Arnold Kling  

10-year Budget Projections

Choose Your Metaphor... Keynes and Hayek, Round 2...

Will Wilkinson writes (on Facebook),

Tell me what you think about the usefulness/uselessness of 10-year budget/deficit projections.

You can think of this as three questions.

1. Are the projections useful? I think they are useful if understood as conditional forecasts. They tell us what will happen if key macroeconomic variables follow a reasonable path. I believe that Congressional Budget Office knows something useful about how GDP and employment affect tax revenues, so I think this is useful. I think trying to go the other way--taking projections about deficits and using them to predict GDP--the CBO knows much less, because that involves theory that is less reliable

2. Is the 10-year window helpful? It can be gamed. You can make a policy look good under the 10-year window even though catastrophe follows. It is possible that the proponents of such a policy will be called out, but that does not always work.

3. Does anyone care about the projections? Here I could argue both ways. On the one hand, the fact that politicians take the trouble to game the projections says that they care about them, at least from the standpoint of appearances. On the other hand, the phenomenon of gaming suggests that in a sense the politicians do not care, except in terms of appearances. The projections provide what Murray Edelman would call symbolic reassurance, and meanwhile the politicians do whatever they want.

Probably the most interesting example is the projection for the effect of Obamacare on health care spending. This illustrates gaming, involving a combination of policy assumptions and the spending window--mostly the policy assumptions, actually. The proponents insisted that cutbacks to doctor payments be scored as part of the legislation, even though past experience would suggest that it would be highly unwise to assume that those cutbacks will take place.

CBO expressed skepticism about the doctor payment cutbacks and made it possible to draw inferences about the path of health care spending under Obamacare under the more realistic assumption that doctor payments are not curbed. However, supporters of Obamacare have chosen to ignore those inferences. Hence the mantra, "repealing Obamacare will raise health care spending and increase the deficit." That mantra is false unless you buy into the projections as gamed.

In the end, I come out strongly in favor of CBO projections, even though in this case they are being used to promulgate nonsense to two decimal places. I think that as long as you have the Journolist types who want to shape the story in Obama's favor, that is where the fault lies. Not with 10-year projections per se.

Comments and Sharing

COMMENTS (10 to date)
Noah Yetter writes:

Expressing recurring spending as a stock rather than a flow is fundamentally incorrect, and this warps the public's thinking on the budget. By expressing it as a single number, politicians disconnect us from the reality that their proposals involve some amount being spent annually, forever. Or until future-politicians change it, which is the other huge lie of multi-year projections.

rpl writes:

Noah, you make a good point, but expressing the spending as a flow has its own problems, namely, that the value of that flow typically changes over time, even for relatively simple policies. Should we report the net cost in the first year? The fifth? The tenth? Wherever you choose to report it, a politician determined to deceive can arrange for costs to look smaller during the benchmark year. Expressing it as a 10-year sum makes that a little harder, since any chicanery must be applied over the entire 10-year period.

There really is no getting around that any attempt to reduce a complicated profile of future spending to a single number is going to be susceptible to gaming. The practice is still useful for comparing honest attempts at policy making, but there is no substitute for being on guard against overt dishonesty.

Phil writes:

10 is an arbitrary number and simultaneously too long and too short. Best practices in public budgeting tend to use medium term expenditure frameworks (3-7 years) updated with each budget cycle, incorporating the associated changes in assumptions and policies. The better governments will publish a direct comparison between the current forecast and the prior year forecast for purposes of both decision-making and accountability. The U.S. national budget provides a five year forecast at the summary level, but does a terrible job of reconciling from year to year. So that's the too long part.

10 years is too short for some policies. Accompanying the annual budget, the US publishes a 75 year projection for programs like social security. 75 is also arbitrary, but I think a long range (at least 25 years) proojection of current policies is a sound practice.

Pete writes:

As I read the conclusion of this post, Arnold is claiming that 10-year projections are simply tools. The problem is not with the tool itself but how the tool is used, in this case by "Journolist types who want to shape the story in Obama's favor."

If that's the case, though, it doesn't make much sense to come out strongly in support of the projections. If we (meaning the CBO in this case) create a tool which is ultimately used for bad ends, then one way to expose the misuse or abuse of the tool is to simply remove it. By getting rid of the ten-year score, we would force those who misuse it to instead write their own scores. This would make it more painfully obvious that they are choosing an arbitrary time window through which to analyze budget proposals, and help readers focus on both the short- and long-term effects of any policy.

David C writes:

Huh? You didn't exactly specify, but I'm guessing you're referring to the Medicare "doc fix" issue which would require an additional $208 billion over 10 years according to the source. But Obamacare didn't make unrealistic cuts in payments to Medicare doctors; those cuts were already in place from previous legislation. Obamacare didn't fix that issue, but it didn't make it worse either. If Obamacare had expanded Medicare, then I could see the point, but I don't see how you can blame a piece of legislation for not solving a problem. Maybe you're referring instead to something related to Medicaid, but I'm not sure what. Also, for those curious, CBO estimated that Obamacare would continue to reduce the deficit after the 10-year window.

AirmanSpryShark writes:

What about replacing exponential discounting to a 10-year horizon with hyperbolic discounting to an indefinite horizon? Removing the sharp line should impair current shenanigans.

The reason for limiting the horizon of exponentially discounted forecasts is uncertainty; the time-inconsistency of hyperbolic discounting is consistent with a rational risk adjustment to exponentially discounted flows (see the URL linked from my name).

Lord writes:

Yes, healthcare reductions due to Obama have nothing to do with "doc fix". It may be unrealistic, but it was already in the projections that were shooting off to the moon before him as well. That just goes to show we have more work to do.

Wayne writes:

Here's an easy way to answer this: look back at their projections, look at what really happened, and then compare, correlate, and quantify.

Kyle writes:

There are lots of reasons simply doing an actual vs expected doesn't work in this kind of forecasting, the most obvious being the fact that the forecast itself impacts future policy, thereby making it wrong as a matter of its existence. If the CBO gives a warning of immediate financial danger and congress makes changes to rectify the situation (hey, we can dream), the "forecast error" is huge - does that mean it was wrong? Well just account for that change, right? But one quickly finds that no one can agree on which event or action exactly caused the improvement. Dueling stories.

There are also some more insidious problems. One is that there are lots of ways to measure the accuracy of a forecast. You'll choose some metrics that seem very good. They'll get good at playing against those metrics, while letting others spin out of control.

This doesn't mean actual vs expected type work shouldn't be done, but it will not be very enlightening.

Phil writes:

CBO actually reports on their own forecasting accuracy:

Kyle makes some excellent points about the challenges, even wisdom, of doing so. I believe there is probably more democratic value than economic value in it, but a public policy guy like me appreciates the fact they make the effort.

Comments for this entry have been closed
Return to top