the fed and the yield curve
Global Policy & Central Banking Strategy January 5, 2018 Krishna Guha 202-872-5260
[email protected]
Ernie Tedeschi 202-872-5262
[email protected]
THE FED AND THE YIELD CURVE
FOMC will not ultimately be deterred by risk of curve inversion; if yields stay low it might even do more
Summary: Going into 2018 there is intense market focus on the possibility that the yield curve could invert if the Fed presses ahead with ongoing rate hikes — and what this would mean in terms of recession risk. Minutes from the December FOMC meeting show policymakers grappling with the same issue, with the Committee divided between those who are worried by a flattening curve and risk of inversion and those who are less concerned. We believe that the Committee will keep a careful eye on the yield curve but will not ultimately be deterred by curve flattening from continuing to raise rates and will likely conclude that in current circumstances even outright inversion would not signal elevated recession risk. In brief we argue: ■
■
■
■
■
■
■
The December FOMC minutes released this week give prominent treatment to a debate over how to interpret the flattening of the yield curve and the possibility that an inverted curve could portend elevated recession risk We think the concern about curve flattening is centered among a subset of regional Fed presidents mostly with more market-oriented / non-academic economist backgrounds and is unlikely to be shared to the same degree by the core leadership of the Fed and its staff There is broad agreement that the current level of the term spread is not particularly alarming today: we run 11 versions of a slope-based recession forecasting model and these give four quarter recession probabilities of between 0.5 per cent and 14.5 per cent with a central estimate around 8 per cent The debate is about how worried the Fed should be if the flattening were to continue and the curve were to invert outright, which has traditionally been a fairly reliable signal of recession We think the Fed institutionally (FOMC leadership, Board staff) regards a narrow focus on the difference between short term and long term yields as a relatively crude way to assess recession risk and favors a richer evaluation of the yield curve including the information that can be distilled about the neutral rate and term premium The debate over how to respond to a flattening curve is an early example of how the Fed reaction function is becoming less predictable with heavy turnover on the Committee and a shift in composition from economists to non-economists; nonetheless we think the institutional Fed view will probably prevail Specifically we believe the Fed institutionally is sympathetic to the view — which we share — that the historic correlation between an inverted yield curve and recession probability is mostly a reflection of a more causal underlying relationship between the level of interest rates relative to neutral and the likelihood of recession
January 5, 2018
2
When the term premium is material and positive, as it was for most of the post-war period, the yield curve can only invert when interest rates rise well above levels that are expected to prevail on average in the future, which roughly correspond to the longer-run neutral rate, meaning that policy is likely quite restrictive But when the term premium is very low or negative, as it appears to be the case today, the yield curve could invert without interest rates rising above levels that are expected to prevail on average in the future, meaning that policy may not be restrictive at all There are a few caveats to this view: we cannot directly observe the term premium so we cannot be certain it is very low / negative; even if it is, curve inversion could still indicate that policy is quite restrictive if the spot neutral rate is below the longer-run neutral rate; and policy might be too restrictive from the perspective of managing inflation expectations / risks even if recession risk is not elevated Given the qualifications above the Fed will not dismiss the recession risk signal from a flattening curve but will cross-check against macro data, information from other asset markets and prices, and surveys including the Primary Dealer survey — to date all of these suggest the economy has strong momentum and near-term recession risk has declined to modest levels Indeed the macro, market and survey data suggest that the spot neutral rate has moved up and closed much of the gap with the longer-run neutral rate, providing a strong fundamental rationale for why the curve should have flattened, and providing less ground for concern in the event that the curve were to invert outright Further the macro, market and survey data suggest that the failure of longer term yields to move higher as the Fed has raised short term rates — the origin of the flattening dynamic — is being driven by global QE spillovers from the ECB and BoJ and not concern about the US outlook The influence of foreign QE and bond yields on the US term premium is a positive policy shock that should boost activity and near-term reduce recession risk If the yield curve flattens because longer-term yields are tightly pinned down by global factors then the Fed may continue to struggle to tighten financial conditions, and this could ultimately give the FOMC further cause to revise up its estimate of spot neutral and raise rates more rather than less to achieve an appropriate stance of policy In our base case however rising global yields (possibly accompanied by some mild upward bias to estimates of the longer-run neutral rate as spot neutral is revised higher) over the course of this year will bail the US out of an inversion scare
January 5, 2018
ATTENTION TO SLOPE OF YIELD CURVE IN FED MINUTES AMID INTENSE MARKET FOCUS1 Minutes from the December FOMC devote substantial space to a debate about the flattening of the yield curve in 2017, the possibility that the curve could invert outright in the period ahead, and what this might portend in terms of recession risk. This follows intense discussion in financial markets as to the risks around yield curve inversion, which has historically been a reasonably good predictor of recessions, with a number of regional Fed presidents including Kaplan,2 Kashkari,3 Harker4 and Bullard5 drawing attention to the importance of this signal in speeches and interviews. The minutes show FOMC participants “generally agreed that the current degree of flatness of the yield curve was not unusual by historical standards.” However “several” thought it would be important to continue to monitor the slope, while “some” expressed concern that a possible future inversion could signal a slowdown or pose a risk to bank profitability and financial stability. By contrast a “couple” viewed the flattening as the expected consequence of raising the federal funds rate and judged that a yield curve inversion under current circumstances “would not necessarily foreshadow or cause an economic downturn.” FOCUS ON THE SLOPE ITSELF IS AT ODDS WITH WHAT WE THINK OF AS THE MODERN FED INSTITUTIONAL VIEW AND COMES FROM OUTSIDE THE LEADERSHIP The attention to the slope of the curve itself is notable because it seems to us more reflective of a market view than what we believe to be the core Fed institutional view. We believe that the Fed institutionally (Board staff, FOMC leadership) regards narrow focus on the slope of the curve and prospect of inversion as a relatively crude indicator of recession risk. This is not to say the Fed ignores the evolution of the yield curve — quite the opposite — but rather to say that the Fed conducts a richer evaluation of the curve, drawing information from the level of current and future expected rates as well as the evolution of the term premium and what appears to be driving this, factors that are lumped together in the simple difference between longer term and short term rates. 1
Andrew Hipolit provided research assistance for this note. “Number one, I actually have been very concerned. So the Fed sets the fed-funds rate, right? And we know the balance sheet, quantitative easing had more effect on the curve, but we’re now unwinding that. So the fed-funds rate is the rate we set. I also look at the two-year, the five-year, the 10-year, the 30-year. And along with – (inaudible) – that as the Fed has raised rates, the 10-year has actually backed down. That’s a little ominous, cautionary for me. I don’t view that as necessarily an easing. I view that as a comment on future economic growth. And what I don’t want to see us do is raise rates so fast we create an inverted yield curve, because history has shown an inverted yield curve has tended to be a precursor to recession. So back to the gradual – (inaudible) – watching the 10-year hang around 2.30, 2.30-something [percent], I’m very cognizant of it because it suggests it limits the ability of the Fed to raise rates too rapidly. It certainly has an effect on the limit – limits to what level we can get it to if you want to avoid a flat or inverted yield curve. That’s number one.” — Robert Kaplan (October 10, 2017), Q&A following appearance at Stanford University. 3 “It’s also possible that we might do it ourselves on the front end. By prematurely raising rates, it’s inverting the yield curve. I understand that the term premium is low now, and so maybe this time is different. But we shouldn’t just dismiss the signals that the bond market is sending us. So it’s possible that – if we continue down this path and inflation expectations don’t come up, it’s possible that we could end the expansion by our own actions.” — Neel Kashkari (December 18, 2017), interview with The Wall Street Journal. 4 “But there are risks to running low on inflation for too long, and we know those. And that’s where we – to me, we should prudently – move rates up to the neutral rate, but prudently. We don’t want to run any risk of disrupting markets, no risk of inverting the yield curve, you know.” — Patrick Harker (October 17, 2017), interview with The Wall Street Journal. 5 James Bullard (December 1, 2017), “Assessing the Risk of Yield Curve Inversion.” 2
3
January 5, 2018
Indeed our best guess — taking the unattributed minutes alongside the public remarks of individual FOMC participants — is that concern about yield curve flattening is coming from outside the core Committee leadership, mostly from more market-oriented non-academic economists who are less deeply imbued with Fed institutional thinking and policy analysis during the 2000s Greenspan “conundrum” period, with the exception of Bullard for reasons we will discuss below. We suspect that the institutional perspective will prevail and ultimately redirect the discussion from the slope and risk of inversion to broader analysis of the yield curve and specific focus on the stance of policy relative to the neutral rate along with the reasons behind a low term premium. But there is some residual uncertainty, part of a larger theme as the Committee is reshaped, loses institutional experience and rebalances from economists towards a larger share of non-economists. AT PRESENT ALL AGREE THAT EVEN AFTER FLATTENING THE YIELD CURVE IS NOT AT PRESENT SIGNALING ELEVATED RECESSION RISK While a flattening yield curve could be an early sign of recession, it is also mechanically exactly what one would expect to see as short-term policy rates move close and closer to longer-term neutral (holding constant any term premium dynamics). Further, while the Treasury curve has flattened, it has not inverted, which has historically been the more reliable indicator of an imminent recession. None of the almost-dozen probit models we ran, varying in sample size and specification, show 12-month ahead recession probabilities above 15 per cent at the moment, and some that include information from the ACM term structure model show probabilities below 1 per cent (see the Appendix for more details on our exercise). And these models draw purely from Treasury yield and inflation data; other financial and real indicators that could be included in a richer model are still running at paces consistent with an expansion. That is not to say that these models counsel complacency: recession probabilities in yield-based models tend to rise suddenly and precipitously, not gradually and with ample warning. On the other hand, while this class of model tends to perform reasonably well in anticipating actual recession, it also produces false positives, as can be seen in the chart below. Still, these models are in broad agreement that recent yield curve behavior is not historically-consistent with a high risk of imminent recession.
Recession probability: 12m ahead 100%
Model 1
Model 2 80%
Model 3 Model 4
60%
Model 5 Model 6
40%
Model 7 Model 8
20%
Model 9 Model 10
0% 1983
Model 11 1987
Source: Evercore ISI.
4
1991
1995
1999
2003
2007
2011
2015
January 5, 2018
RECESSION SIGNAL FROM INVERTED CURVE REFLECTS UNDERLYING RELATIONSHIP BETWEEN CURRENT SHORT RATES AND THE NEUTRAL RATE The issue is rather how concerned the Fed should be if the flattening were to continue potentially to the point at which the curve inverts outright. We believe the institutional Fed view is likely similar to our own — that the historic relationship between an inverted yield curve and recession probability is mostly a reflection of a more causal underlying relationship between the level of short term rates relative to neutral and the likelihood of recession in the period ahead. We can think of the ten year yield as composed of an expected path of future short term rates and a term premium. The expected path of future short term rates ought generally to correspond to the ex ante expected neutral rate over the longer run.6 Ordinarily (though not during the post-crisis period) we would assume that the spot neutral rate is reasonably close to the longer run neutral rate. Therefore when current short term interest rates exceed the average expected future short term rate, the stance of Fed policy is likely to be restrictive. In periods with a material positive term premium (most of the post-War period prior to the financial crisis) the yield curve can only invert when short term interest rates have moved well above the average expected future short term rate7 which is normally a reasonable proxy for the spot neutral rate. In such circumstances on the face of it at least, the Fed would be in very restrictive territory. And a very restrictive Fed would be likely to lead to output falling below potential — a recession. This is why even pre-crisis Fed analysis of the yield curve pointed to the importance of incorporating the current level of the federal funds rate in analysis of recession risk, not just the slope itself.8 YIELD CURVE INVERSION MEANS SOMETHING DIFFERENT IN PERIODS WITH A LOW OR NEGATIVE TERM PREMIUM However as Yellen argued in the December FOMC press conference, in periods when the term premium is low (as in the mid-2000s) or negative (as appears to be the case today9) a very flat or even inverted yield curve may well not convey the same information it does when the term premium is material and positive. Very obviously, the yield curve will naturally invert more easily and more often than when the term premium is materially positive. With little or no term premium, the curve can invert without current short term rates moving materially above expected future short term rates. Indeed with a negative term premium the curve can invert without current short term rates moving above expected future short term rates at all. If as argued above the causal relationship is not between the slope of the curve and the likelihood of recession, but between the stance of policy relative to neutral (proxied by the longer run average short term rate) and the likelihood of recession, then we ought not to worry about curve inversion as long as current short rates remain around or even below their expected longer-run average values. This we think is the Yellen - Dudley view, the staff view, and probably the view that Powell and the Committee as a whole will also embrace, subject to three caveats that require ongoing cross-checks against other market and macro data.
6
Inflation can complicate this assessment. Neutral is strictly a real not nominal concept. Further, the Fed might be seen as seeking to raise or lower the trend rate of inflation on a forward looking basis. 7 Because the longer term yield is equal to the sum of the expected future short term rates and the term premium. 8 See e.g., Jonathan Wright (2006), “The Yield Curve and Predicting Recessions.” 9 Based on decompositions using affine term structure models such as the Adrian Crump Moench (ACM) model. 5
January 5, 2018
CAVEAT NUMBER ONE: WE CANNOT DIRECTLY OBSERVE THE TERM PREMIUM SO WE CANNOT BE CERTAIN THAT IT REALLY IS VERY LOW / NEGATIVE As Bullard argued recently10 dismissing the signal from a very flat or inverted yield curve on the basis that this does not have its historic signaling value in a world with a very low or negative term premium could be risky given that it is impossible to directly observe the term premium. The Fed’s assessment of the term premium is based on a series of affine term structure models such as the Adrian Crump Moench (ACM) model that are regarded as state of the art but are nonetheless sensitive to certain modeling assumptions. We therefore cannot rule out the possibility that the term premium is in fact still materially positive and curve inversion would mean what it has historically meant — that the current short term rate has moved well above the rate that is expected to prevail on average in the future and as such is likely to be quite restrictive relative to neutral. This caution is reinforced by the fact that there is less statistical support than might be expected for the logical proposition that what matters is the relationship between short term rates and the average of future expected short rates as a proxy for neutral; in simple models historically the spread based on the total yield outperforms the spread relative to the risk-neutral rate that summarizes the expected rate path in the ACM model; see the appendix for more detail. This suggests that there may be a cyclical component in the term premium that has additional information regarding the likelihood of a near-term recession, though this is only one of a number of different influences on the term premium. ACM 10y bond yield decomposition
ACM 10y bond yield decomposition 16%
ACM 10y Fitted Yield
ACM 10y Term Premium
ACM 10y Risk-Neutral Yield
14%
3.0%
12%
2.5%
10%
2.0%
8%
1.5%
6%
1.0%
4%
0.5%
2%
0.0%
0%
-0.5%
-2% 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 Source: FRBNY
US 10y-3mo govt bond yield spread - historic 6 5
4 3
2 1
0 -1 -2 -3 -4 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 Source: Bloomberg, Evercore ISI
10
6
3.5%
ACM 10y Fitted Yield
-1.0% 2013 Source: FRBNY
2014
ACM 10y Term Premium
2015
2016
ACM 10y Risk-Neutral Yield
2017
2018
US 10y-3mo govt bond yield spread - since start 2013 3.0 2.5 2.0 1.5 1.0 0.5 0.0 2013 2014 Source: Bloomberg, Evercore ISI
2015
James Bullard (December 1, 2017), “Assessing the Risk of Yield Curve Inversion.”
2016
2017
2018
January 5, 2018
CAVEAT NUMBER TWO: EVEN WITH A NEGATIVE TERM PREMIUM CURVE INVERSION COULD INDICATE THAT POLICY IS RESTRICTIVE IF SPOT NEUTRAL REMAINS BELOW THE LONGER-RUN NEUTRAL RATE If the spot neutral rate is below the longer-run neutral rate — as the Fed argued over the past half-decade and arguably reiterated even in the December statement11 — then curve inversion could indicate that policy has become materially restrictive even with a negative term premium. This is because the comparison between the current level of the federal funds rate and its expected longer-run average value would not be a reliable proxy for the difference between the actual current rate and the neutral rate, particularly if inflation is low today but expected to average a higher value closer to the Fed’s 2 per cent target over the longer run.12 CAVEAT NUMBER THREE: EVEN IF THE FED CAN DISMISS INVERSION BECAUSE THE TERM PREMIUM IS LOW / NEGATIVE IT WOULD STILL NEED TO THINK WHETHER THIS & LOW INFLATION EXPECTATIONS MEANT THAT POLICY WAS TOO RESTRICTIVE Aiming off the slope of the curve itself as a reliable indicator of recession risk on the basis that the term premium is very low or negative would still require the Fed to think deeply about why the term premium is so low, and whether this might still indicate that policy was too tight. As mentioned above, there appears to be a cyclical component in the term premium historically that may have some informational power as to the likelihood of recession in the coming quarters. However, the term premium might also be low for other reasons (structural, US / global QE) and we think that is indeed the case today (see below). Separately, inflation break-evens are below historically normal levels, which might suggest that policy is inappropriately tight less from a recession risk perspective than by way of reinforcing below-target inflation expectations. Model-based decompositions — subject as before to critiques based on their specific properties — suggest the weakness is concentrated in the inflation risk premium, which appears to be negative.13 While this might in principle be associated with nearterm recession risk, in our view the negative inflation risk premium reflects the desire to hedge against the high potential cost of a return to the zero bound in the next downturn rather than elevated likelihood of a recession over the next year or two. This could still imply that policy under the current framework is suboptimal / too tight from a welfare-maximizing risk management perspective. At the same time, a negative inflation risk premium might be better addressed by adopting a different framework (ideally accompanied by non-monetary policy actions) rather than a different rate path; we will write more on this in the coming days.
11
The statement retains the longstanding guidance that “the federal funds rate is likely to remain for some time below levels that are expected to prevail in the longer run” — though this phrase is ambiguous as it does not specify whether the spot neutral rate is likely to remain for some time below the levels that are likely to prevail in the longerrun, or whether policy is likely to remain accommodative for some time. 12 To give a simple concrete example, if the average future short term rate is 3 per cent, reflecting an ex ante longerrun neutral rate of 1 per cent real plus 2 per cent longer run average inflation, but the spot neutral rate is zero real and current inflation is 1.5 per cent, then even with a -25 bp term premium, inversion would occur with the funds rate firmly in restrictive territory at 125 bp above neutral. 13 Again, subject to model-based decompositions. 7
January 5, 2018
THE FED WILL CROSS-CHECK THE SIGNAL FROM THE YIELD CURVE AGAINST THE SIGNAL FROM OTHER ASSET MARKETS, MACRO DATA AND SURVEYS Given the complications around how to interpret the signal from a curve that flattens further (as it did over the course of 2017) or threatens to invert (as it might do in the period ahead if the market reprices for additional Fed rate hikes near term) and the importance of understanding why the term premium is so low, the FOMC will evaluate the signal in the context of the state of the economy, developments in other financial markets, evidence from surveys of economists, and potential influences on US yields other than the outlook for the US economy. In doing so the Committee will be both seeking to assess whether there is any validation for the notion of elevated recession risk and how it should think about the evolution of the key underlying variables: the neutral rate and the term premium. MACRO DATA, SURVEYS SUGGEST US ECONOMY HAS GAINED MOMENTUM AND IS IN LITTLE DANGER OF STALLING We see — and we believe the Fed will see — no validation in the macro data to date of any suggestion that the US economy is in elevated danger of stalling or slipping into recession. Quite the opposite: the US economy appeared to step up a gear in 2H 2017 and is exhibiting considerable momentum in the New Year as reflected in PMIs and our own EVRISI surveys. A strong labor market with some prospect of moderately higher wage gains going forward, slight household tax relief and record net worth (albeit very unevenly distributed), strengthened incentives to invest post tax cuts and immediate expensing with reduced downside risk to global growth and moderation in the dollar, and decent prospects for residential investment all suggest that this momentum will be sustained through 2018. Indeed there is some upside risk to the median December SEP projection that growth will reach 2.4 per cent this year, reflecting in part the Fed’s cautious conventional modeling of the impact of tax cuts.
8
January 5, 2018
US ISM manufacturing & nonmanufacturing PMIs
US PMIs 50+=expansion 60
Composite
Manufacturing
50+=expansion
Services
Manufacturing
62
Nonmanufacturing
60 58
55
56 54 52
50
50
48 45 Jan-15 May-15 Sep-15 Source: IHS Markit
Jan-16 May-16 Sep-16
Jan-17 May-17 Sep-17
46 Jan-15 May-15 Sep-15 Jan-16 May-16 Sep-16 Jan-17 May-17 Sep-17 Source: ISM
ISI Company Surveys Retail, Trucking, Shopping Guide, Banks, Credit Card, Tech, Manufacturing, Cap Goods, Auto Dealers, Airlines 0=Weak, 100=Strong 55 54 53 52 51 50 49 48 47 Jan-15 May-15 Sep-15 Jan-16 May-16 Sep-16 Jan-17 May-17 Sep-17 Source: Evercore ISI
CROSS-CHECK WITH OTHER FINANCIAL MARKET INDICATORS ALSO SUGGESTS RECESSION RISK HAS FALLEN Moreover a cross-check against other asset markets suggests — to date at least — much more support for the idea that US recession risk has fallen than the idea that is has risen as the yield curve has flattened. The credit risk premium is very tight, earnings forecasts are moving up, and the (unadjusted) equity risk premium — while not extremely low by historic standards — has edged down over the past year. The dollar has weakened on net over the past year, but this appears to have been driven more by stronger growth ex-US and the surprise shortfall of US inflation than reconsideration of US growth prospects / recession risk. BofA Merrill Lynch US high yield bond index - option adjusted spread Basis points
Bloomberg US equity risk premium in 2017 7.5%
425
7.4% 7.3%
400
7.2% 7.1%
7.0%
375
6.9%
6.8%
350
6.7% 6.6%
325 Jan-17
Apr-17
Source: BofA / Merrill Lynch
9
Jul-17
Oct-17
Jan-18
6.5% Jan-17 Source: Bloomberg
Apr-17
Jul-17
Oct-17
January 5, 2018
PRIMARY DEALER SURVEY ALSO SUGGESTS RECESSION RISK HAS FALLEN AND IS NOT ELEVATED Consistent with the macro data, surveys including the New York Fed Primary Dealer survey indicate that private sector economists have lowered their estimate of recession risk as the yield curve has flattened to date, and see only moderate recession risk over the next few years in spite of the length of the expansion and the low level of remaining slack conventionally measured. FRBNY Survey of Primary Dealers - what percent chance do you attach to the US economy being in a recession in 6 months? By survey date
25%
25th Percentile
Median
75th Percentile
20% 15% 10% 5% 0% Jan-15 May-15 Sep-15 Source: FRBNY
Jan-16 May-16 Sep-16
Jan-17 May-17 Sep-17
INSTEAD MACRO AND MARKET DATA SUPPORT VIEW THAT THE SPOT NEUTRAL RATE HAS MOVED UP Indeed we read the macro, market and survey data as indicating not only that recession risk is indeed low but also that the spot neutral rate has moved up as the global economy has entered a more synchronized expansion with reduced left skew to the macro distribution and associated endogenous easing of financial conditions, and is now reasonably close to its longer run estimated value. This has been a core part of our thesis since late summer of 2017 and it underpins our view that the Fed has more headroom to raise rates than is widely appreciated without moving into restrictive territory. In the context of the yield curve, this would mean that the Fed need not be concerned about mild curve inversion under a very low / negative term premium as spot neutral would be close to longer-run neutral. AND THE TERM PREMIUM IS BEING HELD DOWN BY GLOBAL QE SPILLOVERS, WHICH COULD FORCE THE FED TO DO MORE TIGHTENING NOT LESS Meanwhile we believe the macro, market and survey data also supports the view that factors other than the US economic outlook are likely depressing the term premium, in particular spillovers from ECB and BoJ QE in integrated global capital markets as well as more underlying structural factors. From the perspective of the US, global QE spillovers constitutes a positive policy shock that ceteris paribus would tend to boost aggregate demand, lower recession risk and increase upside risks.14
14
Of course this is a general equilibrium problem, and the stance of policy ex-US also reflects economic conditions ex-US, so what really matters here is the global policy stance relative to economic conditions ex-US. 10
January 5, 2018
Indeed at some limit if the Fed is unable to push longer term yields higher or otherwise tighten financial conditions (for instance, via renewed appreciation of the dollar) it might reasonably conclude that the spot neutral rate has moved up further and it needs to raise rates further to achieve an appropriate stance relative to neutral. This could lead to a replay of the 2000s cycle in which the Fed ultimately raised rates by much more than was expected at any point along the way, though this is not our base case. GLOBAL FACTORS, EVIDENCE OF RECOVERY IN R* SHOULD RELIEVE INVERSION FEARS In the end however we believe that several factors are likely to prevent material outright curve inversion or unwind it relatively quickly if it occurs. An increase in the spot neutral rate (that we believe has already taken place) might lead market participants to revise higher their expectations of the longer-run neutral rate, at least on a probabilistic basis, even though the closing of the gap between spot neutral and longer-run neutral need not imply a higher longer-run neutral rate, simply convergence on that rate. More reliably, rising foreign yields associated with strong growth and policy transitions in the eurozone in particular but also possibly in Japan, may put upward pressure on the global common component of the term premium as well as plausibly contribute to a revision of views as to both the spot and longer-run neutral rate, which also have a material global common component. This would certainly be the case under the flow view of QE. But it may also be the case, at least to some degree, under the future expected stock view, with a decline in the probabilistic aggregate balance sheet path (little prospect of more, possibility of earlier reductions in the eurozone) at a time when in the base case Fed roll-off continues to increase and the global central bank balance sheet peaks in December against a backdrop of rising net issuance following US tax cuts, with a substantial swing into net transfer of issuance15 ahead in 2019. 10y govt bond yield spreads - US relative to Germany/Japan 2.6
UST-Bund Spread
UST-JGB Spread
2.5 2.4 2.3 2.2 2.1 2.0 1.9 1.8 1.7 Jan-17 Apr-17 Source: Bloomberg, Evercore ISI
15
Jul-17
Oct-17
Jan-18
And duration, holding constant the Treasury’s duration strategy, though this could be adjusted.
11
January 5, 2018
APPENDIX: MODELING RECESSION PROBABILITES WITH YIELDS There is a fair amount of economic literature on the predictive power of yield data in forecasting recessions. For an accessible overview, see Wright (2006). In this section we describe the modelling exercises we performed to illustrate the sensitivity around recession probabilities. The table below summarizes the results from a set of models over two different time periods: Jan 1961 – Dec 2016 and Jan 1983 – Dec 2016 (avoiding possible confounding relationships prior to Volcker disinflation). In each case, we set up a probit model using a 12-month head NBER recession indicator as the dependent variable. The independent variables vary in their attempts to capture different facets of the yield curve, inflation, and inflation expectations:
The spread between the 10-year and 3-month Treasuries The spread between the 10-year Treasury and the effective fed funds rate The level of the effective fed funds rate deflated by the 12-month trailing growth rate of core PCE inflation The basis point change over 12 months of the 10-year ACM risk-neutral rate, deflated by the 12-month trailing growth rate of core PCE inflation The basis point change over 12 months of the 10-year ACM risk-neutral rate, deflated by the Cleveland Fed’s 10-year inflation expectations (available only in the post-1983 sample).
Probit results Dependent variable: 12-month ahead NBER recession indicator (1 = recession) Absolute z-statistics in italics
Model 1 10Y-3M Spread
-0.716 9.76
Real Fed Funds Rate*
1961 - 2016 Monthly Model 2 Model 3 Model 4
Model 5
-0.553 6.73 0.141 3.44
10Y-Fed Funds Spread
Model 6
Model 7
-0.867 6.65
-0.900 5.46
0.055 1.20 -0.479 10.54
-0.435 7.52
12M Change in 10Y Real ACM Risk-Neutral*
1983 - 2016 Monthly Model 8 Model 9 Model 10 Model 11
-0.022 0.33 -0.491 10.32
-0.283 2.94 -0.862 6.46
-1.237 5.99
0.143 2.32
-1.071 6.44 -0.525 3.33
12M Change in 10Y Expected Real ACM Risk-Neutral** Constant McFadden R 2 AIC N Latest Recession Probability * deflated by year-on-year percent change in the core PCE deflator.
-1.305 6.62
-1.110 4.48 -0.449 5.07
-1.000 5.43
-0.975 12.96
-1.15 7.00
-0.96 12.61
-0.220 1.32
-0.126 0.39
-0.667 5.51
0.326 0.93
-0.724 5.66
-0.678 5.11
0.275 368.2 672
0.299 358.4 672
0.311 350.4 672
0.313 351.0 672
0.318 345.5 672
0.300 167.9 408
0.300 169.8 408
0.366 152.5 408
0.405 145.3 408
0.419 142.0 408
0.480 127.8 408
10.6%
4.8%
6.1%
4.6%
7.7%
11.7%
13.0%
4.5%
14.5%
0.8%
0.5%
** deflated by Cleveland Fed 10-year inflation expectations.
Fed funds rate uses Wu-Xia shadow rate when the ZLB binds. Source: FRBNY, Treasury, BEA, Wu-Xia, Evercore ISI.
The models yield several results of note 1. Both the 10Y-3M spread and the real fed funds rate are statistically significant in the specification that includes both variables over the full sample period (Model 2). But the significance of the real fed funds rate vanishes in the post-1982 period (Model 7). 2. In both periods, the models based on just the 10Y-fed funds spread (Models 3 and 8) outperform the models that include both the 10Y-3M spread and the real fed funds rate (Models 2 and 7). 12
January 5, 2018
3. Adding the real fed funds rate to the 10Y-fed funds models leads to virtually no improvement over the full sample (Model 4), but some improvement post-1982 (Model 9). 4. The best performing models of the ones we ran were those that included data on the change in the ACM risk-neutral yield. This term attempts to capture market expectations about shifts in the average policy rate over the next decade. 5. The model that added the ACM risk-neutral change deflated by core PCE (Model 5) had the most explanatory power of the full-sample models. 6. Deflating the change in ACM risk-neutral by Cleveland Fed 10-year inflation expectations improved the model even further in the post-1982 sample (Model 11). 7. Most of the models predict a recession probability below 10 per cent at the moment. Interestingly, the post-1982 models show higher probabilities when they do not include an ACM adjustment, but very low current probabilities (
[email protected]
Ernie Tedeschi 202-872-5262
[email protected]
THE FED AND THE YIELD CURVE
FOMC will not ultimately be deterred by risk of curve inversion; if yields stay low it might even do more
Summary: Going into 2018 there is intense market focus on the possibility that the yield curve could invert if the Fed presses ahead with ongoing rate hikes — and what this would mean in terms of recession risk. Minutes from the December FOMC meeting show policymakers grappling with the same issue, with the Committee divided between those who are worried by a flattening curve and risk of inversion and those who are less concerned. We believe that the Committee will keep a careful eye on the yield curve but will not ultimately be deterred by curve flattening from continuing to raise rates and will likely conclude that in current circumstances even outright inversion would not signal elevated recession risk. In brief we argue: ■
■
■
■
■
■
■
The December FOMC minutes released this week give prominent treatment to a debate over how to interpret the flattening of the yield curve and the possibility that an inverted curve could portend elevated recession risk We think the concern about curve flattening is centered among a subset of regional Fed presidents mostly with more market-oriented / non-academic economist backgrounds and is unlikely to be shared to the same degree by the core leadership of the Fed and its staff There is broad agreement that the current level of the term spread is not particularly alarming today: we run 11 versions of a slope-based recession forecasting model and these give four quarter recession probabilities of between 0.5 per cent and 14.5 per cent with a central estimate around 8 per cent The debate is about how worried the Fed should be if the flattening were to continue and the curve were to invert outright, which has traditionally been a fairly reliable signal of recession We think the Fed institutionally (FOMC leadership, Board staff) regards a narrow focus on the difference between short term and long term yields as a relatively crude way to assess recession risk and favors a richer evaluation of the yield curve including the information that can be distilled about the neutral rate and term premium The debate over how to respond to a flattening curve is an early example of how the Fed reaction function is becoming less predictable with heavy turnover on the Committee and a shift in composition from economists to non-economists; nonetheless we think the institutional Fed view will probably prevail Specifically we believe the Fed institutionally is sympathetic to the view — which we share — that the historic correlation between an inverted yield curve and recession probability is mostly a reflection of a more causal underlying relationship between the level of interest rates relative to neutral and the likelihood of recession
January 5, 2018
2
When the term premium is material and positive, as it was for most of the post-war period, the yield curve can only invert when interest rates rise well above levels that are expected to prevail on average in the future, which roughly correspond to the longer-run neutral rate, meaning that policy is likely quite restrictive But when the term premium is very low or negative, as it appears to be the case today, the yield curve could invert without interest rates rising above levels that are expected to prevail on average in the future, meaning that policy may not be restrictive at all There are a few caveats to this view: we cannot directly observe the term premium so we cannot be certain it is very low / negative; even if it is, curve inversion could still indicate that policy is quite restrictive if the spot neutral rate is below the longer-run neutral rate; and policy might be too restrictive from the perspective of managing inflation expectations / risks even if recession risk is not elevated Given the qualifications above the Fed will not dismiss the recession risk signal from a flattening curve but will cross-check against macro data, information from other asset markets and prices, and surveys including the Primary Dealer survey — to date all of these suggest the economy has strong momentum and near-term recession risk has declined to modest levels Indeed the macro, market and survey data suggest that the spot neutral rate has moved up and closed much of the gap with the longer-run neutral rate, providing a strong fundamental rationale for why the curve should have flattened, and providing less ground for concern in the event that the curve were to invert outright Further the macro, market and survey data suggest that the failure of longer term yields to move higher as the Fed has raised short term rates — the origin of the flattening dynamic — is being driven by global QE spillovers from the ECB and BoJ and not concern about the US outlook The influence of foreign QE and bond yields on the US term premium is a positive policy shock that should boost activity and near-term reduce recession risk If the yield curve flattens because longer-term yields are tightly pinned down by global factors then the Fed may continue to struggle to tighten financial conditions, and this could ultimately give the FOMC further cause to revise up its estimate of spot neutral and raise rates more rather than less to achieve an appropriate stance of policy In our base case however rising global yields (possibly accompanied by some mild upward bias to estimates of the longer-run neutral rate as spot neutral is revised higher) over the course of this year will bail the US out of an inversion scare
January 5, 2018
ATTENTION TO SLOPE OF YIELD CURVE IN FED MINUTES AMID INTENSE MARKET FOCUS1 Minutes from the December FOMC devote substantial space to a debate about the flattening of the yield curve in 2017, the possibility that the curve could invert outright in the period ahead, and what this might portend in terms of recession risk. This follows intense discussion in financial markets as to the risks around yield curve inversion, which has historically been a reasonably good predictor of recessions, with a number of regional Fed presidents including Kaplan,2 Kashkari,3 Harker4 and Bullard5 drawing attention to the importance of this signal in speeches and interviews. The minutes show FOMC participants “generally agreed that the current degree of flatness of the yield curve was not unusual by historical standards.” However “several” thought it would be important to continue to monitor the slope, while “some” expressed concern that a possible future inversion could signal a slowdown or pose a risk to bank profitability and financial stability. By contrast a “couple” viewed the flattening as the expected consequence of raising the federal funds rate and judged that a yield curve inversion under current circumstances “would not necessarily foreshadow or cause an economic downturn.” FOCUS ON THE SLOPE ITSELF IS AT ODDS WITH WHAT WE THINK OF AS THE MODERN FED INSTITUTIONAL VIEW AND COMES FROM OUTSIDE THE LEADERSHIP The attention to the slope of the curve itself is notable because it seems to us more reflective of a market view than what we believe to be the core Fed institutional view. We believe that the Fed institutionally (Board staff, FOMC leadership) regards narrow focus on the slope of the curve and prospect of inversion as a relatively crude indicator of recession risk. This is not to say the Fed ignores the evolution of the yield curve — quite the opposite — but rather to say that the Fed conducts a richer evaluation of the curve, drawing information from the level of current and future expected rates as well as the evolution of the term premium and what appears to be driving this, factors that are lumped together in the simple difference between longer term and short term rates. 1
Andrew Hipolit provided research assistance for this note. “Number one, I actually have been very concerned. So the Fed sets the fed-funds rate, right? And we know the balance sheet, quantitative easing had more effect on the curve, but we’re now unwinding that. So the fed-funds rate is the rate we set. I also look at the two-year, the five-year, the 10-year, the 30-year. And along with – (inaudible) – that as the Fed has raised rates, the 10-year has actually backed down. That’s a little ominous, cautionary for me. I don’t view that as necessarily an easing. I view that as a comment on future economic growth. And what I don’t want to see us do is raise rates so fast we create an inverted yield curve, because history has shown an inverted yield curve has tended to be a precursor to recession. So back to the gradual – (inaudible) – watching the 10-year hang around 2.30, 2.30-something [percent], I’m very cognizant of it because it suggests it limits the ability of the Fed to raise rates too rapidly. It certainly has an effect on the limit – limits to what level we can get it to if you want to avoid a flat or inverted yield curve. That’s number one.” — Robert Kaplan (October 10, 2017), Q&A following appearance at Stanford University. 3 “It’s also possible that we might do it ourselves on the front end. By prematurely raising rates, it’s inverting the yield curve. I understand that the term premium is low now, and so maybe this time is different. But we shouldn’t just dismiss the signals that the bond market is sending us. So it’s possible that – if we continue down this path and inflation expectations don’t come up, it’s possible that we could end the expansion by our own actions.” — Neel Kashkari (December 18, 2017), interview with The Wall Street Journal. 4 “But there are risks to running low on inflation for too long, and we know those. And that’s where we – to me, we should prudently – move rates up to the neutral rate, but prudently. We don’t want to run any risk of disrupting markets, no risk of inverting the yield curve, you know.” — Patrick Harker (October 17, 2017), interview with The Wall Street Journal. 5 James Bullard (December 1, 2017), “Assessing the Risk of Yield Curve Inversion.” 2
3
January 5, 2018
Indeed our best guess — taking the unattributed minutes alongside the public remarks of individual FOMC participants — is that concern about yield curve flattening is coming from outside the core Committee leadership, mostly from more market-oriented non-academic economists who are less deeply imbued with Fed institutional thinking and policy analysis during the 2000s Greenspan “conundrum” period, with the exception of Bullard for reasons we will discuss below. We suspect that the institutional perspective will prevail and ultimately redirect the discussion from the slope and risk of inversion to broader analysis of the yield curve and specific focus on the stance of policy relative to the neutral rate along with the reasons behind a low term premium. But there is some residual uncertainty, part of a larger theme as the Committee is reshaped, loses institutional experience and rebalances from economists towards a larger share of non-economists. AT PRESENT ALL AGREE THAT EVEN AFTER FLATTENING THE YIELD CURVE IS NOT AT PRESENT SIGNALING ELEVATED RECESSION RISK While a flattening yield curve could be an early sign of recession, it is also mechanically exactly what one would expect to see as short-term policy rates move close and closer to longer-term neutral (holding constant any term premium dynamics). Further, while the Treasury curve has flattened, it has not inverted, which has historically been the more reliable indicator of an imminent recession. None of the almost-dozen probit models we ran, varying in sample size and specification, show 12-month ahead recession probabilities above 15 per cent at the moment, and some that include information from the ACM term structure model show probabilities below 1 per cent (see the Appendix for more details on our exercise). And these models draw purely from Treasury yield and inflation data; other financial and real indicators that could be included in a richer model are still running at paces consistent with an expansion. That is not to say that these models counsel complacency: recession probabilities in yield-based models tend to rise suddenly and precipitously, not gradually and with ample warning. On the other hand, while this class of model tends to perform reasonably well in anticipating actual recession, it also produces false positives, as can be seen in the chart below. Still, these models are in broad agreement that recent yield curve behavior is not historically-consistent with a high risk of imminent recession.
Recession probability: 12m ahead 100%
Model 1
Model 2 80%
Model 3 Model 4
60%
Model 5 Model 6
40%
Model 7 Model 8
20%
Model 9 Model 10
0% 1983
Model 11 1987
Source: Evercore ISI.
4
1991
1995
1999
2003
2007
2011
2015
January 5, 2018
RECESSION SIGNAL FROM INVERTED CURVE REFLECTS UNDERLYING RELATIONSHIP BETWEEN CURRENT SHORT RATES AND THE NEUTRAL RATE The issue is rather how concerned the Fed should be if the flattening were to continue potentially to the point at which the curve inverts outright. We believe the institutional Fed view is likely similar to our own — that the historic relationship between an inverted yield curve and recession probability is mostly a reflection of a more causal underlying relationship between the level of short term rates relative to neutral and the likelihood of recession in the period ahead. We can think of the ten year yield as composed of an expected path of future short term rates and a term premium. The expected path of future short term rates ought generally to correspond to the ex ante expected neutral rate over the longer run.6 Ordinarily (though not during the post-crisis period) we would assume that the spot neutral rate is reasonably close to the longer run neutral rate. Therefore when current short term interest rates exceed the average expected future short term rate, the stance of Fed policy is likely to be restrictive. In periods with a material positive term premium (most of the post-War period prior to the financial crisis) the yield curve can only invert when short term interest rates have moved well above the average expected future short term rate7 which is normally a reasonable proxy for the spot neutral rate. In such circumstances on the face of it at least, the Fed would be in very restrictive territory. And a very restrictive Fed would be likely to lead to output falling below potential — a recession. This is why even pre-crisis Fed analysis of the yield curve pointed to the importance of incorporating the current level of the federal funds rate in analysis of recession risk, not just the slope itself.8 YIELD CURVE INVERSION MEANS SOMETHING DIFFERENT IN PERIODS WITH A LOW OR NEGATIVE TERM PREMIUM However as Yellen argued in the December FOMC press conference, in periods when the term premium is low (as in the mid-2000s) or negative (as appears to be the case today9) a very flat or even inverted yield curve may well not convey the same information it does when the term premium is material and positive. Very obviously, the yield curve will naturally invert more easily and more often than when the term premium is materially positive. With little or no term premium, the curve can invert without current short term rates moving materially above expected future short term rates. Indeed with a negative term premium the curve can invert without current short term rates moving above expected future short term rates at all. If as argued above the causal relationship is not between the slope of the curve and the likelihood of recession, but between the stance of policy relative to neutral (proxied by the longer run average short term rate) and the likelihood of recession, then we ought not to worry about curve inversion as long as current short rates remain around or even below their expected longer-run average values. This we think is the Yellen - Dudley view, the staff view, and probably the view that Powell and the Committee as a whole will also embrace, subject to three caveats that require ongoing cross-checks against other market and macro data.
6
Inflation can complicate this assessment. Neutral is strictly a real not nominal concept. Further, the Fed might be seen as seeking to raise or lower the trend rate of inflation on a forward looking basis. 7 Because the longer term yield is equal to the sum of the expected future short term rates and the term premium. 8 See e.g., Jonathan Wright (2006), “The Yield Curve and Predicting Recessions.” 9 Based on decompositions using affine term structure models such as the Adrian Crump Moench (ACM) model. 5
January 5, 2018
CAVEAT NUMBER ONE: WE CANNOT DIRECTLY OBSERVE THE TERM PREMIUM SO WE CANNOT BE CERTAIN THAT IT REALLY IS VERY LOW / NEGATIVE As Bullard argued recently10 dismissing the signal from a very flat or inverted yield curve on the basis that this does not have its historic signaling value in a world with a very low or negative term premium could be risky given that it is impossible to directly observe the term premium. The Fed’s assessment of the term premium is based on a series of affine term structure models such as the Adrian Crump Moench (ACM) model that are regarded as state of the art but are nonetheless sensitive to certain modeling assumptions. We therefore cannot rule out the possibility that the term premium is in fact still materially positive and curve inversion would mean what it has historically meant — that the current short term rate has moved well above the rate that is expected to prevail on average in the future and as such is likely to be quite restrictive relative to neutral. This caution is reinforced by the fact that there is less statistical support than might be expected for the logical proposition that what matters is the relationship between short term rates and the average of future expected short rates as a proxy for neutral; in simple models historically the spread based on the total yield outperforms the spread relative to the risk-neutral rate that summarizes the expected rate path in the ACM model; see the appendix for more detail. This suggests that there may be a cyclical component in the term premium that has additional information regarding the likelihood of a near-term recession, though this is only one of a number of different influences on the term premium. ACM 10y bond yield decomposition
ACM 10y bond yield decomposition 16%
ACM 10y Fitted Yield
ACM 10y Term Premium
ACM 10y Risk-Neutral Yield
14%
3.0%
12%
2.5%
10%
2.0%
8%
1.5%
6%
1.0%
4%
0.5%
2%
0.0%
0%
-0.5%
-2% 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 Source: FRBNY
US 10y-3mo govt bond yield spread - historic 6 5
4 3
2 1
0 -1 -2 -3 -4 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 Source: Bloomberg, Evercore ISI
10
6
3.5%
ACM 10y Fitted Yield
-1.0% 2013 Source: FRBNY
2014
ACM 10y Term Premium
2015
2016
ACM 10y Risk-Neutral Yield
2017
2018
US 10y-3mo govt bond yield spread - since start 2013 3.0 2.5 2.0 1.5 1.0 0.5 0.0 2013 2014 Source: Bloomberg, Evercore ISI
2015
James Bullard (December 1, 2017), “Assessing the Risk of Yield Curve Inversion.”
2016
2017
2018
January 5, 2018
CAVEAT NUMBER TWO: EVEN WITH A NEGATIVE TERM PREMIUM CURVE INVERSION COULD INDICATE THAT POLICY IS RESTRICTIVE IF SPOT NEUTRAL REMAINS BELOW THE LONGER-RUN NEUTRAL RATE If the spot neutral rate is below the longer-run neutral rate — as the Fed argued over the past half-decade and arguably reiterated even in the December statement11 — then curve inversion could indicate that policy has become materially restrictive even with a negative term premium. This is because the comparison between the current level of the federal funds rate and its expected longer-run average value would not be a reliable proxy for the difference between the actual current rate and the neutral rate, particularly if inflation is low today but expected to average a higher value closer to the Fed’s 2 per cent target over the longer run.12 CAVEAT NUMBER THREE: EVEN IF THE FED CAN DISMISS INVERSION BECAUSE THE TERM PREMIUM IS LOW / NEGATIVE IT WOULD STILL NEED TO THINK WHETHER THIS & LOW INFLATION EXPECTATIONS MEANT THAT POLICY WAS TOO RESTRICTIVE Aiming off the slope of the curve itself as a reliable indicator of recession risk on the basis that the term premium is very low or negative would still require the Fed to think deeply about why the term premium is so low, and whether this might still indicate that policy was too tight. As mentioned above, there appears to be a cyclical component in the term premium historically that may have some informational power as to the likelihood of recession in the coming quarters. However, the term premium might also be low for other reasons (structural, US / global QE) and we think that is indeed the case today (see below). Separately, inflation break-evens are below historically normal levels, which might suggest that policy is inappropriately tight less from a recession risk perspective than by way of reinforcing below-target inflation expectations. Model-based decompositions — subject as before to critiques based on their specific properties — suggest the weakness is concentrated in the inflation risk premium, which appears to be negative.13 While this might in principle be associated with nearterm recession risk, in our view the negative inflation risk premium reflects the desire to hedge against the high potential cost of a return to the zero bound in the next downturn rather than elevated likelihood of a recession over the next year or two. This could still imply that policy under the current framework is suboptimal / too tight from a welfare-maximizing risk management perspective. At the same time, a negative inflation risk premium might be better addressed by adopting a different framework (ideally accompanied by non-monetary policy actions) rather than a different rate path; we will write more on this in the coming days.
11
The statement retains the longstanding guidance that “the federal funds rate is likely to remain for some time below levels that are expected to prevail in the longer run” — though this phrase is ambiguous as it does not specify whether the spot neutral rate is likely to remain for some time below the levels that are likely to prevail in the longerrun, or whether policy is likely to remain accommodative for some time. 12 To give a simple concrete example, if the average future short term rate is 3 per cent, reflecting an ex ante longerrun neutral rate of 1 per cent real plus 2 per cent longer run average inflation, but the spot neutral rate is zero real and current inflation is 1.5 per cent, then even with a -25 bp term premium, inversion would occur with the funds rate firmly in restrictive territory at 125 bp above neutral. 13 Again, subject to model-based decompositions. 7
January 5, 2018
THE FED WILL CROSS-CHECK THE SIGNAL FROM THE YIELD CURVE AGAINST THE SIGNAL FROM OTHER ASSET MARKETS, MACRO DATA AND SURVEYS Given the complications around how to interpret the signal from a curve that flattens further (as it did over the course of 2017) or threatens to invert (as it might do in the period ahead if the market reprices for additional Fed rate hikes near term) and the importance of understanding why the term premium is so low, the FOMC will evaluate the signal in the context of the state of the economy, developments in other financial markets, evidence from surveys of economists, and potential influences on US yields other than the outlook for the US economy. In doing so the Committee will be both seeking to assess whether there is any validation for the notion of elevated recession risk and how it should think about the evolution of the key underlying variables: the neutral rate and the term premium. MACRO DATA, SURVEYS SUGGEST US ECONOMY HAS GAINED MOMENTUM AND IS IN LITTLE DANGER OF STALLING We see — and we believe the Fed will see — no validation in the macro data to date of any suggestion that the US economy is in elevated danger of stalling or slipping into recession. Quite the opposite: the US economy appeared to step up a gear in 2H 2017 and is exhibiting considerable momentum in the New Year as reflected in PMIs and our own EVRISI surveys. A strong labor market with some prospect of moderately higher wage gains going forward, slight household tax relief and record net worth (albeit very unevenly distributed), strengthened incentives to invest post tax cuts and immediate expensing with reduced downside risk to global growth and moderation in the dollar, and decent prospects for residential investment all suggest that this momentum will be sustained through 2018. Indeed there is some upside risk to the median December SEP projection that growth will reach 2.4 per cent this year, reflecting in part the Fed’s cautious conventional modeling of the impact of tax cuts.
8
January 5, 2018
US ISM manufacturing & nonmanufacturing PMIs
US PMIs 50+=expansion 60
Composite
Manufacturing
50+=expansion
Services
Manufacturing
62
Nonmanufacturing
60 58
55
56 54 52
50
50
48 45 Jan-15 May-15 Sep-15 Source: IHS Markit
Jan-16 May-16 Sep-16
Jan-17 May-17 Sep-17
46 Jan-15 May-15 Sep-15 Jan-16 May-16 Sep-16 Jan-17 May-17 Sep-17 Source: ISM
ISI Company Surveys Retail, Trucking, Shopping Guide, Banks, Credit Card, Tech, Manufacturing, Cap Goods, Auto Dealers, Airlines 0=Weak, 100=Strong 55 54 53 52 51 50 49 48 47 Jan-15 May-15 Sep-15 Jan-16 May-16 Sep-16 Jan-17 May-17 Sep-17 Source: Evercore ISI
CROSS-CHECK WITH OTHER FINANCIAL MARKET INDICATORS ALSO SUGGESTS RECESSION RISK HAS FALLEN Moreover a cross-check against other asset markets suggests — to date at least — much more support for the idea that US recession risk has fallen than the idea that is has risen as the yield curve has flattened. The credit risk premium is very tight, earnings forecasts are moving up, and the (unadjusted) equity risk premium — while not extremely low by historic standards — has edged down over the past year. The dollar has weakened on net over the past year, but this appears to have been driven more by stronger growth ex-US and the surprise shortfall of US inflation than reconsideration of US growth prospects / recession risk. BofA Merrill Lynch US high yield bond index - option adjusted spread Basis points
Bloomberg US equity risk premium in 2017 7.5%
425
7.4% 7.3%
400
7.2% 7.1%
7.0%
375
6.9%
6.8%
350
6.7% 6.6%
325 Jan-17
Apr-17
Source: BofA / Merrill Lynch
9
Jul-17
Oct-17
Jan-18
6.5% Jan-17 Source: Bloomberg
Apr-17
Jul-17
Oct-17
January 5, 2018
PRIMARY DEALER SURVEY ALSO SUGGESTS RECESSION RISK HAS FALLEN AND IS NOT ELEVATED Consistent with the macro data, surveys including the New York Fed Primary Dealer survey indicate that private sector economists have lowered their estimate of recession risk as the yield curve has flattened to date, and see only moderate recession risk over the next few years in spite of the length of the expansion and the low level of remaining slack conventionally measured. FRBNY Survey of Primary Dealers - what percent chance do you attach to the US economy being in a recession in 6 months? By survey date
25%
25th Percentile
Median
75th Percentile
20% 15% 10% 5% 0% Jan-15 May-15 Sep-15 Source: FRBNY
Jan-16 May-16 Sep-16
Jan-17 May-17 Sep-17
INSTEAD MACRO AND MARKET DATA SUPPORT VIEW THAT THE SPOT NEUTRAL RATE HAS MOVED UP Indeed we read the macro, market and survey data as indicating not only that recession risk is indeed low but also that the spot neutral rate has moved up as the global economy has entered a more synchronized expansion with reduced left skew to the macro distribution and associated endogenous easing of financial conditions, and is now reasonably close to its longer run estimated value. This has been a core part of our thesis since late summer of 2017 and it underpins our view that the Fed has more headroom to raise rates than is widely appreciated without moving into restrictive territory. In the context of the yield curve, this would mean that the Fed need not be concerned about mild curve inversion under a very low / negative term premium as spot neutral would be close to longer-run neutral. AND THE TERM PREMIUM IS BEING HELD DOWN BY GLOBAL QE SPILLOVERS, WHICH COULD FORCE THE FED TO DO MORE TIGHTENING NOT LESS Meanwhile we believe the macro, market and survey data also supports the view that factors other than the US economic outlook are likely depressing the term premium, in particular spillovers from ECB and BoJ QE in integrated global capital markets as well as more underlying structural factors. From the perspective of the US, global QE spillovers constitutes a positive policy shock that ceteris paribus would tend to boost aggregate demand, lower recession risk and increase upside risks.14
14
Of course this is a general equilibrium problem, and the stance of policy ex-US also reflects economic conditions ex-US, so what really matters here is the global policy stance relative to economic conditions ex-US. 10
January 5, 2018
Indeed at some limit if the Fed is unable to push longer term yields higher or otherwise tighten financial conditions (for instance, via renewed appreciation of the dollar) it might reasonably conclude that the spot neutral rate has moved up further and it needs to raise rates further to achieve an appropriate stance relative to neutral. This could lead to a replay of the 2000s cycle in which the Fed ultimately raised rates by much more than was expected at any point along the way, though this is not our base case. GLOBAL FACTORS, EVIDENCE OF RECOVERY IN R* SHOULD RELIEVE INVERSION FEARS In the end however we believe that several factors are likely to prevent material outright curve inversion or unwind it relatively quickly if it occurs. An increase in the spot neutral rate (that we believe has already taken place) might lead market participants to revise higher their expectations of the longer-run neutral rate, at least on a probabilistic basis, even though the closing of the gap between spot neutral and longer-run neutral need not imply a higher longer-run neutral rate, simply convergence on that rate. More reliably, rising foreign yields associated with strong growth and policy transitions in the eurozone in particular but also possibly in Japan, may put upward pressure on the global common component of the term premium as well as plausibly contribute to a revision of views as to both the spot and longer-run neutral rate, which also have a material global common component. This would certainly be the case under the flow view of QE. But it may also be the case, at least to some degree, under the future expected stock view, with a decline in the probabilistic aggregate balance sheet path (little prospect of more, possibility of earlier reductions in the eurozone) at a time when in the base case Fed roll-off continues to increase and the global central bank balance sheet peaks in December against a backdrop of rising net issuance following US tax cuts, with a substantial swing into net transfer of issuance15 ahead in 2019. 10y govt bond yield spreads - US relative to Germany/Japan 2.6
UST-Bund Spread
UST-JGB Spread
2.5 2.4 2.3 2.2 2.1 2.0 1.9 1.8 1.7 Jan-17 Apr-17 Source: Bloomberg, Evercore ISI
15
Jul-17
Oct-17
Jan-18
And duration, holding constant the Treasury’s duration strategy, though this could be adjusted.
11
January 5, 2018
APPENDIX: MODELING RECESSION PROBABILITES WITH YIELDS There is a fair amount of economic literature on the predictive power of yield data in forecasting recessions. For an accessible overview, see Wright (2006). In this section we describe the modelling exercises we performed to illustrate the sensitivity around recession probabilities. The table below summarizes the results from a set of models over two different time periods: Jan 1961 – Dec 2016 and Jan 1983 – Dec 2016 (avoiding possible confounding relationships prior to Volcker disinflation). In each case, we set up a probit model using a 12-month head NBER recession indicator as the dependent variable. The independent variables vary in their attempts to capture different facets of the yield curve, inflation, and inflation expectations:
The spread between the 10-year and 3-month Treasuries The spread between the 10-year Treasury and the effective fed funds rate The level of the effective fed funds rate deflated by the 12-month trailing growth rate of core PCE inflation The basis point change over 12 months of the 10-year ACM risk-neutral rate, deflated by the 12-month trailing growth rate of core PCE inflation The basis point change over 12 months of the 10-year ACM risk-neutral rate, deflated by the Cleveland Fed’s 10-year inflation expectations (available only in the post-1983 sample).
Probit results Dependent variable: 12-month ahead NBER recession indicator (1 = recession) Absolute z-statistics in italics
Model 1 10Y-3M Spread
-0.716 9.76
Real Fed Funds Rate*
1961 - 2016 Monthly Model 2 Model 3 Model 4
Model 5
-0.553 6.73 0.141 3.44
10Y-Fed Funds Spread
Model 6
Model 7
-0.867 6.65
-0.900 5.46
0.055 1.20 -0.479 10.54
-0.435 7.52
12M Change in 10Y Real ACM Risk-Neutral*
1983 - 2016 Monthly Model 8 Model 9 Model 10 Model 11
-0.022 0.33 -0.491 10.32
-0.283 2.94 -0.862 6.46
-1.237 5.99
0.143 2.32
-1.071 6.44 -0.525 3.33
12M Change in 10Y Expected Real ACM Risk-Neutral** Constant McFadden R 2 AIC N Latest Recession Probability * deflated by year-on-year percent change in the core PCE deflator.
-1.305 6.62
-1.110 4.48 -0.449 5.07
-1.000 5.43
-0.975 12.96
-1.15 7.00
-0.96 12.61
-0.220 1.32
-0.126 0.39
-0.667 5.51
0.326 0.93
-0.724 5.66
-0.678 5.11
0.275 368.2 672
0.299 358.4 672
0.311 350.4 672
0.313 351.0 672
0.318 345.5 672
0.300 167.9 408
0.300 169.8 408
0.366 152.5 408
0.405 145.3 408
0.419 142.0 408
0.480 127.8 408
10.6%
4.8%
6.1%
4.6%
7.7%
11.7%
13.0%
4.5%
14.5%
0.8%
0.5%
** deflated by Cleveland Fed 10-year inflation expectations.
Fed funds rate uses Wu-Xia shadow rate when the ZLB binds. Source: FRBNY, Treasury, BEA, Wu-Xia, Evercore ISI.
The models yield several results of note 1. Both the 10Y-3M spread and the real fed funds rate are statistically significant in the specification that includes both variables over the full sample period (Model 2). But the significance of the real fed funds rate vanishes in the post-1982 period (Model 7). 2. In both periods, the models based on just the 10Y-fed funds spread (Models 3 and 8) outperform the models that include both the 10Y-3M spread and the real fed funds rate (Models 2 and 7). 12
January 5, 2018
3. Adding the real fed funds rate to the 10Y-fed funds models leads to virtually no improvement over the full sample (Model 4), but some improvement post-1982 (Model 9). 4. The best performing models of the ones we ran were those that included data on the change in the ACM risk-neutral yield. This term attempts to capture market expectations about shifts in the average policy rate over the next decade. 5. The model that added the ACM risk-neutral change deflated by core PCE (Model 5) had the most explanatory power of the full-sample models. 6. Deflating the change in ACM risk-neutral by Cleveland Fed 10-year inflation expectations improved the model even further in the post-1982 sample (Model 11). 7. Most of the models predict a recession probability below 10 per cent at the moment. Interestingly, the post-1982 models show higher probabilities when they do not include an ACM adjustment, but very low current probabilities (
