Recently, assiduous reader Drew had some questions about the volume-average calculation of the recent new issue CFS.PR.A.
The HIMIPref summary screen indicates that the average trading volume of CFS.PR.A is 119,000 shares, well in excess of its volume over the recent past. Could you please explain this. I am guessing that higher liquidity is positive for valuation and vice versa. If this guess is correct, what sort of impact would the actual trading volume of CFS have on its valuation score?
…that would require an average trading volume of 10000 shares. Unless my memory is deceiving me again, it does not trade anything like that volume on an average basis. Today’s trading volume of 2000 seems closer to average, if a little high.
To understand the design decisions that have been made, it is necessary to understand the two major uses of volume-average in the HIMIPref™ system (which are fed into the model via averageTradingValue:
- the averageTradingValue is used to create a liquidityMeasure, that quantifies the dollar value traded vs. other issues. This measure (which is capped) is fed into the yield curve calculation and the yieldCurvePremiumLiquidity is calculated. In other words, it is considered that there is a spread to the curve assigned to the marketplace for liquidity, just as there is a spread assigned for credit ratings and everything else. When the curve gets re-applied to the instrument’s characteristics to determine the curvePrice of the instrument, the dollars-and-cents value of the instrument’s liquidity then becomes part of its valuation.
- During simulations, it is assumed that a portfolio can sell one-half of the volume-average at the bid price, or buy this quantity at the ask price.
For many years, the first point was moot. I was unable to discern any premium paid in the marketplace for liquidity – every day’s premium was 0.00%, or at at most one or two bp, the effect of mathematical accidents more than anything else. This changed a few years ago. Hesitantly at first, and then with increasing ferocity, the mathematical model started to assign a premium to liquidity which now stands at about 17bp per liquidity unit; the range of values of the liquidity unit varies between -1 and +2. Hence, liquidity has become more important to the marketplace.
Now, what of new issues? When issues are announced, they are valued on a preIssue basis, with the volume-average deemed to be 100,000 shares. This figure is fairly arbitrary – all I can say is that it seems to work pretty well, most of the time. As soon as the the issue starts trading, the volume will decline below this figure after a week (usually!) and the value for volume-average (and hence, the value of the liquidity curvePriceComponent) will decline with it. This will basically take the path of an exponential moving average – see the various links in this post for an explanation.
The trouble with CFS.PR.A is that it turned out to be a ridiculously small issue – only 1,610,000 shares were issued (virtually all to retail, I’ll warrant) – so the 100,000 estimate of average trading volume was very, very high. The daily volume of the issue (see graph) hasn’t been much at all and hence the decay of volume-average has been a relatively smooth curve (graph) that still hasn’t declined to a reasonable estimate of what the volume actually is.
The system works better with issues of normal size and trading patterns, like the recent new issue SLF.PR.E. I have uploaded a graph comparing volume-averages, as well as a graph of the SLF.PR.E daily spot volumes.
So … CFS.PR.A does (as far as I can tell) have a volume-average (and hence a valuation) that is over-estimated by HIMIPref™. That part’s easy. What’s more difficult is deciding what, if anything, to do about it. If I get any ideas, I’ll programme them!