Investors looking for superior returns and turning their attention to high yield bonds should not overlook that liquidity risk has increased dramatically since end 2015. Increased market volatility, the impact of Brexit combined with uncertainty over US interest rates could mean that now more than ever, investors need to watch out for liquidity red flags in their high yield investments.
The high yield bond market is rising strongly– up 12% year-to-date according to the Barclays US High Yield Index. Furthermore, high yield has outperformed investment grade corporate debt by 13% from February 11 through July 29, 2016 again according to Barclays. With approximately $26bn of new debt issued in the past four weeks, high yield corporate debt has attracted lots of buyers and sellers.
But this begs the question: “Is there more danger in high yield bond funds and can investors accurately measure a fund’s liquidity?”
In the winter of 2015, an almost unheard of situation happened. A mutual fund, normally required to guarantee daily liquidity, blocked its clients from withdrawing money. The Third Ave Focused Credit Fund (TFCIX), citing losses and a lack of liquidity in the high yield bond market, put some of its assets into a trust to be sold over time.
For mutual funds that invest in illiquid assets, such as high yield bond funds, liquidity is a major concern. If too many investors try to cash out at the same time, to provide liquidity, a fund may be forced to sell its holdings at fire sale prices – if it can find a buyer at all. Fitch has forecast that $90bn of high yield debt could default by year end, while S&P Global has forecast that the US speculative grade default rate will climb to 5.3% by the end of 2017’s first quarter, up from 3.8% twelve months earlier. Furthermore, high yield debt’s recovery rate – or portion of principal and interest recovered in bankruptcy – was down to 34% in 2015, well below its historical average of 46% according to Lehmann Livian Fridson Advisors.
More defaults with less recovered could set off panic bells for investors in high yield funds. Last September, the SEC even proposed permitting funds to adjust a fund’s redemption price to pass on the costs of immediate liquidity to redeeming shareholders. High yield bond investors could face a rocky ride ahead. If a fund is not able to sell investments quickly without taking a large loss, the ability of the fund to provide daily liquidity is compromised.
Measuring the liquidity of a fixed income portfolio is a challenging exercise. One must be aware of the size and date of issuance of a bond. Large new issues tend to be very liquid in contrast to the often scant liquidity of smaller, older issues. Instrument type matters greatly; liquidity for structured products and credit default swaps has plummeted since the financial crisis. And while stricter capital rules and increased risk aversion have decreased investors’ appetite for corporate and high yield bonds; low interest rates globally have had the opposite effect. In an ideal world, measuring liquidity risk marries this qualitative information with quantitative data that are not easy-to-obtain nor evaluate such a bid-ask spreads, transaction slippage, and dealer inventories. Even harder, measuring a particular fund’s liquidity should also consider the manager’s historical trade sizes, data that are rarely publicly available.
As it may be impractical to obtain the data to measure liquidity directly, we can use a fund’s return history instead. Actual trade prices for securities with low liquidity are recorded at sparse time intervals. High yield funds look for either a recently traded price or dealer quote to calculate a security’s option adjusted spread (OAS) (1). If neither a recent trade nor quote is available for a holding, an earlier OAS is typically carried or extrapolated forward which means the bond’s price may not reflect changing time to maturity, interest rates or other pricing factors. This induces auto-correlation in the return time series of that bond, and as a result, the entire portfolio. Hence, the level of auto-correlation is related to the level of liquidity in a fund (2).
Assessing fund liquidity using the Durbin-Watson Statistic
Fund investors can use the Durbin-Watson statistic to measure auto-correlation.(3,4) Using the Durbin-Watson statistic to assess investment liquidity is both valuable and simple to calculate and can help investors assess if any of their funds potentially stand out as being highly illiquid.
The Durbin-Watson is measured from a value of zero to four, with a value close to zero indicating positive auto-correlation (we could create an equation to predict the next return from prior returns without changing the prior returns’ signs) and a value close to two indicating little auto-correlation (prior returns do a poor job of predicting future returns).
In the graph above, we have plotted the Durbin Watson statistic for 161 high yield bond funds from Morningstar using monthly return data from the past three years on the Y axis. We see a high degree of auto-correlation in the returns of Third Ave Focused Credit, as it has the lowest Durbin Watson statistic value. Fund size does not appear to be related to illiquidity, as both small and larger funds have diverse values, as shown with the fund assets plotted on the X axis. Funds below the black line have highly significant positive serial correlation (5).
The recent strong performance of the high yield market has caught investors’ attention. However, with the liquidity in the market highly fragile, investors need to understand the investment that their managers make into illiquid securities. By using the Durbin-Watson Statistic, we believe that investors in high yield bond funds can be more sensitive to the potential for liquidity related losses.
Sean Ryan is senior analyst, Research, MPI (Markov Processes International)
1. OAS, which stands for option adjusted spread, is a measure of the credit riskiness of a bond and moves inversely to its price.
2. Low liquidity is not the sole source of auto-correlation, although the literature has shown it to be the largest factor in the auto-correlation of hedge fund returns. Auto-correlated returns can also appear due to difficulties in pricing over the counter securities, window dressing, performance smoothing, marking to model, non-synchronous trading, fraudulent accounting, momentum, unexploited market opportunities, time varying expected returns and time varying leverage.
3. We measure the unconditional auto-correlation. Typically, managers will only smooth out the negative returns of their portfolio performance which means that auto-correlation will mostly exist when past returns are negative. Since we focus on the illiquid nature of a fund, not on a manager’s intent, we are focusing on unconditional auto-correlation.
4. The Durbin Watson statistic is biased when lagged dependent variables are part of the regression, as is the case here. An alternative test would be to calculate the beta of the regression between a fund and its lag and check if that beta is positive and significant via the t statistic. We calculated both the Durbin Watson and t-statistic on the high yield funds of this blog and found both tests to be consistent to each other. Both tests highlighted the Third Avenue Fund as an extreme outlier.
5. The line is the lower bound for the 99% confidence interval for the Durbin Watson statistic. That is, if we took a completely random series of numbers and calculated the Durbin Watson statistic, we would expect 99% of those values to be greater than this critical value. In the case of high yield funds, 13 of the 161 funds, or 8%, have Durbin Watson values below the 99% critical value.