Liquidity Fissures in the Corporate Bond Markets

Abstract

In this chapter, we conduct another case study. In particular, we will move from a retrospective of the Volmageddon to a forward-looking study of the US high yield corporate bond markets, in the presence of majority agents. The majority agents are authorized participants (APs) in bond ETFs and to a lesser extent, investors in certain mutual funds. Beyond a certain size, bond market liquidity can no longer support the ETF arbitrage mechanism. This implies that a purely technical equity flash crash can cause lasting damage to the bond market. We combine a survey-based price impact estimate with a model of mutual fund flows to calculate the potential fallout from a limit down move in various high yield ETFs.

Postscript: Tackling the High Yield Bond—ETF—Mutual Fund Feedback Loop During the Covid-19 Crisis in Q1 2020

Let us rewind the clock to February 2020. Once investors realized the implications of lockdowns on economic growth, equities sold off violently and credit spreads spiked from a low base level. The following graph tracks the “spread to worst” of a representative basket of US high yield bonds relative to equivalent duration Treasuries. The spread to worst represents the lowest yield differential investors can capture if a given bond does not default. It accounts for the fact that issuers can retire callable bonds early, changing the effective yield to maturity (Fig. 5.12).

Fig. 5.12

(Source Bloomberg)

Jump in high yield spreads during the Covid-19 crisis

Using DECISION TREE 5.1 as a guide, we can specify that an exogenous event triggered the mega spike in high yield spreads.

High yield spreads reached a peak of 9.23% on March 23, from 4.03% on February 17. As expected, HYG and JNK started to trade at a significant discount as spreads peaked. On the previous trading day (March 20), HYG traded at a −1.27% discount to NAV. JNK was sputtering even more badly, trading at a −1.60% discount. Nominated dealers were obviously in risk averse mode, reluctant to “arbitrage” the discount in a highly illiquid market.

HYG and JNK net flows were dangerously skewed to the downside. Note that we have taken a 10 day trailing average of combined flows in the graph below. In the worst 10 day window, roughly $6 billion (or 10 * $600 million) was taken out of two ETFs whose assets add up to roughly $30 billion (Fig. 5.13).

Fig. 5.13

(Source Bloomberg)

Rolling 10 day average of high yield ETF flows, early 2020

Operating with a slight lag, high yield mutual funds also suffered from large outflows.

As of March 23, the odds of a technical crash in a major high yield ETF were unusually high. We have already mentioned that there were over 500 instances of a stock going limit down in December 2018. From March 1 to March 20, 2020 alone, nearly 6,000 circuit breakers were hit. Under the conservative assumption that 50% of these were limit down events, roughly 200 circuit breakers were hit on a typical trading day! Many of the scenarios we had discussed in 2019 were playing out.

It is likely that the Federal Reserve was alarmed by many of the same indicators we have described above. On March 23, they took a drastic measure, announcing that they would directly buy investment grade corporate bonds and ETFs. (Open market purchases of corporate bonds and ETFs obviously have a more reliable impact on spreads than ordinary Treasury purchases. Buying Treasuries has a more indirect impact on credit spreads and equities. We discuss the usual transmission mechanism in Chapter 7.) From there, the floodgates opened. Dealers and mutual funds in the investment grade space now had access to a willing buyer with a theoretically unbounded balance sheet. VCIT and LQD, two of the largest corporate bond ETFs, popped +5.4% and +7.4%, respectively, on March 23 alone.

While the Fed did not announce its intentions toward high yield bonds, investment grade bond purchases clearly reduced the pressure on dealer balance sheets. By March 27, HYG and JNK were trading at a significant premium (+2.09% and +1.24%, respectively) to NAV. They had resumed their roles as leading indicators, suggesting improved credit market conditions.

Had the Fed not intervened directly in the credit markets, we suspect that HYG and JNK would have fallen into a destructive feedback loop. Recall that, once dealers buy shares at a discount to NAV, they have to sell bonds into an illiquid market. This can cause bond prices to overshoot to the downside. Investors with long credit exposure may then be inclined to hedge by selling more units of HYG, forcing HYG back to a discount.

We can see from the graph below that dealer risk aversion generally increases as a function of market volatility (Fig. 5.14).

Fig. 5.14

(Source Bloomberg)

As the VIX rises, HYG’s link to the cash bond market becomes increasingly unstable

After March 23, net flows to HYG and JNK became strongly positive, as we can see in the next graph. Again, we have smoothed the data with a 10 day trailing moving average (Fig. 5.15).

Fig. 5.15

(Source Bloomberg)

Money flows back into HYG and JNK once Fed’s intentions are digested by the market

However, conditions in the high yield space remained volatile. From March 20 to April 9 (consisting of 15 trading days), the standard deviation of HYG and JNK’s premium/discount to NAV remained roughly 10 times above its 2019 average. The following chart demonstrates this point (Fig. 5.16).

Fig. 5.16

(Source Bloomberg)

The alchemy of liquidity fails when investors require liquidity the most

In early April, US high yield mutual funds also suffered significant outflows, again operating with a lag to high yield ETFs.

On April 9, 2020, the Federal Reserve announced that it would increase the size of its corporate bond and ETF asset purchases by over $550 billion. Significantly, new provisions also allowed for the purchase of high yield bonds if they had been downgraded during the crisis.

If the Fed had not taken extraordinary action on March 23 and April 9, our implementation of the GJN model would have anticipated additional mutual fund outflows in the $7 billion range from April to June 2020. The toxic feedback loop may have continued. The graph below focuses on Scenario 8 (illiquid funds with negative alpha in high risk regimes) outflows from earlier in the chapter. It calculates expected flows on a monthly basis (Fig. 5.17).

Fig. 5.17

(Source Bloomberg)

High yield mutual fund outflow expectations, based on the GJN paper methodology

Looking into the future, we believe that structural problems in the high yield markets remain. According to order book-related measures, equity liquidity has been drop** sharply over the past few years. A limit down move in HYG or JNK would not require an exogenous shock, as in Q1 2020. The overhang of high yield debt also suggests a bloated market, as we can see from the graph below. The paths taken in DECISION TREE 5.1 are still entirely possible, especially if the Fed chooses not to make direct ETF purchases during the next crisis.