When “Forecasts” Are Not Really Forecasts

Jeff Madrick

I have written about the deep and misleading flaws inherent in Congressional Budget Office (CBO) forecasts before. But given last week’s projections of job losses due to a proposed minimum wage hike, the inadequacy and misleading character of CBO pronouncements needs addressing again.

What particularly provokes me now is a quasi-debate between economist Jared Bernstein, former economic adviser to Vice President Joseph Biden, and Republican economist Douglas Holtz-Eakin, former head of the CBO, on CNBC. The CBO had issued its customary over-simplified statement that “Once fully implemented in the second half of 2016, the $10.10 [per hour minimum wage] option would reduce total employment by about 500,000 workers, or 0.3 percent, CBO projects.” They next point out that this is merely the midpoint of a range of possibilities they derive from existing research on the relationship between the minimum wage and jobs.

But why make the declarative statement in the first place? The damage is done. Politicians and the media pick it up as if it is a forecast, not a midpoint based on a wide range of conflicting research.

Here’s my point. It wasn’t only unsophisticated commentators who treated the 500,000 as an outright forecast. So did Holtz-Eakin, and for a while Bernstein went along with it. Holtz-Eakin kept saying he was very concerned about the 500,000 who would lose jobs. He didn’t even hint that the CBO said the distribution went from almost no job loss to one million. And even this only covers two thirds of the possible range (I assume a simple standard deviation estimate under a bell curve is what the CBO provided).

If two trained economists treat the 500,000 job loss as a real forecast, what are the media, the pre-committed politicians, and the nation to do? Bernstein finally said he thinks the CBO “high-balled” the estimate, which they almost certainly did. But then he said the loss might be only 200,000 or 250,000. In fact, the job loss might well be zero or there might even be a slight gain in jobs due to increased demand for goods and services.

There is an informal pressure to present a simple, one point analysis for Congress, much the way Wall Street economists forecast growth rates to decimal points. Their clients demand it. But my point is that CBO pronouncements are given far more weight than their due because of the over-simplified way they are presented.

This is very bad economics. The real CBO—if we actually had one—should be explaining to Congress and the public how difficult it is to make such projections, what the sensitivities are to different events, and how conflicted the research is. A midpoint in a distribution is not a forecast.
To cover themselves, of course, the CBO economists place their qualifications and footnotes in appendices that are dense. But even these are not always clarifying. In the case of the job loss due to the minimum wage hike, they say they took account of unpublished research that had fairly low job loss predictions, and also emphasized state-to-state methodologies over national employment growth rates. The latter would have lowered their job loss forecast.

They also apparently relied on a very recent study that reported large job loss effects. The methodology of this study is highly dubious. Arinjadrit Dube of University of Massachusetts at Amherst has chewed it up. He noted, for example, that the paper found bigger job loss effects in manufacturing, where there are few minimum wage workers who could be affected, compared to modest job loss effects in retailing, where there are many.

The problems with the CBO are bigger than this latest brouhaha. First, they have structural and institutional problems. The pressure to come up with an unambiguous forecast on which to base policy leads to bad, over-simplified economics, no matter how many qualifications are put in the appendices. The qualifications should be emphasized in the primary text; the CBO should teach Congress about the hypothetical nature of economics, not over-simplify economics for legislators. They should provide a range of answers, not one.

Second, the CBO regularly makes ideological assumptions that take neo-classical propositions at face value. For example, in recent years they have assumed that a rising federal budget deficit could avoid or shorten a recession in the short run, but that the higher level of debt would invariably lead to slower growth down the road. More disturbing, they invariably assume weak economies are self-adjusting. It’s Say’s Law, pure and simple. The economy will automatically rise from the ashes in three or four years at the latest. In 2012, the CBO assumed that if the U.S. took a recession in 2013 by failing to increase the deficit, it would have been significantly better off by 2016 or so than if it had increased the deficit, a la Keynes. The economy would have grown more rapidly—it’s as simple as that.

And, frankly, they often make preposterous assumptions. The CBO budget report of mid-2012 included an alternative forecast to account for the unlikelihood that the Bush tax cuts would be eliminated. (As a reminder, they must make forecasts according to current law.) They forecast a debt-to-GDP level of about 200 percent by 2037.

In a response about a year ago, the Center for American Progress economist Michael Linden showed that the underlying assumptions were pretty wacky. First, the CBO assumed that tax revenues would not rise as the economy grew from that point. But tax revenues usually rise with a growing economy, so to keep it constant would require a tax cut. Second, they assumed that discretionary social programs would be entirely restored to their old levels against GDP, again highly unlikely. With these and other assumptions corrected, the debt ratio fell closer to 100 percent in 2037, calculated Linden.

And that’s the level the CBO projected in its 2013 report a year later—around 100 percent of GDP in 2038. Now, don’t misunderstand. The CBO does give alternative projections. They are simply rarely cited by the public because they are either buried or no one really cares, once the main forecast is made. In 2013, the CBO they noted, for example, debt-to-GDP could range from 65 percent of GDP to 156 percent of GDP in 2038. Did you hear many talking about this wide forecast range?We should have.

The CBO needs serious reform based on acknowledgement of how tentative future forecasts are. It simply should not present one dominant forecast, or a single number like the 500,000 jobs loss. It should present a range and explain the variables that might affect it. In its more detailed analysis, it should be more inclusive of minority or heterodox views, against which it is usually biased.

It would be splendid if an economic group analyzed the accuracy of CBO forecasts as well as their political influence. My guess is it would be highly damaging to the CBO’s reputation, and place much-needed skepticism on their announcements. If we can’t reform the CBO, we should at least develop a “shadow” CBO, like the old shadow monetary policy committee, which can offer sensible alternatives and criticism of the CBO conclusions. Economics is mostly a “science” of unproven hypotheses and a CBO, true to its calling, would acknowledge that.