 years, then 2) they use those modeled growth rates to project forward another 10 years and check costs/savings then, without taking the growth rates seriously after that.

I agree that the second sentence you quote is a bit confusing with its reference to the "commitment", but read carefully, he seems to be saying that CBO's methodology allows them to make a very rough guess, 20 years out, of whether the bill is adding to or subtracting from the deficit -- which is just a matter of what the costs vs. savings will be at the end of 20 years, not what the growth rates will be.

I wouldn't claim that this interpretation is clear, since I wound up reading a good chunk of the post a couple of times to make sure I was getting it, but if you look carefully at Elmendorf's language, he is saying that their methodology doesn't allow him to estimate growth rates 20 years out.

Clearly this is kind of a lie, because in testimony he regularly talks about the kinds of things that will bend the (federal spending) curve, like taxing employer health benefits. But I think the key point is that while it's pretty easy to say "subsidies will bend the curve the wrong way for the feds, taxing benefits will bend it the other way" and be confident that that's right, 20 years out, if you've got a proposal that does both of those things, it's really really really hard to figure out which side wins out a couple decades from now.

Posted by: Mike_Russo | August 3, 2009 5:25 PM | Report abuse

Adding to be clear: per your first post, Ezra is oversimplifying by saying that CBO doesn't score growth rates, since it clearly does over the first 10 years. It just doesn't project growth rates after that. It's kind of an awkward position to be in, since growth rates a decade from now are kinda far enough out that you start to get a sense of whether the curve's been bent, but really it isn't until 20 or so years out that you have a firm sense of whether things are working or not (assuming 5 years of phase in, 5 years of implementation, so one decade isn't actually a lot of time to gain much traction).

Posted by: Mike_Russo | August 3, 2009 5:29 PM | Report abuse

visionbrkr could you source that claim about the 1991 projection? I see it all the time but it reeks of urban legend of the type we always see about Social Security.

Per this source Medicare expenditures went from 9.7% of national health care costs in 1967 to 19.4% in 1997
http://findarticles.com/p/articles/mi_m0795/is_1_22/ai_74292512/
To accept your point we would have to believe that the 1965 projection was off by almost 6X implying they predicted that percentage of national health care costs covered by Medicare going down from 9.7% to something like 3.3%. The numbers don't seem to add up and a link would be much appreciated.

Posted by: BruceWebb | August 3, 2009 6:19 PM | Report abuse

http://cboblog.cbo.gov/?p=328

Who says the CBO does not project growth rates after ten years? Check out the third figure from the Director's Blog of July 16th of this year. That shows combined Social Security and Medicare spending over the entire 75 year actuarial window.

If you want to see Medicare broken out the statement to Elmendorf's testimony to the Senate show not just one but two different sets of projected growth rates for Medicare in figures 1 & 2.
http://www.cbo.gov/ftpdocs/104xx/doc10455/07-16-Long-TermOutlook_Testimony.pdf
Once again over the 75 year window.

Someone is badly misreading Elmendorf here. What I see him saying is that CBO doesn't have a good methodology for projecting overall national health care spending but does have one for projecting federal health care spending.

It is not that CBO doesn't have long term modeling, in fact they explicitly lay out that model in the following
http://www.cbo.gov/ftpdocs/103xx/doc10328/06-26-CBOLT.pdf
It just that their Long Term Outlook methodology used to project long term economic factors is somehow somewhat at odds with their methodology for scoring legislation.

Most of their reports are limited to the ten-year window, but that doesn't mean they don't peak through the window now and again.

Posted by: BruceWebb | August 3, 2009 6:45 PM | Report abuse

visionbrkr

the earliest data I could come up with in a hurry shows that actual spending in 1970 was $7.5 billion
http://www.cms.hhs.gov/ReportsTrustFunds/downloads/tr2002.pdf (~1 MB)
See tables II.A1 and II.A2 on page 32. In find it hard to believe that anyone projected that spending 20 years out would only be up to $9 billion, even inflation adjusted.

Per this source Medicare spending was $3 billion in 1967 (its first full year). Again I doubt that anyone looking forward 24 years would see that only rising to $9 billion.
http://www.kff.org/medicare/upload/Medicare-Chart-Book-3rd-Edition-Summer-2005-Section-6.pdf Figure 6.5

But I'll be happy to look at your link.

Posted by: BruceWebb | August 3, 2009 7:00 PM | Report abuse

Bruce -- yeah, clearly CBO does project things significantly further out (as in that terrifying chart of health spending as % of GDP graph stretching almost to 2100).

I presume the difference is that it's a lot easier to project the status quo forward than a giant bill with tons of different policy tweaks. You can do the first-order approximations, but it gets tricky after that. As Elmendorf says in the linked post:

"Although we publish projections of the federal budget 75 years ahead, those projections are inherently uncertain and are designed to identify broad trends rather than to reflect specific pieces of legislation. Trying to project several decades ahead not just the evolution of the health care system under current law but also the effects on that system of a particular comprehensive and interacting set of reforms is extremely difficult."

And he goes into some examples of what those interacting pieces look like.

In the last paragraphs of his post, he also says that CBO doesn't have quite as much analytic power when it comes to overall health spending as they do with respect to federal spending, but that's a separate point.

Posted by: Mike_Russo | August 3, 2009 7:29 PM | Report abuse

I'm sorry, but I look at the record of CBO forecasting (Surpluses for the next 10 years - 1998) and I still think you would do better with a good shaman, a goat and a sharp knife.

Posted by: lensch | August 3, 2009 9:09 PM | Report abuse

"All this talk of curves is getting a bit baroque,"

I.e., it's not going Ezra's way.

Posted by: tomtildrum | August 4, 2009 10:31 AM | Report abuse

The comments to this entry are closed.

 
 
RSS Feed
Subscribe to The Post

© 2010 The Washington Post Company