In “Bursting Bund Bubble: 2 Charts And Some Lessons From History,” we recapped the sell-off in German government bonds, touching on the severity of the yield spike (with emphasis on Thursday’s intraday move above 77bps), the breakdown in the historical relationship between Bunds and Treasurys/Gilts, and parallels between the rout and the 2003 sell-off in JGBs. On the latter issue, we presented the following chart from Barclays which shows the degree to which this most recent incarnation of government bond carnage mirrors the 2003 manifestation.
While various exogenous factors have likely contributed to the big Bund battering — including profit-taking, the frontrunning of positive EGB supply in May, and short calls from several bond market heavyweights — it’s worth noting that at the heart of the carnage are two embedded self-fulfilling feedback loops which, together with decreasing liquidity, have made for a rather precarious situation.
The first loop is related to the structure of PSPP (i.e. no purchases below the depo rate) and has been described by Goldman as follows:
As the decline in yields that has followed the liquidity injections has made its way to intermediate maturities, the market has extrapolated that the Bundesbank would have to purchase a larger share of longer maturity bonds to fill its quota. This is a self-reinforcing expectations loop, where lower yields beget lower yields. Given its nature, the loop can also switch direction. As yields rise, more bonds become eligible for central bank purchases, and the price action goes into reverse.
The second self-feeding dynamic is something we’ve discussed at length before, most notably in 2013 when volatility-induced selling — reminiscent of the 2003 JGB experience — hit the Japanese bond market again, prompting us to ask the following rhetorical question:
What happens to JGB holdings as the benchmark Japanese government bond continues trading with the volatility of a 1999 pennystock, and as more and more VaR stops are hit, forcing even more holders to dump the paper out of purely technical considerations?
The answer was this: A 100bp interest rate shock in the JGB yield curve, would cause a loss of ¥10tr for Japan's banks.
What we described is known as a VaR shock and simply refers to what happens when a spike in volatility forces hedge funds, dealers, banks, and anyone who marks to market to quickly unwind positions as their value-at-risk exceeds pre-specified limits.
Predictably, VaR shocks offer yet another example of QE’s unintended consequences. As central bank asset purchases depress volatility, VaR sensitive investors can take larger positions — that is, when it’s volatility times position size you’re concerned about, falling volatility means you can increase the size of your position. Of course the same central bank asset purchases that suppress volatility sow the seeds for sudden spikes by sucking liquidity from the market. This means that once someone sells, things can get very ugly, very quickly.
Here’s more from JPM on the similarities between the Bund sell-off and the JGB rout that unfolded two years ago:
The sharp rise in bond volatility over the past week or so is reminiscent of the VaR shocks of October 2014 in US rates and April 2013 in Japanese rates. The common feature of these rate volatility episodes was that there was no clear fundamental trigger. Instead, positions and flows experienced a sharp swing making these VaR episodes appearing more technical and unpredictable in nature. In October 2014, a violent capitulation on short positions at the front-end of the US curve had caused a collapse in UST yields. In April 2013, profittaking in long duration exposures post BoJ's QE announcement caused a sharp rise in JGB yields that started reversing two months after.
What is causing VaR shocks and why are they happening often? We argued before that one of the unintended consequences of QE is a higher frequency of volatility episodes or VaR shocks: investors who target a stable Value-at-Risk, which is the size of their positions times volatility, tend to take larger positions as volatility collapses. The same investors are forced to cut their positions when hit by a shock, triggering self- reinforcing volatility-induced selling. This, we note, is how QE increases the likelihood of VaR shocks.
The proliferation of VaR sensitive investors, such as hedge funds, mutual fund managers, risk parity funds, dealers and banks raise the sensitivity of bond markets to self- reinforcing volatility-induced selling. These investors set limits against potential losses in their trading operations by calculating Value-at-Risk metrics. Value-at-Risk (VaR) is a statistical measure that investors use to quantify the expected loss, over a specified horizon and at a certain confidence level, in normal markets. Historical return distributions and historical market volatility measures are often used in VaR calculations given the difficulty in forecasting volatility. This in turn induces investors to raise the size of their trading positions in a low volatility environment, making them vulnerable to a subsequent volatility shock. When the volatility shock arrives, VaR sensitive investors cut their duration positions as the Value-at-Risk exceeded their limits and stop losses are triggered. This volatility induced position cutting becomes self- reinforcing until yields reach a level that induces the participation of VaR-insensitive investors, such as pension funds, insurance companies or households.
The VaR shock in the JGB market in April 2013 contained most of the above characteristics. By looking at quarterly Flow of Funds data from the BoJ, it was Japanese banks, Broker/Dealers and foreign investors who sold JGBs at the time. And it was VaR insensitive investors, such as Pension Funds and Insurance Companies and Households (via investment trusts) who absorbed that selling along with the BoJ.
And as for just how illiquid the Bund market (where Mario Draghi sees no signs of PSPP-induced trouble) has become...
The 30-year Bund futures contract in particular has been at the core of the latest VaR shock as it was the one experiencing the highest volatility. Figure 1 shows a price to volume based indicator of volatility for the first 30y Bund futures contract [and] shows how big the deterioration in 30y Bund futures market breadth has been over the past week since April 29th. A stabilization in this, or similar volatility indicators, is perhaps needed to make us confident that this latest VaR shock is behind us.
But it is not only market breadth which deteriorated sharply over the past week or so. Another important dimension of market liquidity, market depth, i.e. the ability to execute a large trade without moving markets too much has also seen a decline. This is shown in Figure 2 where we proxy market depth by the 5-day average of the tightest three bids and asks each day, measured in number of Bund futures contracts. What is also striking in Figure 2 is how big the deterioration in the market depth of the 30y Bund futures contract has been during 2014 before even the ECB started its QE purchases. For example, our market depth proxy of Figure 2 suggests that one could have traded around 100 contracts without moving markets too much at the beginning of 2014 but this number has fallen to below 20 this week. One of the reasons for this deterioration is perhaps the almost secular contraction in the German repo market by 30% over the past five years. This contraction in German repo markets started five years ago as shown in Table 1 and continued up until the most recent data in 2014. It intensified during the euro debt crisis as flight to quality made investors unwilling to depart from their “precious” German collateral. The net withdrawal of Bund collateral as a result of ECB’s QE and lower issuance by the German government makes it likely that the German repo market contraction will continue if not intensify this year hampering market trading liquidity further.
* * *
Between a hopelessly illiquid market, deeply negative German supply in June and July, a central bank determined to prove it has not lost control in spite of all evidence to the contrary, and all of the dynamics discussed above, it's safe to say the fireworks aren't over yet.