Macro and Credit - Aleatoricism

"An investment in knowledge pays the best interest" - Benjamin Franklin

Like any good behavioral psychologist, we watched with interest the debt ceiling theatrical drama unfold positively, not being surprised of the outcome given we tend to focus on the process rather than on the content. Given we have seen the Fed consistently lower their dot plot forecasts to meet market expectations, when it came to selecting our title analogy for this week's musing we decided to go for "Aleatoricism". Aleatoricism is the incorporation of chance into the process of creation, especially in the creation of art or media, or central banking one would opine. The word derives from "alea", the rolling of a dice. Aleatoric methods have been used in artistic composition for thousands of years and were popularized in the early 20th century by the Dada movement. Leonardo Da Vinci once said:

"Look at walls spotted with various stains or with a mixture of different kinds of stones. If you are about to invent some scenes, you will be able to see in it a resemblance to various different landscapes adorned with mountains, rivers, rocks, trees, plains, with valleys and various groups of hills."

In similar fashion, one could look at the Fed's dot plot and with Fischer resigning wonder if indeed they do matter anymore, or are just an artistic creation given  markets are acting like the Fed will only hikes approximately once between now and Dec 2018 when their Dot Plot showed 4 hikes:

- source Bloomberg

The median forecast still had the central bank making three quarter-point increases in 2018; the end-2019 rate was seen at 2.9 percent, a slight change from 3% from their March projections. One might wonder if indeed we have a case of "Aleatoricism". Interest-rate projections for 2018 and 2019 are becoming less reliable guides to future policy with the likelihood of a complete revamp of the Fed’s Board of Governors next year. With the Fed officially entering its blackout period, where officials no longer make public appearances or grant interviews, one thing we are certain is that the US yield curve is immune to their "Jedi tricks". We continue to hold a flattening stance, and while we have missed the opportunity to reach more the long end of the curve we will happily do so and increase our duration exposure accordingly. After all, New York Fed president William Dudley took a less hawkish stance than previously on Friday when he admitted that the call on further rate hikes was up in the air. This is in our book is clearly "the rolling of a dice" but we ramble again...

In this week's conversation, we would like to look at the predictability of the US Yield curve in forecasting US recession, without having it inverted as the economic and credit cycle matures. 

Synopsis:

• Macro - Yielding to the predictability of the US Yield curve
• Credit - US High Yield - Beware The Ides Of September
• Final chart - Is the "Buck" breaking bad?

• Macro - Yielding to the predictability of the US Yield curve

As we pointed out in our August conversation "Gullibility", the US yield curve has been immune to the old tricks played by the Jedis at the Fed. While the Fed's dot plot might be a case of "Aleatoricism, it certainly isn't the case when it comes to using the US Yield curve as a recession predictor, with its ongoing flattening stance. Same goes with credit curves. In our book a flattening of the US High Yield curve (we use the US CDX High Yield CDS Index implied curve as a proxy), is a predictor for growing pressure on credit, which is generally a leading indicator for equities as shown in late 2015 before the early sell-off of 2016. For instance, as of late, we have been closely watching with interest the rise in the front-end of the curve as displayed by ICE CMA:

- source ICE CMA

No need yet to panic on the above, but it is certainly worth monitoring going forward, as another indicator you need to track. The next CDS roll for CDS indices will mark the launch of new credit index series (series 28 in Europe and series 29 in the US) and the introduction of new "on-the-run" single reference contracts (bearing a December 2022 maturity rather than the current June 2022 "on-the-run" maturity) and is only a couple of weeks away as a reminder.

When it comes to the US Yield curve being a recession predictor, its reliability over time is certainly not a case of "Aleatoricism". What is of interest to us all, is the recent question asked recently by Wells Fargo on the subject in their note from the 8th of September entitled "Do We Need to Wait for a Yield Curve Inversion to Predict a Recession? No":

"Others have seen what is and asked why. I have seen what could be and asked why not." – Pablo Picasso

Be Mindful of Elevated Recession Risks in 2018-2019

Predicting recessions is one of the most important elements of decision-making in the public and private sector. As such, a different set of policy tools is needed during a recession than that used for an economic expansion. The yield curve (spread between the 10-year Treasury and federal funds rate, for example), in particular the inversion point of the yield curve, is thought to be a very good predictor of a recession. An inverted yield curve has led all recessions since the 1969-70 recession. Furthermore, the Federal Open Market Committee (FOMC) has raised the federal funds target rate (fed funds rate) twice in 2017, which has brought the inverted yield curve topic (and the impending risk of a near-term recession) back into the spotlight. Other analysts are raising questions surrounding the yield curve’s effectiveness in predicting recessions. Presently, the fed funds rate is “recovering” from a historical low level (zero-lower bound) and, as such, the yield curve may not invert in this cycle as it has in the past. That is, the yield curve stayed in the positive territory (did not invert) during the 1954-1965 period and that era experienced two recessions (1957-1958 and 1960-1961 recessions).2

In this report, we propose a new framework that identifies a threshold between the fed funds rate and the 10-year Treasury yield (we call it FFR/10-year threshold). In a rising fed funds rate environment, the threshold is breached when the fed funds rate touches/crosses the lowest level of the 10-year Treasury yield in that cycle. When this occurs, the risk of a recession in the near future is significant. Our framework has successfully predicted all recessions since 1955 with an average lead time of 17 months. Furthermore, our framework predicted several recessions before the yield curve inversion point and, therefore, serves as a more effective tool in predicting recessions. That is, with our framework, we do not need to wait for the yield curve to invert to predict a recession.

Why is our analysis important for decision-makers? In the current monetary cycle, the lowest 10-year Treasury yield was 1.36 percent (hit on July 5, 2016) and the current fed funds rate is 1.25 percent. We are forecasting one more rate hike by the FOMC (December 2017), and, if we are correct, the fed funds rate will be 1.50 percent, thereby surpassing the lowest level of the 10-year Treasury (1.36 percent) and thus breaching the threshold. Historically, when the threshold is met, there is 69.2 percent chance (average probability) of a recession within the next 17 months  (average lead time). Therefore, in the December rate hike scenario, the chances of a recession in 2018 through mid-2019 are elevated.

Summing up, our proposed framework (FFR/10-yr threshold) produced 13 signals since 1955 and 9 of the 13 signals are associated with recessions (there are only 9 recessions in that time period, thus, we did not miss any recessions) with an average lead time of 17 months. The remaining four signals are connected with changes in the monetary policy stance (moving from rising fed funds to cutting interest rates) with an average lead time of 8 months. Typically, during long economic expansions, monetary policy will shift due to a “mid-cycle softening” and the FOMC will reduce interest rates to boost the economy. This phenomenon occurred in the 1960s, 1980s and 1990s. Therefore, our proposed framework did not produce a single misleading signal (neither false positive or false negative). Given the robust performance of our framework, we suggest decision makers to watch for a recession during 2018 and mid-2019 (17 months from December 2017, in the case of a rate hike).

The Inverted Yield Curve: A Song of Policy Tightening and Recessions

The yield curve (spread between the 10-year Treasury yield and fed funds rate) is one of the most cited recession predictors. In fact, the yield curve is the part of the Conference Board’s index of leading economic indicators (LEI). In particular, the inverted yield curve has led the last seven recessions (all recessions since the 1969-70 recession), Figure 1.

That is, the yield curve inverted before each of the last seven recessions (although with a wide range of 8-23 months lead time).

Most studies in the past have utilized the spread between the 10-year Treasury yield and fed funds rate as the yield curve, and we followed that tradition (for more detail see Adrian et al. (2010)). There are several benefits of using the yield curve based on the 10-year Treasury and fed funds rate. First, Bernanke and Blinder (1992) utilized the nominal level of the fed funds rate to measure the monetary policy stance, and thereby we include the policy stance in the yield curve. Second, moments in the 10-year yield reflect financial market participants’ expectations about the economic outlook and the monetary policy stance.

For example, typically, investors seek higher yield in an inflationary era and take refuge in Treasuries when fears of a recession rise (lower yield), all else equal. Therefore, our measure of the yield curve represents both the policy stance and market expectations.

Third, Adrian and Estrella (2009) utilized the fed funds rate to identify monetary cycles, which are good predictors of recessions and another competitor of recession prediction methods. The fourth and final method, our FFR/10-year threshold framework, also utilizes the fed funds rate and 10-year yield to predict recessions. Therefore, all three methods of recession prediction utilize the fed funds rate and 10-year yield and, thereby, we can compare performances of these methods to find which approach is most effective in predicting recessions.

Typically, an inverted yield curve suggests that the financial markets are not very optimistic about the near-term economic outlook and seek “safety” of their investment by buying 10-year Treasuries, which creates higher 10-year Treasury demand and reduces the yield, all else constant. The fed funds rate, on the other hand, is set by the FOMC and, usually, the FOMC takes time (lag effect) to change the stance of monetary policy (transitioning from a dovish tone to hawkish one). The lag response from the FOMC is because the FOMC utilizes realized data along with its forecast to set the monetary policy stance. Therefore, in the case of an inverted yield curve, the 10-year yield drops below the fed funds level because changes in the economic outlook influence both financial market expectations and monetary policy.

Is This Time Different for the Inverted Yield Curve to Predict Recessions?

As mentioned earlier, the inverted yield curve has predicted the last seven recessions but missed the 1957-1958 and 1960-1961 recessions. That is, the yield curve remained positive (did not hit the inversion point) during the 1954-1965 period. Furthermore, the misses associated with the 1957- 1958 and 1960-1961 recessions (false negative) raises questions about the yield curve’s effectiveness in predicting the next recession. This begs the questions: are there some potential factors which may prevent the yield curve to invert in this cycle similar to the 1950s/mid-1960s episodes? Likewise, is there an alternative method for recession prediction that is more effective than the yield curve?

We believe that the yield curve’s effectiveness to predict the next recession may be different than the last seven recessions and may repeat the 1957-1958 and 1960-1961 recessions’ experiences. That is, the yield curve may not invert and, thereby, be useless in predicting the next recession. There are several major reasons to support our view. First, the fed funds rate is recovering from the lowest level (from 0-0.25 percent range) in our analysis, which covers the January 1954- July 2017 time period. Furthermore, the fed funds hit the 0-0.25 percent range on December 2008 and that was the lowest level since July 1958 (0.68 percent). In addition, before December 2008 there are only two episodes of below 1 percent fed funds rate and both of them occurred in the 1950s (several months in 1954 and a few months of 1958 observed below 1 percent fed funds rate) and the yield curve did not invert in the 1950s (1954-1959). Since both the 10-year yield and fed funds rate remained positive in our sample period (1954-2017), a historically lower level of the fed funds rate may pose a hurdle for the yield curve to invert. For instance, the 10-year yield dropped below 2 percent for the first time (in our sample period) on September 2011 (1.98 percent) and never dropped below 1 percent in our analysis. The fed funds rate, on the other hand, hit the zero-lower bound on December 2008. By the same token, in the 1954-1965 period, the lowest fed funds rate was 0.63 percent (May 1958) and the lowest 10-year yield was reported as 2.29 percent (April 1954). It is important to note that, in the pre-December 2008 era, the lowest level of the fed funds rate and 10-year yield were reported in the 1950s and the yield curve did not invert during the 1954-65 period (and missed two recessions). Therefore, the recovery from the historically low-level of the fed funds rate may alter the yield curve’s effectiveness in predicting the next recession compared to the last seven recessions—we may not see an inverted yield curve before the next recession.

Second, the current and near-term economic outlook, in particular inflation expectations, may not be ideal for a faster pace of monetary policy tightening (faster fed funds rate hikes, for example). That is, the dual mandate (maximum employment and price stability) of the FOMC dictates that officials consider the labor market/inflation expectations (in addition to other factors) in setting the monetary policy stance. For example, Taylor (1999) suggested that the fed funds rate responds to economic variables (unemployment and inflation rate, for example), even in the periods when the FOMC was targeting some other variables (i.e. money growth and/or reserves targets). Furthermore, Romer and Romer (2002) concluded that inflation expectations, typically, play a crucial role in setting the monetary policy stance. Therefore, the economic outlook may influence the fed funds rate in a way to slow any pace of federal funds rate.

Echoes From the 1950s 

In our view, the current economic outlook, in particular realized and expected inflation, is more in line with the 1954-1965 period than the last seven recessions (1969-2007 period). For example, the FOMC’s inflation target is 2 percent (the PCE deflator is the preferred inflation measure of the Fed) and the PCE deflator is just 1.22 percent for the May 2012-June 2017 period (Figure 3).

Furthermore, the PCE deflator (year-over-year [YoY] percent change) stayed below 2 percent for the May 2012 to June 2017 period (with the exception of two months: January/February 2017) and it is the longest stretch in the past 50 years. In addition, before 2012, the last time the PCE deflator stayed below 2 percent for over five years was between January 1960 and January 1966 (slightly more than six years). It is worth mentioning that the PCE deflator series only goes back to 1960, and we utilize CPI (YoY) for the pre-1960 period. The CPI (YoY) was below 2 percent for the November 1952 and July 1956 period (with an average of just 0.41 percent) and between November 1958 and January 1966 (with an average of 1.28 percent). The possible consequences of the lower inflation of the 1950s and early 1960s is the lower fed funds rate as the rate dropped below 1 percent, the first time ever in our analysis, in 1954 and then in 1958. On the other hand, the PCE deflator hit the 2 percent target on February 1966 and then stayed above the 2 percent line for the next 30 years, and the fed funds rate never dropped below 3 percent during that time period.

Silvia, Iqbal and Bullard (2016) identified low-inflation episodes (where a low-inflation episode is defined as six consecutive months of PCE deflator below 2 percent) using the 1975-2016 time period. Out of eight total episodes identified between 1975 and 2016, three of them occurred in the 2008 2016 period. Between November 2008 and July 2017 (last data point at the time of this writing), the PCE deflator was below 2 percent for 85 of 105 total months (with an average of just 1.33 percent). Therefore, inflation rates since the Great Recession are at historical lows, on average. The unemployment rate, the other focus of the FOMC, took almost nine years to drop below the pre-Great Recession era (Figure 3 above). That is, the unemployment rate dropped to 4.3 percent in May 2017, just below the 4.4 percent rate registered in May 2007. Historically low inflation rates, along with a slow decline in the unemployment rate, are associated with a historically low fed funds rate (dropped to zero-lower bound on December 2008 and stayed there for the next six years) and several rounds of quantitative easing, which boosted the Fed’s balance sheet to around $4.5 trillion (a historically high level). All of these factors, in our view, are supportive of a slower pace of fed funds rate hikes in the near future. The FOMC has also announced its intentions to reduce its balance sheet starting in 2017. Furthermore, a lower fed funds rate was unable to invert the yield curve in the 1954-1965 period, although that period experienced two recessions. Therefore, we need to look for methods, other than the simple yield curve, to accurately predict the possibility of the next recession.

A Probit Model Using the Yield Curve to Predict Recessions

One method that utilizes the yield curve to predict recessions involves building a probit model and employing the yield curve as the predictor variable of the model (Figure 4).

The probit model (or any standard econometric model) needs a longer data history for estimation purpose. Therefore, we started forecasting recessions since 1980 (use 1960-79 period for estimation). The probit model has predicted (50 percent probability as a threshold) all recessions since 1980 except the 1990-91 recession. Furthermore, due to the lack of data for the pre-1950 period, we are unable to check the probit model’s accuracy regarding the 1957-1958 and 1960-1961 recessions (these recessions were missed by the traditionally calculated inverted yield curve). Thus, the probit model may not be the answer. Therefore, we move forward to utilize another tool, monetary cycles, to try to predict recessions.

- source Wells Fargo

It seems apparently that not only the Fed's Phillips Curve model is outdated, but waiting for an inversion of the yield curve to predict a recession going forward might not be working this time around. 

Could the monetary cycles be more useful in predicting the next recession and avoid "Aleatoricism"? Wells Fargo in their very interesting notes try to give us some clues on this very subject:

"The Game of Monetary Cycles: Is Recession Coming?

Adrian and Estrella (2009) identify monetary cycles and suggest that monetary cycles are good predictors of economic activities, as these cycles are typically associated with business cycles.11 Adrian and Estrella (2009) concluded that there have been 14 monetary cycles since 1955 and that a monetary cycle ends when the fed funds rate peaks. Moreover, the end of a monetary cycle is a prediction for an upcoming recession. According to the NBER, however, there have been nine recessions since 1955, which indicates that not all monetary cycles are associated with recessions as there were 14 monetary cycles in the same time period, see Table 1 and 3 for results.

For example, the first monetary cycle ended on October 1957 but the recession start date is August 1957 (missed by two months). Similarly, January 1980 is the recession start date but the monetary cycle end date is April 1980 (missed by three months). Monetary cycles predicted the rest of the recessions with a lead time range of 1-16 months. Although monetary cycles have predicted several recessions (and missed the 1957-1958 and 1980 recessions), the real time effectiveness of the monetary cycles is a big question mark. Why? In our view, the monetary cycle approach to predict recessions is a backward-looking method, and that is because of the way a monetary cycle is defined. For example, Adrian and Estrella (2009) said that a monetary cycle ends when either one of the two criteria is met: (1) the fed funds rate is higher than at any time from the 12 months before to nine months after and is at least 50 bps higher than at the beginning of this period, or (2) the fed funds rate is higher than at any time from six months before to six months after and is 150 bps higher than the average at these endpoints. Basically, the peaking of the fed funds rate is the ending of a monetary cycle; in other words, the peaking of the fed funds rate is a prediction for the upcoming recession.

However, life is not that simple as we have to wait for at least six months (2nd criterion) to confirm whether the fed funds rate has peaked six months ago. Why does that matter? It matters and, to some extent, changes (reduces) the real time effectiveness of monetary cycles to predict recessions completely. For example, the first monetary cycle end date is October 1957 and that cycle missed the recession by two months (the recession start date is August 1957). However, in reality, we have to wait for at least six months (using the 2nd criterion, which has a shorter waiting period) to determine the fed funds rate’s peak point. That is, in April 1958 we were able to declare that the October 1957 was the peak month for the fed funds rate (end of a monetary cycle) and then we can make a recession prediction. But, April 1958 is also the end date of the recession, and therefore, in real time analysis, the monetary cycles missed the 1957-1958 recession completely. What we are suggesting is that the monetary cycles method’s lead time to predict recessions should be longer than the waiting period of 6-9 months to gain financial benefits (predict recessions) from this approach. For example, using real time analysis, monetary cycles have missed all of the recessions in the 1955-1989 period as the lead time for all these recessions is less than six months (Table 1). Monetary cycles did predict the 1990-1991 and the Great Recession as the lead time was 16 and 15 months, respectively. The lead time for the 2001 recession was eight months and that could be considered a miss using the first criterion to define a monetary cycle (at least nine months are needed to confirm the fed funds rate peak).

Therefore, in our view, monetary cycles are effective recession predictors, but the method is backward-looking. Furthermore, in real time analysis, monetary cycles are only able to predict recessions in the post-1990 era. We need a method that effectively predicts recessions in real time so decision-makers have enough leeway to prepare policies for the upcoming recession. We will now discuss the final method, our proposed method. We believe this approach is more effective in predicting recession than the other methods discussed so far in this study.

In similar fashion to the monetary cycles, looking at default rates is backward-looking when trying to assess the state of the credit cycle. Looking at the deterioration of financial conditions through the prism of the credit impulse as well as the Fed Quarterly Senior Loan Officer Opinion Survey (SLOOs) appears to us much more relevant:

- Lending conditions - source Bank of America Merrill Lynch

US C&I Loans last three months: 31st of August 0.7, 31st of July 0.2, 30th of June 0.6. Since the beginning of the year and relative to 2016, the trend has been weakening hence our more cautious tone we have been having relative to the credit cycle.

If the Norwegian Blue Parrot aka the Phillips Curve is simply "resting" and if waiting for the US Yield Curve to invert would be akin to waiting for Godot, then again you might be wondering what kind of framework would enable you to predict an upcoming recession. Wells Fargo in their very interesting note is coming up with an interesting proposal we think:

"A New Framework to Predict Recessions: The FFR/10-Year Threshold

We have discussed the limitations of the yield curve (inverted yield curve and the probit/yield curve modeling) and monetary cycles’ approaches to predict recessions. In addition, one major challenge in this monetary cycle is that the fed funds rate’s recovery from the lowest level, along with a low inflation environment, may block inversion of the yield curve similar to the 1954-1965 period. Therefore, we need a framework that is more effective in real time recession forecasting than the yield curve/monetary cycle methods. The framework should also be able to predict recessions accurately in different economic regimes such as lower inflation/fed funds rate regimes (the 1954-1965 period and the era since the Great Recession, for example), higher fed funds rate/inflation (the 1970 to mid-1980s time period) and in the moderate inflation/fed funds rate time periods (the 1990-2007 time period, for instance). We believe our proposed framework would predict recessions accurately in all those economic regimes.

Our framework identifies a threshold between the fed funds rate (FFR) and the 10-year Treasury yield (10-year). The crossing of the threshold is an indicator for an upcoming recession. We labeled the framework as the FFR/10-year threshold to predict recessions. The threshold is best explained by the following description: in a rising fed funds rate period, when the fed funds rate crosses/touches the lowest level of the 10-year yield in that cycle, then that is a prediction for an upcoming recession. For example, the Fed started raising the fed funds rate in December 1954 (fed funds rate increased from 0.83 percent to 1.28 percent) and the 10-year yield hit 2.61 percent (lowest level in that cycle) on January 1955. Furthermore, the fed funds rate crossed the lowest level of the 10-year yield on April 1956 (fed funds rate was 2.62 percent) and, thereby, signaled an upcoming recession. The recession start date is August 1957 (a 16-month lead time for our framework’s prediction), (Table 2). It is worth mentioning that both the yield curve and monetary cycles’ methods were unable to predict the 1957-58 recession.

Before we discuss the effectiveness of the FFR/10-year threshold, we raise a few questions to highlight the intuition behind the threshold method. Why is the rising fed funds rate the starting period for the threshold method? Why is the lowest level of the 10-year Treasury yield in a cycle matter? Why is the threshold (FFR crossing/touching the lowest 10-yr) approach a good recession predictor? The rising fed funds rate, outside recessions, is a sign of a change in the monetary policy stance and, typically, the Fed starts raising interest rates when the economy enters expansion. Naturally, a recession comes after an expansion phase and therefore a rising fed funds rate environment is a better policy stance to utilize in recession predictions, which is the objective of the threshold framework. The 1980 recession is an exception as the next recession (the 1981-1982 recession) started within a year of the ending month of the 1980 recession. Therefore, a rising fed funds rate represents a change in the monetary policy stance and the FOMC’s expectations about the strength (expansionary phase of a business cycle) of the economy.

By the same token, the 10-year yield’s lowest level in a cycle serves as an inflection point in the market’s expectations about the economic outlook and monetary policy stance. That is, market participants are not looking for a refuge/safety in Treasuries, which reduces the Treasuries’ demand and consequently a rise in the yield, all else equal. Furthermore, financial markets are also expecting a better economic outlook (perhaps the beginning of an expansionary phase) and a change in the policy stance (rising fed funds rate) in the near future. Basically, both policy makers and market participants are expecting a better economic outlook/expansion phase and, therefore, the rising fed funds rate and lowest 10-year yield level are inflection points and help to predict recessions.

Looking Beyond the Absolute Length of an Expansion 

While it is true that the end of an expansion phase is the beginning of a recession, the length of expansions vary significantly as the longest expansion in our analysis lasted for 10 years. The current expansion is the third longest at the time of this writing, September 2017. Therefore, attempting to predict a recession based on an expansion’s start date and current length is not a useful exercise. Moreover, we are looking for a threshold that will help us to predict recession, in real time, in a timely matter. For example, the threshold for the yield curve approach is the inversion point and peaking of the fed funds rate is a benchmark for the monetary cycles. However, the lower fed funds rate along with a low inflation rate may prevent an occurrence of an inversion point in this cycle similar to the 1954-1965 period. We have to wait for 6-9 months to declare a peak month of the fed funds rate to predict recessions in the monetary cycles’ analysis and the waiting period reduces the effectiveness to predict recessions significantly, as mentioned earlier. Therefore, our proposed threshold of FFR crossing/touching the lowest 10-year point is effective in real time recession prediction. There are several reasons for this unique ability that will be discussed.

First, a lower fed funds rate may prevent a yield curve to invert, but our method does not incorporate an inversion point in predicting recessions. Second, there is no waiting period to declare whether the threshold has been met, unlike the monetary cycle method, which is largely backwards looking. Moreover, we can also predict the possible timing of the threshold point, which we show in a latter part of the study. Third, we do not impose a specific value of the 10-year yield as a benchmark (2 percent 10-year as a threshold, for example) because different economic regimes (higher or lower inflation and/or stronger or weaker recoveries, for instance) would produce different lows/highs of fed funds rates and 10-year yields in a business cycle. Therefore, using the cycle low yield for the 10-year, accounts for the heterogeneity of business cycles. Another reason for not using a fixed level for either the 10-year yield or the fed funds rate as a threshold is that the effect of a rising fed funds rate on 10-year yield varies depending on the cycles. For example, the FOMC raised the fed funds rate from 1.00 percent to 5.25 percent during the June 2004-June 2006 period and the 10-year increased only by 37 bps (from 4.73 percent to 5.11 percent) during the same time period; Greenspan labeled it as “interest rate conundrum.” In sum, the accuracy of our proposed framework would not be affected by the fact that the fed funds rate is recovering from the zero-lower bound, or by a low inflation environment or by the fact that the relationship between the fed funds rate and 10-year has changed overtime.

The final and fourth reason is that the FOMC can raise the fed funds rate up to a certain level and, typically, when the fed funds rate peaks, that is an indication that the expansion is close to its peak, and a recession is in the neighborhood. Furthermore, historically, (Table 2 and 3) when the fed funds rate crosses/touches the lowest level of the 10-year, in a monetary cycle, that is an indication that fed funds rate’s peak is approaching. Therefore, meeting the threshold is a prediction for the upcoming recession.

Now we discuss the accuracy of our proposed method, with the results reported in Table 2. Since 1955, our framework predicted all recessions with an average lead time of 17 months, with a range of 6-34 months. It is important to note that our method is the only approach discussed in this study that did not miss any recessions in the sample period. This means that it is more effective than the yield curve and monetary cycle approaches. Furthermore, our framework has a better lead time than the yield curve to predict recessions for all recessions except the 1969-1970 and 1981-1982 recessions where both approaches have the same lead time, Table 2.

The Exception to the Rule: Recessions versus Changes in the Monetary Policy Stance

Sometimes there is an exception to the rule, and thereby our framework produces four calls that are not associated with recessions. It is worth mentioning that our framework is the only approach (discussed in this study) that did not miss any recession since 1955, whereas the yield curve missed two recessions and produced three calls, which were not related with recessions, Table 2 and 3. Similarly, the monetary cycle method missed two recessions and five of its signals are without a recession, Table 1 and 3. Does that mean that our framework produces false positives? In our view, the answer is no. Although four of the 13 calls are not associated with recessions, those four calls are connected with changes in the monetary policy stance (from raising/unchanged fed funds rate to cutting interest rates). For example, the framework produced a recession call on December 1964 (threshold met) and the Fed started reducing the fed funds rate on December 1966 (24-month lead time). The Fed reduced the rate from 5.76 percent to 3.79 percent between November 1966 and July 1967, Table 3. Similarly, the remaining three calls  (August 1984, February 1995 and July 1998) are followed by changes in the monetary policy stance, see Table 3 for details.

Put differently, four of the 13 calls are associated with a change in the monetary policy stance with an average of eight months lead time—with a range of 1-24 months. Another reason not to declare these four calls as false positives is that, during long economic expansions, the Fed has reduced interest rates to “boost” the economy from a “mid-cycle slowdown.” Furthermore, the 1960s, 1980s and 1990s experienced some of the longest expansions on record and, thereby, changes of monetary policy stance during those expansions.

Summing up, our framework has produced 13 recession calls since 1955, and nine of those are associated with recessions. Therefore, whenever our framework produced a recession call, there was a 69.2 percent (9/13) chance of a recession within the next 17 months (average lead time).

Why Our Analysis Matters for Decision-Makers?

The FOMC raised the fed funds rate on December 2015, the first time in the post-Great Recession era, so the first condition of our framework is fulfilled—in a rising fed funds rate environment. The 10-year yield hit 1.36 percent on July 5, 2016, which is the lowest level in this cycle, Figure 5. Therefore, two conditions of the threshold framework are accomplished. The current level of the fed funds rate is 1.25 percent, which is lower than the 1.36 percent 10-year yield.15 As mentioned earlier, our proposed framework is not only good in a real time analysis but also is a forward-looking approach. That is, we are forecasting one more rate hike by the FOMC in 2017 (December), and, if that happens, then the fed funds rate will be 1.50 percent in December 2017. Therefore, in the case of one more rate hike, the threshold will be met in December 2017 as the cycle low for the 10-year is 1.36 percent (lowest daily closing yield on July 5, 2016, and 1.50 percent for the monthly average of July 2016, still the lowest level in this cycle). Therefore, starting in December 2017, again in the case of a rate hike, there will be a 69.2 percent chance of a recession during the next 17 months (average lead time).

Final Thoughts: Be Mindful of Elevated Recession Risks

We have proposed a new framework using the fed funds rate and the 10-year yield to predict recessions. Our framework has predicted all recessions since 1955 with an average lead time of 17 months. Furthermore, we are forecasting one more rate hike in 2017 (December), and, in the case of a rate hike, the threshold will be met. Therefore, starting December 2017, there is an increasingly elevated probability of a recession in the coming years. It is important to note that, at present, our official call is for continuously moderate growth in 2018-2019 (around 2.5 percent GDP growth rate) and this framework suggests a downside risk to our forecast. Therefore, we are not making an official call for a recession over the two-year forecast horizon. Instead, decision-makers may want to watch 2018 through mid-2019 for potential slowdown/recession. But, be mindful of the analysis we have performed in this report, we will be watching incoming data closely to determine whether conditions that could lead to a recession/slowdown starting late 2018 are developing. We would encourage decision-makers to do so as well." - source Wells Fargo

If indeed, the Fed goes for another hike in December, then again the recessionary path could be starting from that point at least from a high probability threshold and not due to "Aleatoricism". The Wells Fargo analysis is interesting in the sense that it shows that macroeconomic frameworks need to be reassessed and one must not take for granted that what has been working, will continue to work  the same like the Phillips Curve though the Phillips Curve debate is still raging.

As we move towards Autumn, the real season that is, not the credit cycle season in which we believe we are, there are indeed interesting seasonality factors for credit markets and in particular US High Yield of interest as per our next credit bullet point.

• Credit - US High Yield - Beware The Ides Of September

Summer might be over as we move towards autumn, but, when it comes to assessing the credit cycle, we clearly think we are way part Summer and most likely in Autumn from our perspective. On the subject of seasonality and credit, we read with interest Barclays US High Yield note from the 8th of September entitled "Autumn Years":

"Seasonality in market conditions is often tracked as a way to measure the extent to which technicals drive returns. September, in particular, poses seasonal headwinds for high yield investors. As shown in Figure 1, it is the only month that has produced an average negative return and has ranked among the top quartile of best-performing months in only four times out of the total 33 observations considered.

This trend has persisted in recent years, with September the worst-performing month in 2014 and 2015.

While we are skeptical of any investment strategy that relies purely on return seasonality, we believe the drivers of the calendar effect can provide a useful guidepost. We first note that one significant factor has been a seasonal uptick in September gross issuance, which has, in turn, weighed on returns for the month. Figure 2 shows the month-over-month increase in supply is highest in September, with supply more than doubling following a slow August.

We do not expect an increase in supply of the same magnitude this year, considering that high yield markets priced roughly $20bn in August despite a litany of volatility-driving headlines. But a sizable forward high yield calendar and the tight pricing of recent deals suggest roughly $25-30bn in September supply. Furthermore we expect to see more opportunistic issuance considering currently attractive financing yields. Our expectation is that new supply will be well absorbed and that refinancing will account for a bulk of the calendar through year-end; but beyond September, we do not believe incremental supply will produce a drag on returns.

We find that the increased supply in September is often not met y increased demand from retail investors. Fund flow seasonality shows that retail sentiment has historically been softer leading up to September, with funds registering outflows in August followed by only modest inflows in September (Figure 4).

That said, US high yield valuations have remained largely intact despite the $12bn in year-to-date outflows registered through last week. Therefore, we are not overly concerned about a seasonality driven weakening of retail sentiment, but recognize that it could be a factor that prevents the market from tightening in the face of higher supply. In the event that volatility spikes meaningfully and outflows increase, we expect this to be counterbalanced by less opportunistic issuance." - source Barclays

Despite Labor day, and the seasonality factor pointed out in Barclays note, we do agree that flows matter and particularly in High Yield with the retail feeble crowd playing the beta game through ETF exposure. We note that last week inflows to US High Yield jumped to $1.1bn following the previous week's $258mio gain, the most since July and the third largest inflow year to date according to Bank of America Merrill Lynch US High Yield Flow Report from the 7th of September. Yet, year-to-date total returns were down a bit during August, but to a still-impressive 6.1% through August 31. Within developed markets, this is the 2nd best performance and only trails the S&P’s +11.9% return. The beta game is still the trade du jour for many. We do think that at this stage of the credit cycle, you should continue moving higher the quality spectrum and on that note we agree with Bank of America Merrill Lynch's take from their High Yield Strategy note from the 6th of September entitled "Labor Day letdown in credit":

"Where does this leave us?

Moving forward, we remain cautious on high yield. Although earnings growth remains strong and issuer fundamentals continue to improve, in our opinion the macro landscape remains concerning. With US and North Korea relations heading in the wrong direction, the Fed on a normalization path, and all eyes on Washington as it faces a multitude of uphill tasks for the remainder of the year, we think too many risks still persist. Consider the amount of spread investors earn per turn of leverage in HY, which declined from 85bps to 83bps last quarter (Chart 2).

This is below the historical median of 111bps and suggests that investors are currently taking on more risk for lesser compensation. However, high yield spreads have proven resilient in the face of macro uncertainty for the first half of the year. We could therefore envision a scenario where the FOMO (fear of missing out) attitude extends into 2018, with high yield clipping coupon through December. But not before enduring a few speed bumps along the way." - source Bank of America Merrill Lynch.

Sure there is still plenty of punch being served at the credit party but it doesn't mean you need to get yourself drunk beyond reason by playing the beta game a little bit too long we think. What matters once again is the velocity of credit, the rate of change of credit availability.

• Final chart - Is the "Buck" breaking bad?

While our contrarian stance served us well in early 2017 as per our conversation "The Woozle effect" given we told you that we were not part of the long US dollar crowd at the time:

"If indeed the US administration is serious on getting a tough stance on global trade then obviously, this will be bullish gold but the big Woozle effect is that it will be as well negative on the US dollar

It appears that from a "Mack the Knife" perspective, it will be rather binary, either we are right and the consensus is wrong thanks to the Woozle effect, or we are wrong and then there is much more acute pain coming for Emerging Markets, should the US dollar continue its stratospheric run. From a contrarian perspective we are willing to play on the outlier."  - source Macronomics, January 2017

Of course we were right...But the most important question is about the US dollar aka the "Buck" and if it is on course to "break bad". Our final chart comes from Nomura FX Insights note from the 6th of September entitled "The dollar's seven-year feast-famine cycle" and displays the dollar cycle:

It’s all too easy to get caught up in the short-term dynamics of currency markets and miss the big picture. Alternatively, if one did see the big picture, it’s too easy to erase the pre-2008 period as the “old normal” and believe that the financial markets are now driven by new forces. This ahistorical perspective forgets that every decade is characterised by some new paradigm that impels us to believe the old rules no longer apply. Today the new paradigm is secular stagnation or balance sheet constrained growth models. In the 2000s, it was the disruption caused by China, the 1990s saw the dot-com boom, the 1980s saw Reaganomics, and the 1970s saw stagflation.

Through all of this the US dollar has tended to follow a cycle where it enjoys many years of gains followed by many years of declines. On average the dollar trends tend to last about seven years (Figures 1). The number seven seems almost pre-ordained. The Bible talks about seven years of plenty and seven years of famine (Genesis 7:2). There are the seven chakras in various Indian traditions. Then we have the seven-day week, the seven metals of antiquity and the seven endocrine glands. It’s perhaps easy to see why chartists in currency markets have been known to use the moon cycle or sunspots in forecasting the dollar.

Admittedly, seven is an average. Dollar trends can be longer or shorter, but importantly there are fundamental causes for this dollar cycle. The crucial point is that the business cycle which most investors focus on tends to be much shorter than the dollar cycle. The combined expansion and recession phases of the US have typically lasted six years, while the full dollar cycle – both the uptrend and downtrend – has lasted 14 years. That means business cycle or monetary policy explanations alone will not be adequate in predicting dollar trends. Instead, forces such as relative price levels, terms of trade, current account dynamics and capital flows need to be included to provide the context for multi-year turns in the dollar. Taken together, they indicate that the dollar is currently at the early stages of a multi-year decline. This could see the euro eventually reach 1.40 or higher and USD/JPY fall towards 90 or lower."  - source Nomura

We don't know if it as case of seven years of bad luck for the US dollar or an imaginary curse or another illustration of "Aleatoricism", but we do know that when it comes to the US Yield Curve and the credit cycles it does follow patterns much more precise than the Fed's dot plot, that's a given no matter how good their "Jedi" tricks.

"Luck is a matter of preparation meeting opportunity." - Lucius Annaeus Seneca

Stay tuned !