What matters to an investor when they are choosing assets to invest in? Risk-return is the most obvious trade-off to balance. One can narrow down by asset class and sector, dividing up to achieve diversity (or an illusion thereof).
The memory of the latest crisis still being as fresh as it is, many investors are focused on the liquidity component that sits under the broader category of risk. How fast, and how efficiently, can an asset be cashed in?
Liquidity metrics are, however, very tricky. For example, over-the-counter derivatives can’t be measured the same way as exchange-traded cash products. The availability of potential data inputs varies widely.
In addition to that, one could have many a drunken conversation of what liquidity even means. Lob onto that the tendency for liquidity metrics to be calibrated with only the recent past, and some users may find the numbers to be bad indicators when they are needed most — in a crisis, the shape of which is unlikely to be known in advance.
All of which didn’t keep strategists at Citi from having a crack at it in a note published last week. The dataset in question is a collection of 6,000 TRACE-eligible bonds in the US debt market. TRACE is a tracking system for over-the-counter bond trading, started in 2002.
The analysts at Citi piled five factors into their model to measure liquidity, employing a 60-day horizon:
1) Absolute number of block trades. These are defined as a trade size of over $5m for high-grade bonds, and $1m for high-yield. Once they have the count over the last 60 days, they logarithmically rank them within their own sector (high-grade or high-yield).
2) Consistency of block trading activity. Consider the first factor more closely — if there were, say, 45 block trades in the 60-day window but they were all on one day, that doesn’t exactly shout liquidity! If, on the other hand, the trades were spaced out, that’s probably a better sign. This factor also involves a logarithmic ranking.
3) Client trading activity. This is where the Citi strategists start to get creative, and where it also becomes apparent that they are building this liquidity measure for their clients. Here they count up the number of trades that actually involved a client rather than just the Street selling to the Street, i.e. client-to-dealer rather than dealer-to-dealer. Additionally, they adjust this variable by factor 1) above.
4) Market balance. This is a measure of the amount of two-way activity seen in the client trades of the previous factor. If there are as many client buys as sells, then it’s more likely that trades can be executed without moving the market. This factor is also weighted by 1) above.
5) Volatility or something. We’re not exactly sure what’s going on with this one. Here’s what they write:
To capture this component we use our Quantitative Strategy team’s volatility model, which essentially isolates the amount of each issuer’s spread that can be attributed to mark-to-market risk (Market-Implied Default Probabilities: Update dated September 9, 2011).
How much spread is because of mark-to-market risk? Well, isn’t the mark calculated off of the spread? It’s possible what they mean is the model divvies up the spread observed in the market into implied probability of default and everything else, where the everything else is substantially down to ‘compensation’ for volatility and liquidity, furthermore affected by rates, expectations, etc, etc.
In general though, lower volatility is loosely associated with more trading, hence potentially more liquidity. Here’s a scatterplot of the data — to demonstrate how little this fifth factor is likely to matter compared to the previous four more than anything else:
Once all five factors are plugged in, using 60-days of data up to August 29, 2012, the scores look a bit like this (just plucking out a sample of GE bonds with similar maturities):
The Citi strategists use the above chart to illustrate how much variation in liquidity their model finds for just one maturity bucket of a single-issuer.
This is, of course, where it becomes most important to note that the above is just a point-in-time estimate, valid as of August 29, 2012, using 60-days of (summer!) data. Perhaps more interesting would be to see how liquidity evolved over time for these bonds, how they performed in a stressed market, etc.
Whether this matters to the investor at all depends on who the investor is. Again, it’s all incredibly subjective.
Perhaps the more interesting message from this particular effort (and the Citi strategists are welcoming comment on it) can be divined through a great deal of aggregation — by sector, that is.
End result? Graphs like this one (click to expand):
On how to interpret the bubble colours and sizes:
A green dot reflects a positive fundamental outlook for the sector relative to consensus expectations, red a negative view, and the larger the dot the more conviction we have on our outlook. We use a +2 to -2 scoring system, with a +2 representing a very favorable fundamental view and -2 the opposite.
So big green dots are good and big red dots are bad, fundamentally-speaking. The arrows are indicating recommendations for sector shifts depending on whether one values liquidity or could sacrifice it for some pick-up in spread.
Too add liquidity:
Many investors have a bias toward owning more liquid bonds, and for such an investor one rotation that jumps out is a shift from REITs into Telcos (we acknowledge the completely different business models of the two sectors). First, from a fundamental perspective we strongly believe that telecom has a more favorable outlook relative to consensus expectations than the REITs sector (+2 for telecom vs. -1 fundamental score for REITs).
In addition, this rotation would increase liquidity by 18 percentage points (score of 26% for Telcos vs. 8% for REITs) and while there is a nominal spread give up of 27 bp (184 bp vs. 211 bp), the spread difference is at multi-year tights (Figure 7).
To sell liquidity:
In the current low-yield environment other investors are looking for opportunities to sell liquidity in an effort to boost yield and meet investment bogeys. For such an investor, consider a rotation from banks into insurance sector. Insurance is not a terribly liquid sector (10% vs. 28% for banks), but we like it better from a fundamental perspective and it is a lower-beta sector. And importantly, one gets paid for giving up liquidity in this space. On average insurance trades at a similar spread as the typical bank (209 bp vs. 207 bp), and the spread difference between the two sectors is at the one-year tights.
But given how much diversity there is even for a single maturity bucket of a single issuer, we’re not sure how useful this is going to be. Liquidity metrics, ladies and gentlemen! Ever the holy grail of market data…