Arnold Kling  

Concerns with the CBO

PRINT
"How Deserving Are the Poor?":... The Case for More Skilled Immi...

The Wall Street Journal reports


The inquiries of the Congressional Budget Office, which haven't been made public, concern the CBO's analyses of some of Washington's most complex and controversial measures, including bills on financial regulation, health care, small-business lending and efforts to aid the housing market, said people familiar with the matter.

The article refers to Republican staffers, and from what I can gather the inquiries do not have formal standing.

Meanwhile, apparently at least one left-wing think tank is up in arms over the CBO study that showed that Federal workers are overpaid. Andrew Biggs gives his perspective on that brouhaha.

In its overall budget forecasting, the CBO offers a useful check on politicians of both parties. I admit to being a big fan of the CBO for that reason. See the latest budget outlook.

I have two concerns about the CBO. One is that they are regarded as an all-purpose think tank by Congress, and this leads them into issues that are controversial and outside of their core competence. The most prominent example is forecasting the economic impact of fiscal policy, for which they use models that are not well regarded.

The study of Federal pay is another example of engaging in needless think-tankery.* In the end, I think CBO has more to lose than to gain by taking on such projects. At a time when the budget outlook is such a crucial issue, I wish that the CBO were focused laser-like on budget forecasting, without political distractions.

The other concern I have is that CBO has become, like the rating agencies during the mortgage securitization boom, a target for gaming. In my view, the Democrats gamed the CBO process in producing a health care bill that had "scorable savings" that almost surely were not real savings. Other watchdogs, such as the chief actuary for Medicare, called them out on this, but CBO was not able to do so. I suspect that the Republican resentment toward CBO stems from that experience.

It would be harder for legislators to game the CBO scoring system if CBO were less transparent. If legislators understood nothing about the details of the process and simply had to make rough guesses about CBO analyses, then they could not fine-tune bills to take advantage of scoring quirks.

This is a difficult issue with which to deal. On the one hand, legislators should obtain estimates of the fiscal impact of legislation before it is enacted. They should certainly make changes when the impact is not what is desired. What they should not be doing is playing games to make legislation appear to be better than it really is.

I do not see a simple solution for this problem. Perhaps a limit on the number of iterations of a bill that can be scored would (a) conserve CBO staff resources and (b) make gaming more difficult. However, it is hard to see how to devise and implement such a limit.

(*Part of my problem is that I reject statistical methods of determining government pay comparability in favor of the simple market test of looking at quit rates and application rates for government jobs.)


Comments and Sharing





COMMENTS (10 to date)
Jeremy N writes:

Could you elaborate on "Democrats gamed the CBO process in producing a health care bill that had 'scorable savings' that almost surely were not real savings."?

Why do you have that impression? Can you point me to anyone else who's discussed your point of view?

Thanks

Justin writes:

While it obviously doesn't completely resolve the problem, wouldn't a great deal of the problem with CBO forecasts go away if the CBO were allowed to make reasonable assumptions about the future even if they conflict with the legislative baseline? For example, when Congress under both parties has repeatedly failed to enforce the sustainable growth ratio (SGR) with a series of "Doc fix" bills, it seems reasonable for the CBO to assume a baseline that does not include savings from the SGR. Similarly, if Congress passes an entitlement program like CLASS, it would make sense for the CBO to be allowed to either refuse to count the premiums as savings or to consider the net present value of future payouts under the program as costs. If you allow the folks at the CBO to apply their judgement about what savings are realistic and what savings are illusionary, it would at least eliminate the most obvious forms of gaming.

This would, of course, lead to an increasing desire to politicize the CBO since both parties would want to add analysts that would tend to apply judgement that was favorable to one party or the other. The current structure of the CBO, however, seems at least reasonably well positioned to withstand those sorts of political pressures.

PrometheeFeu writes:

"It would be harder for legislators to game the CBO scoring system if CBO were less transparent."

The flip side is that nobody would respect the CBO if nobody could double-check their methodology.

Daublin writes:

Jeremy, I'd be interested in Arnold's examples as well. Meanwhile, I can contribute a couple of examples as food for thought:

1. Phase in tax increases to happen in 8 years and benefit increases to happen in 12 years. A 10-year forecast will look much rosier than it deserves.

2. Include a steep benefit decrease that will nominally occur in 4 years, but make it very easy for Congress to change their mind in 3.5 years and not implement the decrease after all. The forecast would then include benefit decreases that won't really occur.

I don't see a good way to make it ungameable. Attacks like these seem fundamental.

Jeremy, Alabama writes:

To amplify the "gaming" aspect, it should be obvious that if models predict different outcomes, then the most favorable prediction will be preferred. Since models will show different results for different policies, models will (and of course have) become politicized.

One answer might be to deliberately politicize models, and have an "official republican" version and an "official democrat" version. Over time, one might be able to determine which one gives better predictions.

Hasdrubal writes:
The flip side is that nobody would respect the CBO if nobody could double-check their methodology.

Yeah, that's the real problem.

One idea would be to generate three forecasts using transparent methodology:
- An as-is forecast assuming no changes in laws currently active. (Like their current methodology.
- A worst-case forecast assuming all possible changes that will negatively impact the fiscal results that can happen will happen.
- An in-between forecast using the most likely legal changes, explaining what laws they think will be in effect, and why they think so.

This would make gaming more obvious and provide lawmakers and the public a better idea of what CAN happen, not try to predict what WILL happen. Down side, it'll create a LOT more work for the analysts.

In fact, didn't the CBO just recently put out a best/worst case pair of estimates on the budget deficit?

Alternatively, they could eschew point estimates and only give ranges: This legislation will result in something between a $100 billion surplus and a $1 trillion deficit over the next 10 years.

Arnold Kling writes:

Two examples:

1. About $50 billion in savings were supposed to be generated by the CLASS act, which subsequently was declared non-implementable by the Obama Administration.

2. Also, the Chief Actuary for Medicare issued an analysis that was much more pessimistic than that of the CBO.

Fralupo writes:

I'm going to have to come down on the opposite side of this: I see no point in having a CBO high priesthood that has the exclusive right to determine what the fiscal effects of a bill will be. Far better, in my opinion, for Congressmen and staffers to have a good idea of the costs of what they want to implement as soon as possible.

If this means that staffers write provisions into laws that make them look less costly or higher-taxing, then it should also be the case that such shenanigans can be more easily exposed. The fact that you can cite the cost of the CLASS feature in Obamacare is a good example of this.

Ultimately Congress will pass whatever its going to pass. If the CBO has an influence it shouldn't be to restrain or propel laws, it should be only to inform our representatives of what their laws are actually going to cost.

Chris Koresko writes:

This discussion about the CBO and how its rules are being gamed reminds me of the financial sector and the way its regulations are being gamed.

If we've got a rough consensus that it's hard to regulate the financial industry because the rules get gamed, it seems at least plausible that the same arguments apply to the CBO, too.

Perhaps the solution, if there is one, lies not with the "production" side where the forecasts are made, but with the "consumption" side where the results are interpreted. In other words, make people more aware of what these forecasts mean and let them decide how much to trust them.

Chris Koresko writes:

Jeremy N: Could you elaborate on "Democrats gamed the CBO process in producing a health care bill that had 'scorable savings' that almost surely were not real savings."?

In addition to the examples Arnold Kling and Daublin mentioned above, I'm pretty sure ObamaCare includes some big new taxes on things that have nothing to do with health care. Those taxes would properly have been separate laws, to be debated and passed or rejected on their own merits. But putting them in the ObamaCare bill made the forecast for ObamaCare's impact on the budget deficit look better.

I also vaguely remember that there was double-counting of money, i.e., the same pile of money from some source was counted as being spent twice to fund two different ObamaCare programs.

Comments for this entry have been closed
Return to top