Saturday, June 18, 2016

Demise of the Dots?

St Louis Fed President Bullard made news yesterday with the publication of a note titled, The St. Louis Fed’s New Characterization of the Outlook for the U.S. Economy, in which he argues for replacing the current FOMC notion of a long-run steady state with a regime-dependent policy framework. In particular, Bullard argues that FOMC member interest rate projections (the 'dot plots') have been consistent neither with market pricing nor with actual rate outcomes over time and that these projections should be replaced with a regime-dependent policy, in which policy is made under the assumption of indefinite regime persistence. In fact, Bullard declined to offer a long-run projection at the June 14-15 meeting, given his emphasis on this new framework.

In the media call coinciding with the release of his note, Bullard offered the view that inflation expectations likely would increase were his colleagues on the FOMC to adopt this framework, presumably because the policy rate that Bullard believes is appropriate for the current regime is only 0.63% -- well below the projections currently offered by his colleagues (and previously by Bullard) under the existing framework.

A key characteristic of Bullard's framework (mentioned four times in six pages of text) is the notion that economic regime switches are not forecastable. In fact, given the importance of this presumption in Bullard's argument, it's worth considering this concept in greater detail.

Nowhere in his note or the accompanying media call does Bullard define the sense in which he uses the term forecastable. But unless Bullard has some rather specific definition in mind, a reasonable presumption would be that regime switches could be characterized as forecastable if their associated states could be identified via observation and if the probability of entering those states could be quantified -- or at least characterized (eg, into high, medium, and low categories).

Bullard clearly believes regime shifts can be identified, since he includes in his note a schematic (in this case, a binomial tree) in which he not only lists these states but also illustrates the dependencies between the states. So Bullard's view that regime switches are not forecastable appears to rely on a view that it's not possible to characterize the probabilities of entering these regimes.

While I appreciate the empirical difficulty of estimating parameters of regime-switching models, particularly when the frequency of regime switches is low, this difficulty in and of itself doesn't prohibit the characterization of regime-switching probabilities. In this regard, I'm reminded of Phil Tetlock's research on forecasting, which for many years has involved asking a large sample of participants to associate probabilities with a wide variety of outcomes, most of which have no precedent. For example, one question currently being asked at Tetlock's web site involves the upcoming US Presidential election: Will an independent or third-party candidate win at least 5% of the popular vote? Probabilities offered for this question appear in graphical form going back to early April and have increased from roughly 5% at that time to 25% as of June 13.

As far as I'm aware, there's nothing preventing Bullard and his colleagues at the St Louis Fed from associating probabilities in a similar manner to the various regimes they identify for the economy. As a result, I don't see the inherent forecasting problem that appears so central to Bullard's argument. At a minimum, perhaps Bullard could simply engage Tetlock to organize an economic forecasting competition for the St Louis Fed.

Otherwise, I see two other economic problems with Bullard's argument. First, he argues that government bond yields in the current regime are subject to a large liquidity premium, which he estimates to be 137 bp. In the presence of such a large liquidity premium, presumably government bond yields in the market would be well below the yields that would prevail under the pure expectations hypothesis of interest rates. And in that case, one would expect market yields to be well below the interest rate projections offered by FOMC members. But of course this discrepancy is one of the factors motivating Bullard to dispense with the current framework. At a minimum, the discrepancy between FOMC member projections and actual market outcomes is less of a conundrum if the liquidity premium for government debt is as large as Bullard estimates.

Second, Bullard's estimate of this liquidity premium appears based in part on the notion that returns to capital in the economy more generally are not low and that this premium applies only to government bonds. In a footnote Bullard cites Ricardo Lagos' paper, Asset Prices and Liquidity in an Exchange Economy, but the data set used in this paper (at least in the working paper I've read) is not contemporary and therefore of limited use in characterizing the current situation.

More important, I suspect Bullard may be comparing ex post realized returns on equity and other forms of capital to ex ante expected returns implicit in government bond yields. For example, the realized return on many bond funds in recent years has been greater than the stated ex ante yields on the bonds, since these yields have been declining, resulting in capital appreciation of the bonds. Of course a similar effect is present with equities (ie, as the expected return of an equity declines, ceteris paribus, it's realized return increases).

Specific issues aside, an intriguing question is whether Bullard's attempt at leading the FOMC members away from their dot plots will be successful. My expectation is that Bullard won't persuade many of his colleagues on the FOMC to adopt his perspective, given the problems with his framework and the fact that some members of the FOMC (eg, Fischer) are formidable theoreticians

On the other hand, Bullard has a point in noting that actual market outcomes have been well below the FOMC projections since the advent of the dot plots. As FOMC communication policy continues to evolve, it's not hard to imagine the dots being replaced with something relying less on point estimates, such as the repo rate fan charts provided by the Riksbank. Alternatively, it's not to hard to envision the FOMC dispensing with the dot plots altogether in the next few years, given the repeated discrepancy between FOMC projections and actual market outcomes.

Finally, it's worth noting that Bullard's attempt to exert more intellectual leadership on the FOMC may be a sign that Yellen doesn't have as firm control over her FOMC as Bernanke and (especially) Greenspan did over their Committees. In particular, it's difficult to imagine anyone on the FOMC under Greenspan taking any action that could be interpreted as undermining the Chairman, such as failing to submit a long-run projection when requested. In this regard, it's also worth noting the efforts of Minneapolis Fed President (and recent gubernatorial candidate) Neel Kashkari to exert leadership on the Too Big to Fail issue, which appears to have bedeviled Yellen at times (eg, when questioned by Senator Warren during semi-annual appearances before the Senate Banking Committee.)

For now, Bullard appears to have created a slightly awkward situation, in which he'll be declining to offer a long-run projection while his colleagues continue to provide theirs. While I suspect this will have no discernible outcome on actual policy over the next few quarters, over time it may add to a sense of diminishing coordination among Committee members, with the FOMC operating increasingly like the MPC at the Bank of England. And in the meantime, it will be interesting to see the way Yellen and particularly Fischer respond to Bullard's proposal and the economic analysis on which it's based.