Informe especial
Highlights Butterfly Strategy: A butterfly fixed income strategy is a combination of a barbell (a weighted combination of long- and short-term bonds) and a bullet (the medium-term bonds that sit within the yield curve segment selected in the barbell) designed to provide investors exposure to specific yield curve changes while being insulated from parallel shifts. Yield Curve Models: Simple yield curve models, based on the positive relationship between the slope of the yield curve and butterfly spreads – and to a certain extent, implied interest rate volatility – can be used to identify which part of the yield curve is most attractively valued by comparing what change in the slope is being discounted with our own macro views. Current Valuation: The overall message from our new suite of global yield curve models is that trades favoring barbells over bullets are attractive across all the developed market countries covered in our analysis. Feature In February 2002, BCA Global Fixed Income Strategy (GFIS) introduced a framework for measuring market expectations for changes in short-term interest rates embedded in the slope of government bond yield curves.1 By comparing those discounted changes with our own macro view on where rates were headed, this framework provided signals on potential value in trades focusing on the shape of the yield curve. This analysis originally focused on one specific yield curve (butterfly) strategy across six developed markets; the US, Germany, the UK, Japan, Canada, and Australia. Table 1Most Attractive Butterfly Trades More recently, our sister service US Bond Strategy applied this framework to each different butterfly spread combination across the entire US Treasury curve, creating a tool to identify the most attractively valued parts of the US yield curve at any point in time.2 In this Special Report, we revisit the original GFIS methodology for identifying attractive yield curve trades in global government bond markets. Furthermore we extend the analysis to all butterfly combinations and add three additional European countries to the list - France, Italy and Spain. The overall message is that trades that favor barbells over bullets are attractive across all the developed markets covered in this analysis. Table 1 displays the most attractive combinations of barbells over bullets for each country. Going forward, we will rely on the readings from our refreshed yield curve models, combined with our macro views, to populate our new Tactical Trade Overlay framework with yield curve trades in global government bond markets. What Is A Butterfly Strategy? A butterfly fixed income strategy involves two main components: a barbell (a weighted combination of long-term and short-term bonds) and a bullet (a medium-term bond that sits within the yield curve segment selected in the barbell). This strategy owes its name to the resemblance that barbells and bullets can have with the wings and body of an actual butterfly, not to lepidopterology.3 To implement a butterfly strategy, a bond investor would go long (short) the barbell while simultaneously going short (long) the bullet. In general, barbells are expected to outperform bullets in a flattening yield curve environment, and vice-versa. The reason butterfly strategies are so widely used is that they provide fixed-income investors exposure to specific changes in the slope of the yield curve, while being neutral to small parallel shifts. This immunization to small parallel shifts is achieved by setting the weights of the short- and long-term bonds in the barbell such that the weighted sum of their dollar duration (referred to as DV01 – the dollar value of a basis point) equals the DV01 of the bullet. In the event of large parallel shifts in the yield curve – which are quite rare – the barbell will outperform the bullet since the former will always have a greater convexity than the latter in the absence of convexity-matching between each leg of the trade. We illustrate how a 2/5/10 butterfly strategy works for US Treasuries, using hypothetical constant-maturity par bond yields, in Table 2A.4 Table 2AThe Butterfly (Strategy) Effect Illustrated As can be seen in the ”Weighted DV01” column of Table 2A, the DV01 of each leg of the trade (the bullet and the two combined bonds in the barbell) are identical. Importantly, the weighted DV01 contribution to the barbell from the 2-year note and the 10-year bond differ substantially, meaning that the barbell is more sensitive to changes in the 10-year yield than changes in the 2-year yield. This mismatch is precisely what gives a butterfly strategy exposure to the slope of the curve. Table 2A also presents three yield curve scenarios to demonstrate the benefits of butterfly strategies. In the parallel shift scenario, yields across the entire yield curve rise by 10bps. This parallel shift is neutralized as the two legs of the strategy cancel out. In the steepening curve scenario, the 2-year yield falls by 10bps, the 10-year yield rises by 10bps and the 5-year yield remains flat. In this case, the small gains on the 2-year note cannot offset the losses on the 10-year bond; hence the barbell underperforms the 7-year bullet. Finally, the “Flattening” column in the table shows that the barbell outperforms the bullet when the curve flattens. Our government bond yield curve models rely on the positive relationship typically observed between the butterfly spread and the slope of the yield curve. Bottom Line: A butterfly fixed income strategy is a combination of a barbell (a weighted combination of long- and short-term bonds) and a bullet (the medium-term bonds that sit within the yield curve segment selected in the barbell) designed to provide investors exposure to specific yield curve changes while being insulated from parallel shifts. Dusting Off The GFIS Yield Curve Models Chart 1Butterfly Spreads & Yield Curves Our government bond yield curve models rely on the positive relationship typically observed between the butterfly spread and the slope of the yield curve. When the curve steepens, the butterfly spread widens, and vice-versa (Chart 1). This has to do with mean reversion: as the curve steepens, it increases the odds that the curve will flatten in the future since it cannot steepen indefinitely. Consequently, investors will ask for greater compensation to enter a curve steepener trade when the curve is already steepening. As a result, we can create simplified models of the yield curve by regressing any butterfly spread on its corresponding curve slope. Deviations from these fair value models indicate which butterfly strategies are cheap or expensive. While positive, the correlations between yield curve slopes and butterfly spreads vary widely across butterfly combinations and also among countries – in Japan, for example, the historical relationship seems dubious (Chart 1, panel 4). We can further improve the fit of some of our yield curve models by including the MOVE US bond volatility index as a second independent variable. As our colleagues at US Bond Strategy have pointed out, implied interest rate volatility is also positively correlated with the slope of the yield curve (Chart 2, top panel). This matters for butterfly trades because of the convexity mismatch between the barbell and the bullet, particularly given the fact that high convexity is beneficial when implied interest rate volatility is elevated. Simply put, a larger convexity mismatch between the two legs makes them more sensitive to changes in the slope of the curve, and therefore easier to model (Chart 2, bottom panel). Importantly, one other useful application of the relationship between yield curve slopes and butterfly spreads is that we can reverse the yield curve models to calculate what amount of curve steepening or flattening is being discounted in current butterfly spreads. In other words, our models allow us to calculate change in the curve slope that would force the butterfly spread to be equal to its fair value (Chart 3). Chart 2Taking Into Account Implied Vol Armed with that information, we can then apply our macro views to determine potential butterfly spread trades. Chart 3Case In Point: US 2/5/10 Spread Fair Value Model For example, the 2/5/10 butterfly spread in the US (the 5-year bullet yield minus the weighted combination of 2-year and 10-year yields) is, at the moment, below its fair value with 46bps of steepening discounted over the next six months (Chart 3, panels 2 & 3). That means the bullet is expensive as per our model and therefore the recommended butterfly strategy would be to go long the 2/10 barbell and short the bullet. However, in the event the 2/10 Treasury slope steepens by more than 46bps over the next six months, the 5-year bullet would be expected to outperform the barbell. In other words, when the butterfly is initially below its fair value, more curve steepening will be needed for the bullet to outperform the barbell. Conversely, if it is above fair value, more curve flattening will be required for the barbell to outperform. In light of this, let’s consider the example of curve steepening from before, but this time looking at two scenarios: the butterfly spread is at fair value the butterfly spread is initially different from its model-implied fair value, but is then expected to revert to fair value by the end of the investment horizon. Under the first scenario, the bullet outperforms the barbell when the curve steepens, as expected given that the butterfly spread is at fair value (Table 2B). Now, in the second scenario, the bullet actually ends up underperforming the barbell, although it is the same curve steepening environment. Table 2BButterfly Strategy Performance And Deviations From Model-Implied Fair Values The reason for this underperformance is that the butterfly spread is now below the fair value shown in scenario #1, thus requiring more steepening for the bullet to outperform the barbell. Ultimately, we have to rely on our macro view of how the slope of the yield curve will change alongside the message from our yield curve models to choose the right butterfly strategy. This means that, ultimately, we have to rely on our macro view of how the slope of the yield curve will change alongside the message from our yield curve models to choose the right butterfly strategy. Bottom Line: Simple yield curve models, based on the positive relationship between the slope of the yield curve and butterfly spreads – and to a certain extent, implied interest rate volatility – can be used to identify which part of the yield curve is most attractively valued by comparing what change in the slope is being discounted with our own macro views. The Message From Our Butterfly Strategy Valuations In the remaining pages of this Special Report, we present the current read-outs from of our yield curve models for each of the major developed market. More specifically, we provide the deviations from fair value for different combinations of bullets and barbells and highlight the most attractive butterfly strategy. The deviations from fair value shown in Tables 3-11 are standardized to facilitate comparison between the different butterfly combinations. Also, for each country we provide a quick assessment of the performance of these butterfly strategies over time by applying a simple mechanical trading rule. Every month, we enter the most attractive butterfly strategy, i.e. the one with the highest absolute standardized deviation from its model fair value. The overall message is that barbells appear attractive relative to bullets across all the countries shown. Trades that favor barbells over bullets are attractive across all the developed markets covered in our analysis. This is consistent with our near term macro view. Global government bond markets have been experiencing bull flattening pressures ever since the COVID-19 virus outbreak sparked a generalized flight-to-safety. Markets woke up to the recent news about the spread of the virus in countries outside of China – namely Italy, South Korea, Japan, Iran and Israel – and all traces of complacency have now vanished.5 There is too much uncertainty about COVID-19 in terms of severity and duration, and government bond yields may very well continue falling until the threat is contained. In the meantime, this may force major central banks to provide even easier monetary policy. While this may be difficult for the ECB and the BoJ, which both already seem out of ammunition, the other central banks could very well end up delivering the rate cuts currently discounted in the overnight index swap curves.6 Looking back at our Central Bank Discounters, the largest amount of rate cuts over the next year are now discounted in the US (-53bps), now discounted in the US (-53bps), Australia (-38bps), Canada (-37bps) and the UK (-23bps). At the same time, the fewest cuts are priced in Japan (-8bps), the euro area (-6bps) and New Zealand (-25bps). The resulting bull steepening would likely be mild, however; after all, rate cuts cannot fight a pandemic, but can only try and cushion the blow to growth. In the event COVID-19 virus does not turn into a pandemic and we observe a decline in the daily change of the number of cases, then global government bond yields would rebound from their current lows. Given the current valuation cushion, we would expect barbelled portfolios to do well, especially since we would not expect more steepening than what is currently being discounted (i.e. we do not expect the 2/30 Treasury slope to steepen by more than 73bps in the near term). Jeremie Peloso Senior Analyst jeremiep@bcaresearch.com US There are presently three butterfly combinations standing out in that they appear attractive according to our yield curve model. One of them is going long the 2/30 barbell and shorting the 10-year bullet, which currently displays a standardized residual of -1.42 (Table 3). Table 3US: Butterfly Strategy Valuation: Standardized Residuals The bullet appears 21bps expensive according to our model and would only outperform its counterpart given a steepening in the 2/30 Treasury slope greater than 73bps, which we view as unlikely given the current environment (Chart 4A). Chart 4AUS: 2/10/30 Spread Fair Value Model Chart 4BUS Butterfly Strategy Performance Following the mechanical trading rule looks promising (Chart 4B). In fact, we observe few periods of negative year-over-year returns. Germany The most attractively valued butterfly combination currently on the German yield curve is going long the 2/30 barbell and shorting the 10-year bullet, which is currently a little bit more than one standard deviation above its implied-model fair value, with a standardized residual of -1.09 (Table 4). Table 4Germany: Butterfly Strategy Valuation: Standardized Residuals The bullet appears 14bps expensive according to our model and would only outperform its counterpart given a steepening in the 2/30 German curve slope greater than 36bps (Chart 5A). Chart 5AGermany: 2/10/30 Spread Fair Value Model Chart 5BGerman Butterfly Strategy Performance Over time, picking the cheapest butterfly combinations based on our yield curve models works relatively well (Chart 5B). Importantly, we observe very few episodes of underperformance since 1990. France The most attractively valued butterfly combination currently on the French OAT yield curve is going long the 5/30 barbell and shorting the 10-year bullet, which currently displays a standardized residual of -1.13 (Table 5). Table 5France: Butterfly Strategy Valuation: Standardized Residuals The 10-year bullet appears 11bps expensive according to our model and would only outperform its counterpart given a steepening in the 5/30 French OAT curve slope greater than 44bps (Chart 6A). Chart 6AFrance: 5/10/30 Spread Fair Value Model Chart 6BFrench Butterfly Strategy Performance The mechanical trading rule appears to also work relatively well when applied to butterfly combinations in the French OAT government bond market (Chart 6B). Italy & Spain Turning to European countries in the periphery, the most attractively valued butterfly combinations appear to be going long the 5/30 barbell and shorting the 7-year bullet in the Italian government bond yield curve (Table 6), and favoring the 7/30 barbell versus the 10-year bullet in the Spanish government bond market (Table 7). Table 6Italy: Butterfly Strategy Valuation: Standardized Residuals Table 7Spain: Butterfly Strategy Valuation: Standardized Residuals In the case of Italy, the 7-year bullet appears 7bps expensive according to our model and would only outperform its counterpart given a steepening in the 5/30 Italian curve slope greater than 41bps (Chart 7A). The mechanical trading rule appears to work well when applied to Italian butterfly combinations, displaying better excess returns than for most other countries we’ve looked at (Chart 7B). Chart 7AItaly: 5/7/30 Spread Fair Value Model Chart 7BItalian Butterfly Strategy Performance Looking at Spain, the 10-year bullet appears 8bps expensive according to our model and would only outperform its counterpart given a steepening in the 7/30 Spanish curve slope greater than 64bps, which seems highly unlikely at this point in time (Chart 8A). The mechanical trading rule works well when applied to Spanish butterfly combinations and shows very few periods of underperformance since the early 90s (Chart 8B). Chart 8ASpain: 7/10/30 Spread Fair Value Model Chart 8BSpanish Butterfly Strategy Performance UK The most attractively valued butterfly combination currently on the UK Gilts yield curve is holding a 2/30 barbell versus the 10-year bullet, which currently displays a standardized residual of -1.64 (Table 8). Table 8UK: Butterfly Strategy Valuation: Standardized Residuals The 10-year bullet appears 21bps expensive according to our model and would only outperform its counterpart given a steepening in the 2/30 curve slope greater than 62bps (Chart 9A). Chart 9AUK: 2/10/30 Spread Fair Value Model Chart 9BUK Butterfly Strategy Performance Chart 9B shows that applying the simple mechanical trading rule works well over time. Canada The most attractively valued butterfly combination currently on the Canadian yield curve is favoring a 1/30 barbell versus the 10-year bullet, which currently displays a standardized residual of -1.11 (Table 9). Table 9Canada: Butterfly Strategy Valuation: Standardized Residuals The 10-year bullet appears 18bps expensive according to our model and would only outperform its counterpart given a steepening in the 1/30 curve slope greater than 60bps (Chart 10A). Chart 10ACanada: 1/10/30 Spread Fair Value Model Chart 10BCanadian Butterfly Strategy Performance Once more, following the mechanical trading rule looks promising (Chart 10B). In fact, we observe only four periods of negative year-over-year returns. Japan The most attractively valued butterfly combination currently on the Japanese JGBs yield curve is going long the 5/10 barbell and shorting the 7-year bullet, which is currently below one standard deviation above its implied-model fair value, with a standardized residual of only -0.86 (Table 10). Table 10Japan: Butterfly Strategy Valuation: Standardized Residuals The bullet appears slightly expensive, by 5bps, and would only outperform its counterpart given a steepening in the 5/10 JGB curve slope greater than 32bps (Chart 11A). Chart 11AJapan: 5/7/10 Spread Fair Value Model Chart 11BJapanese Butterfly Strategy Performance The mechanical trading rule also performs well when selecting the most attractive butterfly combinations in the Japanese government bond market (Chart 11B). Australia The most attractively valued butterfly combination currently on the Australian government bond yield curve is going long the 3/10 barbell and shorting the 7-year bullet, which presently displays a standardized residual of -1.52 (Table 11). Table 11Australia: Butterfly Strategy Valuation: Standardized Residuals Please note that we excluded the 20- and 30-year government bonds from our analysis since they were first issued only a few years ago. Our yield curve model suggests that the bullet is 10bps expensive and would only outperform its counterpart given a steepening in the 3/10 Australian curve slope greater than a whopping 92bps (Chart 12A). Chart 12AAustralia: 3/7/10 Spread Fair Value Model Chart 12BAustralian Butterfly Strategy Performance The mechanical trading rule also performs well when picking the most attractive butterfly combinations in the Australian government bond market (Chart 12B). Footnotes 1 Please contact your sales representative to request a copy. 2 Please see US Bond Strategy Special Report, "More Bullets, Barbells And Butterflies", dated May 15, 2018, available at usbs.bcaresearch.com. 3 The scientific study of butterflies. 4 Using benchmark Treasury yields would only result in slightly different weightings for the bonds in the barbell without affecting the outcome. 5 Please see Global Investment Strategy Weekly Report, "Markets Too Complacent About The Coronavirus", dated February 21, 2020, available at gis.bcaresearch.com. 6 Please see Global Fixed Income Strategy Weekly Report, "What Central Banks Are (Or Should Be) Watching", dated February 19th, 2019, available at gfis.bcaresearch.com.
