If the decline in the trading volume was not enough, credit traders must now face up to their obsolescence. The brave new world is here and pithily captured by FT Alphaville’s headline “Culled UBS traders replaced with algos”.
The incessant march of technology cannot be halted and it was only a matter of time until the machines forced their way into the OTC-trading bastion. Moreover, two developments have greatly accelerated the rise of the machines. First, the credit crash exposed the myth of the alpha generating trader. What many banks considered superior trading ability was largely found to be some variant of a leveraged carry trade. Second, post-crisis regulation pushed for greater transparency, squeezed profit margins and increased capital requirements on trading positions. The pre-crash business model of taking enormous (albeit hidden) risk with borrowed money is no longer tenable. The new object is to match buyers and sellers, taking a cut from both without risking much capital. In this, machines are cheaper and faster than humans as demonstrated by the domination of HFT algorithms in equity markets.
Large desks manned by scores of highly paid traders and salespeople were anachronistic especially in commoditised OTC credit products. When a client wishes to trade a particular bond or CDS, price is usually the only determinant for a trade. In the process, there is no value added by salespeople who take the client’s call, relay the message to the trader and relay back a response. The value added by a trader is only realised when the bond/CDS position is closed out through an offsetting trade. This can be negative if the market moves against the position in the interim (Therefore traders demand a margin to compensate for this risk and hope to convert it into actual added value). The level of competition in the market impacts the margin and thus the value realised by the trading operation. The example below makes it clearer.
Consider a buyer of bond A, who wishes to pay $99 for it. Unknown to him a seller B is ready to sell at $98. Now if there is only one bank who knows both the buyer and seller then it can make $1 profit by buying from B at $98 and selling to A at $99. However, if after buying from B, some news causes A to revise his bid to $97 then the actual value added is negative $1. Therefore the less time that elapses between the two offsetting transactions, the better it is for the bank. Now presence of another bank which also knows A and B will reduce the margin demanded due to each bank’s incentive to win the trade by undercutting the other. Not only does it reduce their realised profit from $1 but it also increases the potential loss. If a bank buys at $98.5 hoping to sell to A for $99 but news leads to a revised bid of $97 then the loss is $1.5 not $1 as it would have been earlier.
Before the crash, excluding the innumerable smaller firms, there were more than 20 large banks all clamouring for the same business. The obvious result was zero margins in market-making. These were supposed to be compensated through proprietary trading which is no longer an option.
To make money in market-making requires maximising the speed of trade execution to avoid being undercut and minimising the delay in closing out trading positions to reduce the risk of deleterious market moves. This is exactly how HFT algorithms operate in equity markets and even though the implementation might differ, there is no reason why market-making fundamentals should differ for OTC credit markets. Electronic platforms which enhance the speed of trade execution have already been adopted by most banks. They obviate the need to maintain an expensive army of salespeople whose main job is to intermediate between clients and traders. It enables the bank to focus on the few salespeople who actually add value. This is a move towards corporate banking type relationships where a few people have extensive and strong relationships with a firm's senior managers ensuring that they at least get asked when there is business to be done. The next logical step is disbandment of the expensive army of traders which the introduction of algorithms integrated with electronic platforms would enable. This would result in lower bid-offer spread (i.e. advertised margin) and quicker post-trade price adjustment. Consequently the probability of capturing client trades would go up reducing the delay in offsetting trading positions. Human traders will still be needed but as in sales, far fewer than earlier and of a higher calibre than average.
The gloom over employment following the UBS cull is justified since not only is banking undergoing a structural change but artificial intelligence is fast replacing human intelligence. However there is no need to start the search for John Connor. A chosen few will still be needed by banks. More importantly for those who have real talent there is nothing to fear from a rise of the machines. Even though machines have an edge in market-making, long-term trading or investment is a multidimensional field where human intelligence beats artificial intelligence hands down. Ask a computer to compute the probability of eventual Greek default and how best to express that view and chances are that a blue screen will pop up. Opportunities abound in the current market and the financial turmoil has led to a general thinning of ranks (leveraged coin tossers and momentum chasers are dying out) which has made pricing anomalies more numerous, longer lived and more profitable. Even better, the rise of the machines means that execution becomes easier and cheaper. So raise a toast to the machines and follow the one consistent and proven winning strategy: that of using the machines to one’s advantage.