- In many financial markets, technology has improved price discovery and transparency.
- Some parts of the fixed-income market have not yet benefited from improved technology, leading to concerns about liquidity.
- Some heavily traded markets, such as listed equities, have produced enough data that patterns are discernible and reasonable predictions can be made about transactions costs in many different conditions. However, it’s not possible to do this in many parts of the fixed-income market.
- While we cannot predict fixed-income liquidity costs precisely, we can use guidance from the repo market and other sources to classify securities into five liquidity levels.
- This liquidity quantification system helps us understand where we might need to use additional liquidity-providing techniques to better match liquidity supply with liquidity demand.
Financial markets provide places where buyers and sellers can find each other. Technology ranging from eBay to high-frequency trading has improved the ability of buyers and sellers to find each other and to discover the going price of an item they want to trade. Cartels controlling markets are under threat from technologies providing transparent price discovery—for example, Tesla Motors has shaken car dealers by refusing to use or establish a dealer network. Instead Teslas are offered and bought online.
Human intermediaries have been eliminated in many cases. “Shares actually traded on the [New York Stock Exchange’s] historic floor... amount to perhaps 15% of the total, maybe less. The rest of the volume is handled electronically...”1 A February 4, 2015 Bloomberg article, “CME Cuts Ties to 19th Century by Shutting Once-Chaotic Pits,” explains that because virtually everything is done electronically, “few exchanges still maintain spaces where humans interact.”2
And then there is the fixed-income market. While a few of the larger government bond markets use modern technology, trillions of dollars still trade via human-to-human exchanges.3 Consider, for example, bonds issued by the health care company Johnson & Johnson (JNJ). They trade in the $8 trillion US corporate bond market, while JNJ’s common equity trades on the $20 trillion New York Stock Exchange.4 There is only one kind of JNJ common equity, so trading is brisk. There are 255 JNJ bond issues, which trade far less frequently. That is because natural matches between buyers and sellers are harder to come by.
In fixed-income markets, banks act as dealers, not just brokers. If someone wanted to sell a 7-year JNJ bond and someone else wanted to buy a 5-year JNJ bond, a dealer would step in and buy the 7-year and sell the 5-year. Over many such transactions the dealer can manage interest-rate and credit risk by hedging and by using its capital as a buffer, while making a nice profit on the transactions using asymmetric information and a privileged market franchise.
Since the global financial crisis of 2008–2009, banks have been increasingly restricted in their ability to use their capital to provide this type of buffer. In the land a great hue and cry has arisen warning that because of this, fixed-income markets have become fragile and a liquidity disaster is waiting to happen. Cynics suspect that this hue and cry is led by the banks, in whose interest it might be to make a fuss so they get relief from the capital rules that are restricting their profits.
Whatever the cause, attention has been focused on bond market liquidity. Western Asset has, naturally, a keen interest in this issue and has been both analyzing and acting on it. Recent Western Asset white papers by Thomas McMahon (Analyzing Credit Market Liquidity, April 2015) and Michael Buchanan (Q&A: Liquidity in the Fixed-Income Market, July 2015) have discussed aspects of bond market liquidity as they affect our portfolios. In this white paper, we will discuss the method that Western Asset has been using to quantify bond market liquidity since 2013.
Expanding on Professor Harris’s pithy definition, we can say that liquidity is the ability to turn an asset into cash or cash into an asset over a desired time frame. That involves finding a buyer, a seller and a price they agree to. Good liquidity requires:
- Large and relatively balanced numbers of buyers and sellers
- Transparent price discovery
- A matching mechanism so that buyers and sellers can be brought together
When buyers and sellers are plentiful and relatively balanced, transactions are easy and liquidity is high. When there is an imbalance or a difficult search—buyers and sellers may exist but they don’t know about each other—then transactions are difficult and liquidity is low.
There are some obvious factors that affect a security’s liquidity during a transaction:
- The characteristics of the security. The common stock of a large corporation like Apple is usually very liquid; Apple (AAPL) is currently trading about $6 billion a day. A small bond issue of a small municipality is usually very illiquid.
- The urgency of the transaction. The shorter the time available to match a buyer and a seller, the more costly the transaction will be to the initiating party.
- The size of the transaction. A transaction that is a small percentage of typical trading volume is generally easier to arrange than one that is a large percentage. However, very small transactions might be more difficult because they are odd lots.
- The direction of the transaction—buy or sell.
- Current market conditions. In disrupted markets—such as parts of the fixed-income market in late 2008 and early 2009—transactions that might otherwise be reasonable, are not.
- The price history of the security—for example, positive or negative momentum.
The overarching goal of analysts attempting to model liquidity is to produce an algorithm that takes as input the factors listed above and produces, as output, the cost of a transaction that has not yet occurred. For example if a security’s current price is $100 and we want to sell urgently, a liquidity algorithm might say that we would get only $98, i.e., the cost of transacting is $2.
Unfortunately no such algorithm exists in fixed-income markets. The information required to produce an accurate estimate of costs for a trade that has not yet occurred is simply not available—the markets are too variable and the data are too sparse. There is a tremendous amount of data across a large number of diverse market participants in many market conditions for all trading in AAPL. Conversely, there are very little data about trading in the $9 million issue of the town of Cohasset, Massachusetts 3% due November 15, 2017.
Western Asset’s approach to liquidity quantification is influenced by the humbling observation of the previous paragraph: trying to predict fixed-income liquidity costs definitively is simply not possible. We might come up with a number, but it would be a misleading level of false precision. What we can realistically do is to rank order securities from most liquid to least liquid. Even rank ordering can veer toward false precision if we try to get too granular. Our method simply classifies securities into one of five liquidity levels. These levels are intended to hold through normal markets and disrupted markets.
Our liquidity framework starts with repo6 haircuts. In the repo market, securities are pledged as collateral against borrowing. For example, $102 worth of US Treasury bonds (USTs) might be pledged against $100 worth of overnight borrowing (cash). If the borrower does not repay, the lender can cash in the USTs to get back the missing principal. The $2 “haircut”—the difference between the value of assets pledged and the amount of cash borrowed—is set as a buffer so that the lender can reasonably expect to cash in the collateral and still get back at least the principal. Thus the repo haircut contains a joint prediction of how illiquid and how volatile the collateral will be in a market that is likely to be disrupted since the borrower has defaulted.
The Federal Reserve Bank of New York (NY Fed) conducts a regular survey of dealers that find the haircuts—more formally the margin levels—for various kinds of assets. Exhibit 1 shows its survey from June 2015.
Looking at the median numbers, we can see that collateralized debt obligations (CDOs) sensibly require a much larger median haircut (15%) than USTs (2%).
We obtained historical NY Fed data and looked at repo haircuts during the 2008–2009 global financial crisis.9 They were not too different between the credit crisis period July 2008 to July 2009 and the relatively stable period July 2009 to January 2010. For example, the mean haircut on commercial paper was 4.2% in the crisis period and 3.9% in the stable period. Except for equity—which had a lower haircut in the crisis than in the stable period—the asset class rankings were roughly preserved. This contributed to our belief that we can rely on these levels in both normal and disrupted markets.10
From this basis, we consulted with Western Asset trading desks to get market color and fill in more detail, especially for instruments that were not covered in the repo surveys. We found that credit rating was important for corporates, while the nature of the issuer was important for sovereigns. Over the more than two years that we have had the system in place, we have made a number of refinements to reflect the feedback we received from our traders and dealers. Our current basic level assignments are shown in Exhibit 2.
We further refine our scale with adjustments up or down based on issue size, bond age, and years to maturity. For example, an A rated investment-grade corporate (Level 3) might be boosted to Level 2 if it is recently issued (on-the-run) and very large in issue size. On the other hand, the AA rated Cohasset, Massachusetts bond we referenced earlier would get a penalty for small size, being off-the-run, and maturing in over a year, and would be demoted from Level 2 to Level 3.
Another supplement to our scale adds equities, which fall into levels 2 to 5 depending on the median daily trading volume. Listed equity markets are deep enough and transparent enough that past trading patterns are reasonably predictive of future patterns. In fact, in times of stress, equity trading volume often goes up not down. Of course this type of increased liquidity is accompanied by higher price volatility.
We cannot count on being able to predict when market conditions will move from normal to stressed. What will most dramatically change from a normal to a stressed environment will be the costs of transacting. But if we have been successful in our assignments, Level 5s should be more expensive than Level 4s in all environments. This helps us understand which portfolios, and which parts of portfolios, would be relatively robust to stressed liquidity demands and which would not.
Lack of liquidity has both a downside and an upside. In the ordinary course of managing a portfolio, lower liquidity will increase the cost of repositioning the portfolio to react to changing market conditions. On the other hand, market participants know this and build in an illiquidity premium—that is, securities that might be less liquid are cheaper than their fundamentals would indicate. A patient investor relying on good fundamentals can profit. The manager’s skill in navigating these offsetting effects will determine whether or not the portfolio benefits.
Portfolios managed for a single client retain control of liquidity demand that arises through subscriptions and redemptions—the single client does not have to worry about someone else making a decision to force the portfolio to transact.12 A manager can position a separately managed portfolio to benefit from waiting out a liquidity storm and reaping the liquidity premium if the client feels comfortable with such a strategy. On the other hand, clients who know they might have liquidity demands on their portfolios might not want to take such risks. Clearly it is important to have good communication between managers and clients about these issues.
However, the SEC’s reference to “fund liquidity needs” points to another consideration: in collective investment vehicles (as opposed to separately managed portfolios with a single client), some investors might demand liquidity while others might supply it. For example, in both US mutual funds and in European UCITS funds, investors selling their shares into a disrupted market would be demanding liquidity while investors buying or even just staying invested would be supplying it.
We routinely look at all portfolios that might have exogenous liquidity demands in order to determine what natural liquidity supply the portfolio has. Sometimes natural liquidity supply is low. “Fixing” this may not be a simple matter of shifting assets to better liquidity levels. For example, Exhibit 2 indicates that a corporate high-yield fund cannot meet its mandate to invest mainly in high-yield securities without having a high percentage of Level 4s and Level 5s.
When the natural supply of liquidity is comparatively small, we look to a variety of measures including increasing liquidity supply in the better liquidity levels, as long as that is compatible with the portfolio’s mission. For example, Exhibit 3 shows a sample global high-yield bond fund.
Because this is a high-yield fund, both the fund and its benchmark have a preponderance of assets in Levels 4 and 5. However there are enough assets in the more liquid levels (1–3) to meet the worst one-day redemption the fund has ever experienced over any period (including the 2008–2009 global financial crisis). While normally a fund would try to sell all of its assets proportionally to meet redemptions, it is good to know that the portfolio manager has the option to make the decision to sell some more liquid assets immediately and to rebalance the rest of the fund more patiently. This might be done if, in the manager’s opinion, everyone would benefit from this approach.
This particular fund had a worst five-day redemption of 18% of assets.13 In order to meet this massive redemption with its current assets, the fund would have to have sold some less liquid securities. This seems reasonable: it is not feasible to have 18% of a high-yield fund—or worse, the 23% that is the size of the worst 20 business day redemption—in non-high-yield securities.
Because of the possibility of such large outflows, it is tempting to compare liquidity in collective investment vehicles to liquidity in banks. It is well known that a bank run of sufficient magnitude will break even the most perfectly sound bank.14 That is why banks have deposit insurance and central banking in most countries.
A collective investment vehicle facing a big enough flood of redemptions similarly cannot meet those redemptions without investors losing money. However, there the similarity ends: a collective investment vehicle does not guarantee deposits the way a bank does. Unlike bank depositors, investors in such vehicles must expect to bear the risk that the value of their investments may drop due to a combination of falling markets and transactions costs.
Thus Western Asset’s liquidity framework is not intended to manage away any possible pain that investors might feel due to liquidity stress. Instead, it helps us understand where we might need more tools to alleviate—although we cannot cure—liquidity costs. For example, we might hold some cash that is securitized by index credit default swaps in a high-yield fund such as the one shown in Exhibit 3. Index credit default swaps provide overall exposure to the appropriate asset class and are often more liquid than cash bonds. Although active “alpha” is not provided by securitized cash, the liquidity buffer it provides can be useful. Similarly, we might hold heavily traded ETFs. There is basis risk in each of these strategies, but, properly managed, these techniques can increase our liquidity supply. Other tools our liquidity analyses have considered and put in place in some of our vehicles include lines of credit and interfund lending.
The fixed-income market has not benefited from market making technology as much as other asset classes have. Due to post-crisis capital constraints, the dealer market has become less capitalized. New and better markets may eventually emerge, but until they do, fixed-income liquidity requires a high level of scrutiny. It is unrealistic to think that a precise estimate of the cost of trading can be produced in these markets. Instead, Western Asset has developed a five-level liquidity scale that helps us understand—both in normal and stressed markets—the liquidity supply that our portfolios have both on an absolute basis and a relative basis. Especially in cases where there can be exogenous liquidity demands, this helps us determine what steps to take to be ready to meet those demands.
- Telephony (1876) and instant messaging (mid-1960s) are used, grudgingly.
- http://sifma.org/research/statistics.aspx and http://www.nyxdata.com/nysedata/asp/factbook/viewer_edition.asp?mode=tables&key=333&category=3
- Source: Bloomberg
- “Repo” is short for “repurchase;” the borrower has technically sold the collateral to the lender for a short period and has agreed to repurchase it for a slightly higher price that is the cost of borrowing.
- Fedwire-eligible means the collateral is accepted by Federal Reserve Banks. Non-Fedwire-eligible collateral may be used between private institutions.
- http://www.newyorkfed.org/research/staff_reports/sr506.pdf, Table II.
- We used a similar survey for European markets conducted by the International Capital Market Association (ICMA). http://www.icmagroup.org/Regulatory-Policy-and-Market-Practice/short-term-markets/Repo-Markets/repo/latest/, Table 2.11.
- An “exogenous liquidity demand” is a transaction that is forced by a market event or a party other than the portfolio manager. While a portfolio owned by a single client is insulated from subscriptions and redemptions that might affect one class of shareholders differently from another, it might still have exogenous liquidity demands. For example, a portfolio that invests in euros but is automatically 100% hedged into dollars might have to liquidate some assets to pay for its FX hedges if the euro does well.
- The worst 5-day outflow for this fund during the 2008 credit crisis was 9%. It also had an 18% 5-day outflow in mid-2014. While the 2014 outflow was due to unique circumstances that are unlikely to be repeated, we use it as our worst case stress test.
- Diamond DW, Dybvig PH (1983). “Bank runs, deposit insurance, and liquidity.” Journal of Political Economy 91 (3): 401–419.