Friday, 29 May 2015

Benchmarking on the outliers.

There is often a big disconnect between what people expect to receive an what they actually receive. Management or client relationship attempts to stick an elastoplast over this difference is called ‘expectation management’. In the case of pricing financial trades a client will expect their fill to be at the market price and the stresses involved in managing expectations between what rate they expect and the one they receive are legion. But what is the market price?

Imagine a huge fruit market with all 1000 stalls selling oranges. There will be an average price at which they are all willing to sell oranges and it would be reasonable to expect that average to be the price you are content to pay and also a reasonable price to quote as the representative price of oranges at that market. Yet amongst all of those stalls one of them will have the best selling price and one will have the best buying price. So the best bid and offer are not representations of the average market price but instead are the extreme outliers of the price distribution only representing 1/1000 of the market.

Both the best bid and offer suffer ‘The Winner’s curse”

"The winner's curse is a phenomenon which can occur in common value settings—when the actual values to the different bidders are unknown but correlated, and the bidders make bidding decisions based on estimated values. In such cases, the winner will tend to be the bidder with the highest estimate, and that winner will frequently have bid too much for the auctioned item".

Expand that logic to the financial markets with millions of participants and we can see just how misrepresentative of the average market bid or offer those best bids and offers are. Yet everyone seems to expect to be able to deal their amount of interest on them and any deviation is a rip off.

It’s as if in that fruit market, someone turned up and demanded that as Manuel in Saville was giving away 3 oranges free from his bumper crop then the market should be expected to supply free oranges too, or at least the market stall owner expected to drive down to Spain at their own expense to buy those 3 oranges, drive back and hand them free to the demanding customer. Good luck if you try that at your local supermarkets, but that is a regular expectation in financial markets.

When it comes to benchmarking the absurdity goes one stage further with clients expecting to deal on the average of the best bid and best offer - the mid price. As we have just seen, the best bids and offers are the last extreme of the distribution of all bids and offers so the mid must therefore lie outside the universal set of dealable prices. Yet the mid is expected to be provided as the dealt price.

How refreshing it would be if instead of picking these extremes of the distributions for benchmarking we should average all the prices that the market is willing to sell at and the average of all the prices that everyone is willing to buy at to produce an average bid and an average offer. Absurd? Well it happen with all other statistical sampling methods. The Consumer Price Index doesn’t take the highest or lowest price to work on from all their samples.

This can be averaging order boards. Sum(Volume x Price) / Total Volume of order book to give the average volume weighted price of the bid side and the same for the offer. Repeat for all market participants (including all those Bureau de Change 30% spread prices) and your benchmark exchange rate suddenly moves from one of dealing impossibility to one of real world actuality.

What is more, fund managers would be absolutely delighted as not only do they have constant slippage against an unachievable benchmark, they now have a very good chance of outperforming it.

Now whilst I am proposing a mad logic testing alternative that I really can’t see happening, the key point I am trying to make is that the best bid and offer are not representative of the average of the market. They are virtual particles flitting in and out of existence at the very end of a huge distribution curve and to use them as benchmarks pretending that they represent reality is nuts.


Unclear Wessels said...

Only trouble is that this is so easy to spoof: pile on a zillion offers away from market and average offer would move up. Wait to get filled some, then ditch the spoof orders, causing the avg offer to move back down. You can now cross the market into this lower average offer that you've just caused to move down!

Polemic said...

UW - agree. there is no easy solution,

theta said...

a "solution" would be no trading all day long, but just one auctions per day, like the current opening process or closing auction. Everybody gets filled at the same price or not at all.