Part I of our series “Bid This Bond” discussed how a host of regulations, such as Basel III, MiFID II, EMIR, Dodd-Frank, and the Volcker Rule provisions, will change trader and sales-trader “workflow” by adding more steps and validation points to the process of pricing a bond.
We will now take a closer look at the quantitative approaches that firms and their traders can use to practically implement the liquidity risk management measures needed to comply with these regulatory guidelines. Until recently, regulatory intent was well ahead of the market’s ability to comply because there were few quantitative approaches that supported compliance. New quantitative evaluation techniques and technologies are now making it possible for firms to practically implement the detailed requirements underpinning these regulations.
Buy side, take notice
To-date most of the regulators’ focus and prescribed guidelines are directed at the sell side. Market leading sell-side firms have already implemented, or are actively trying to implement, quantitative methods to help them create and enforce internal policies to assess asset quality and inventory composition. This quantitative, data-driven insight is extremely valuable for the buy side because it can help traders and portfolio managers understand the factors that their sell side counterparts are considering when they ask them to “bid this bond.”
Moreover, the techniques the sell side uses to assess asset quality will be of particular value to the buy side as more and more regulators in the years ahead shift focus from the sell side to the buy side to address liquidity and redemption risk. Proposed regulations in the U.S. (i.e. SEC 22e-4) and Europe (i.e. MiFID II) are beginning to address how regulators want firms to assess, monitor, manage and adjust trading methods based the liquidity risk of their portfolios.
Quant metrics and capital management
When a bond comes in for a bid, regulations require the trader to factor the cost of capital and compare the asset’s quality with the firm’s internal policy. Best practice should also include acceptable asset classifications, market liquidity risk profiles, some measure of anticipated carry time and whether the asset can be positioned in the firm’s current inventory (Figure 1). Fixed income, more than any other asset class, is characterized by many as relatively illiquid – many securities are not frequently traded. Moreover, trade data is not publicly available in the case of non-U.S. instruments.
Quantitative methods for measuring the liquidity of fixed income securities, such as the Bloomberg Liquidity Assessment Tool, are emerging to help guide firms in creating objective rule sets around which internal policies can be formed. Bloomberg LQA is now available on the Bloomberg Professional service at {LQA}.
Let’s assume that 50MM XYZ of ‘46 is for sale. For the buy side seeking a bid, a liquidity score analytic can provide a view into what they are asking of their sell side relationship. Bloomberg LQA is a powerful analytic that provides holistic insight into liquidity risk management. For now we will focus on two straight-forward outputs of Bloomberg LQA – the liquidity score and the time to liquidation of a given position. These outputs provide two examples of how quantitative data can inform internal procedures. More advanced internal policies may look beyond these parameters to use more complex and customizable metrics such as the probability of executing at the firm’s reference bid price or better. For example, a firm’s policies can mandate traders account for probability percentages in their trading decisions, e.g. to offer a bid for a bond, the probability of executing at the bid price or better is greater than 50 percent.
Bloomberg LQA is built on a market impact model that uses machine learning to account for the lack of available trade data in the fixed income markets (Figure 2). The model considers a matrix of variables that describe a bond’s structure, regulatory classifications and its secondary market characteristics like Bid/Ask spread and volatility. Machine learning then dynamically uses these variables to “cluster” bonds into nearest neighbor families. Trade data for bonds within the same family cluster are weighted and used to assess activity and reaction to market conditions for the bonds within the cluster. It is this activity from which the liquidity estimations are formed. For example, a families’ trade velocity and an individual bond’s relationship to the family can provide a basis for approximating demand at given prices and volumes to produce a “time to liquidation” estimation. Bonds can then be compared relatively to the rest of the population in a normalized fashion to provide a relative liquidity score of the target bond against the universe of securities. These two examples can provide the backbone for a structured, non-biased, repeatable practical implementation of the internal policies developed to comply with regulation. Let’s look at an example.
The model estimates a bond’s liquidity score based on a range from 0 (illiquid) to 100 (liquid). An institution may decide to use the score to guide traders to assess asset quality and segment client requests in a consistent manner across the firm. The firm may segregate assets based on the score out of 100 where internal policy would dictate trader bid/no-bid action. Consider the following groups/bands/categories:
- Group 1 – Score 100–81 – the bond is highly liquid, has a high velocity and the size of the inventory has a low anticipated holding (carry) period giving the trader (firm) confidence to allocate balance sheet and make a price;
- Group 2 – Score 80–66 – the bond is a less liquid bond but the trader is allowed to provide a bid to facilitate a trade with a “good” customer;
- Group 3 – Score 65–51 – may imply that the trader will facilitate the customer order by buying-half and working the other half. This type of trading – “getting my good relationship started” – is popular in equities and may apply only to more liquid fixed income issues. This is where the regulation may change the dealer capacity (role) from principal to agency (or riskless/matched principal);
- Group 4 – Score 50–21 – may imply that although the trader cannot position the asset into inventory, the firm has clients that may be interested in the asset so sales-traders will “shop” the bond; and,
- Group 5 – Score <= 20 – may imply that the dealer neither has the capacity nor network of clients so trying to find an interested counterparty would be not be an effective use of the firm’s resources.
After the initial liquidity quality assessment, the next regulatory hurdle that needs consideration is the composition of the inventory. In order for a trade to be considered “customer facilitation”, there needs to be an assessment of an approximate time to liquidation. Inventory carried past a certain holding period could be considered proprietary position. Moreover, in addition to conventional risk measures, institutions need to have a “target” liquidity profile of their inventory. This is very similar to an emerging buy side regulatory push that will mandate investment managers to maintain a buffer of liquid assets that can be converted into cash within three business days, and report an estimated time to liquidate their position in all assets without materially affecting the price of those assets. To this point, Bloomberg LQA provides institutions with corresponding position size and estimated time to liquidation (TTL).
Inventory can be segmented into buckets based on TTL. Highly liquid bucket could be defined by positions with TTL of 0 – 3 days; a semi-liquid bucket could be defined by TTL of 3-12 days, etc. and a rule could be enforced that traders would not be able to position sizes that are estimated to require greater than 20 days TTL (Figure 1).
The Complexity of Setting a Price
Assuming our above examples have been written into internal procedures, if the bond’s quality aligns with asset and inventory management groups of 1, 2 or 3, then the trader may provide a price.
What is the price?
In Part I, we discussed how the price of a bond is determined in part by the bond’s relationship to a relative value benchmark and perhaps carry costs. However, with Basel III’s new “haircut” regime, “bid this bond” is actually the buy side asking to rent the sell side balance sheet to facilitate execution immediacy. Because of this “haircut,” the use of the balance sheet has a direct cost associated with it because the “cost” of the balance sheet is at rates well above financing costs. With credit, if the haircut is 50%, then half of the notional value of the bond is being financed against the balance sheet rate of, perhaps 350bps. Bloomberg LQA’s TTL can then be used to determine the approximate holding cost of the bond for the firm and the corresponding price to the client could be marked down accordingly.
Because it is an objective data-driven measurement, Bloomberg LQA can help the sell side explain why they cannot position the bond, or why their price is at a discount to the relative benchmark. Moreover, it can provide the information to the front line (the sales trader) in near real time to enable them to calculate or corroborate the price as quickly as possible. Most, if not all of the above factors are being considered by various middle and back office functions already within firms; the key differentiator for market leaders is bringing this information seamlessly into the trade workflow before an order is executed.
We will talk about that more in Part III in our series.
