Jun 26, 2012
The bigger picture
Mike Earlywine, head trader at buy-side firm Ecofin’s New
York office, explains why the current approach to pre-trade analysis used on
the trading desk may not always reflect intended investment objectives.
The TRADE USA: In your opinion, how has the approach to
pre-trade analysis evolved over recent years and where is there room for
further improvement?
Mike Earlywine: Pre-trade analysis
was initially about finding the trade that was too large to be efficiently
executed due to the average volume of the stock in question. It was about
alerting traders to the challenges in front of them and provided a feedback
loop between the trader and the portfolio manager. Orders that consisted of two
or three times average daily volume prompted discussions about the cost of implementing
the trade, time span for the trade, near-term catalysts, and to some degree,
the risk associated with the trade.
With the evolution of program
and list trading, pre-trade analysis focused on finding the one or two names in
a list that would slow the completion of that list or program. Staying
dollar-balanced and maintaining the risk profile of the list were techniques
used to avoid unintentional bets. Liquidity analysis and projections influenced
the trading technique and eventually led to recommendations for techniques and
timing of trades that optimised cost and also minimised opportunity costs.
My concern is that this process
doesn’t go far enough. It needs to incorporate the intent or context of trading
activity with the implementation optimisation, i.e. knowing how a single trade
can influence the risk associated with an entire portfolio. As traders we have
lost our way in the quest for the lowest-cost implementation strategy.
There is nothing wrong with
trying to minimise friction in the implementation process – in fact, it is a
requirement and a necessary goal for all modern equity traders. But the
obsession with reducing friction costs has obscured the intent of the trade.
The TRADE USA: So what can traders on the desk do to gain
a better understanding of the contextual impact of a trade?
Earlywine: If we adhere to modern portfolio theory (MPT), then all trading is
done to facilitate movement along the efficient frontier through the addition
and deletion of risk (our goal is to maximise return while minimising risk).
Therefore, in deciding the types
and extent of resources to devote to each trade, understanding the impact on
the overall portfolio (risk on vs. risk off) is as important as the cost of
implementation. No matter what the thesis of a trade, no matter what the alpha
profile of a trade, the net result is an increase or decrease to the risk of a
portfolio.
Although great progress has
been made in improving the sophistication of the pre-trade analysis process as
well as its functional content, I would argue that the evolution of pre-trade
analysis has focused too much on the execution side of the equation and not
nearly enough on the risk and portfolio management side.
A concept that I have been
working on over the past couple of years involves incorporating Value-at-Risk
(VaR) analysis into my pre-trade routine.
For the uninitiated, VaR is a
measurement that seeks to estimate what an extreme profit-and-loss (P&L)
event could look like for a portfolio or an individual position.
Risk management technology
provider PortfolioScience have done more than speculate on this concept and
have teamed with TradingScreen to make dynamic VaR analytics a working part of
the pre-trade analysis process. I am not a client of Portfolio Science, but through
their collaboration with TradingScreen, of which I am a client, I was able
to put some of my theories to the test. Having spent some time evaluating the
Trading Screen/PortfolioScience product, I believe more than ever that VaR
analysis at the point of execution provides an important and previously missing
element to the pre-trade process: context.
The TRADE USA: How does the added knowledge on the
overall impact of the trade impact the buy-side trader’s priorities?
Earlywine: By monitoring a
trade’s contribution to the portfolio VaR, I am able to prioritise and make
trade-off decisions between cost of implementation and the net effect of
overall risk to the portfolio. I still look to minimise implementation costs
but I do so within the context of the individual risk contribution of each
trade and the net effect of portfolio risk due to a combination of trades.
Using risk metrics like
portfolio VaR and VaR contribution, I am able to trade in concert with the
cumulative intentions of all of the day’s trades. I can focus my efforts and
resources on the trades that will affect the portfolio’s risk profile the most,
not just the executions that are the most expensive to implement. Sometimes
these two types of trades will be the same, but in my experience they are more
often than not different.
Knowing the impact of a trade on
a portfolio’s VaR can help traders to better understand the intentions of their
portfolio managers and facilitate smoother dialogue.
The TRADE USA: How do you incorporate the measurement of
VaR with more traditional methods of pre-trade analysis that only look at
implementation costs?
Earlywine: My ideal form of
pre-trade analysis would include a VaR measurement and a specific individual
contribution to VaR along with the more traditional cents per share execution
cost number.
The measurement is typically
expressed in dollar terms, such that the portfolio with X dollars of VaR
implies that a future P&L event of size X dollars is potentially likely. My
goal is to understand the overall effect of each trade on a portfolio – specifically,
how each affects the movement along the purported efficient frontier curve
(have we added risk as we seek return, and so forth).
The movement along the curve
tells me if we are looking to add or subtract risk. It sets the tone and gives
context for trading. Are we adding to or reducing risk in the aggregate?
Measuring potential trading costs in the context of the net effect of
additional portfolio risk allows for a comparison of implementation costs on a
relative basis.
For example, is a
20-cents-per-share (cps) implementation cost too high? It might not be if the
result is a reduction in portfolio risk by a significant amount. Likewise,
there is an argument to be made that five cps implementation can be quite
expensive if the trade does nothing to add or reduce risk from the portfolio.
The TRADE USA: How do you practically use measures of VaR
at Ecofin to influence the execution process?
Earlywine: Below is an example
of how we use VaR to triage executions. The table details a series of orders
that we recently traded over an afternoon.
While not a perfect measurement,
the stock prices were similar enough in prices for our purposes. Of the four
trades A-to-D, C is the most expensive to trade at five cps, but it is not very
efficient in adding risk to the portfolio. Five cents of impact only buys you US$660
worth of potential additional P&L. Trade B, on the other hand, is
relatively expensive at 2.6 cps, but buys five times more VaR.
Trades A-to-D were part of a ‘risk
on’ strategy going into a short-term market turn. A and B were ‘add on’ trades
to existing positions, while C and D involved covering a low-conviction short
and adding on to a higher-conviction short. Trade E was an opportunistic sell
at the close of trading. We had a mature long position nearing a fundamental
price target that spiked higher on no discernible news (as a rule, we try to
take advantage of these moves and sell into them as part of the profit-taking
discipline).
To summarise trades A-to-D: We
added US$471,000 worth of net long exposure and US$11,000 worth of risk (or
potential P&L move) to the portfolio.
While this is not an exhaustive
study, I think it shows the potential for using VaR at the execution level. By
providing context to the purpose and effect of each trade, the trader is able
to take that “intent” into account and make a much more informed cost/benefit
decision. Looking at VaR in terms of cps is an interesting concept and one that
is worth looking into further. MPT tells us we can’t have return without risk,
but looking at trades this way could highlight the cost of adding or reducing
that risk.
However, because measures like
VaR are often divorced from the implementation of a trade, combining these
measures with electronic execution tools can be challenging. The majority of
algorithms are based on the historical liquidity curves of the securities they
target, which can be out of synch with the need to schedule a trade based on
its VaR in relation to the entire portfolio.
Anish Puaar
+44 (0)20 7397 3817
anish.puaar@thetrade.ltd.uk