The results from the triennial BIS forex survey are out. Unsurprisingly, trading volume continues to rise, the dollar retains its dominance in forex transactions, and the dollar/euro currency pair is the most heavily traded. But, notably, algorithmic trading is on the rise.
To me, the most interesting trend is the increasing importance of trading on electronic systems, as opposed to execution by other means.
Source: BIS, Triennial Central Bank Survey: Report on global foreign exchange market activity in 2010 (Basel, December 2010).
An important structural change enabling increased FX trading by these
customers is the spread of electronic execution methods. Electronic trading
and electronic brokering are transforming FX markets by reducing transaction
costs and increasing market liquidity. These changes, in turn, are encouraging
greater participation across different customer types.
Continued investment in electronic execution methods has paved the way
for the growth of algorithmic trading. In algorithmic trading, investors connect
their computers directly with trading systems known as electronic
communication networks (ECNs). Examples of ECNs in FX markets are
electronic broking systems (such as EBS and Thomson Reuters Matching),
multi-bank trading systems (such as Currenex, FXall and Hotspot FX) and
single-bank trading systems. A computer algorithm then monitors price quotes
collected from different ECNs and places orders without human intervention
(Chaboud et al (2009)). High-frequency trading (HFT) is one algorithmic
strategy that profits from incremental price movements with frequent, small
trades executed in milliseconds.
See also their VoxEU post.
In a 2001 paper, Yin-Wong Cheung and I documented the early and rapid transition to electronic platforms, from 2% to about half going from 1993 to 1998. Since the samples are not comparable, we can’t say what has happened going from 1998 to 2010.
The issue of algorithmic trading is of interest to me, if only because we haven’t completely figured out the origins of the May 2010 “flash crash” in the equity markets. 
Source: M. King and D. Rime, “The $4 trillion question: what explains FX growth since the 2007 survey?” BIS Quarterly Review, December 2010. (Basel, December 2010).
King and Rime attribute a large share of the growth in trading volume (particularly in spot transactions) to high frequency trading (closely linked to algorithmic based trading):
The growth in electronic execution methods in FX markets has enabled
algorithmic trading. Algorithmic trading is an umbrella term that captures any
automated trades where a computer algorithm determines the order
submission strategy.13 For example, FX dealers use algorithms to
automatically hedge risk in their inventories or to clear positions in an efficient
manner. Customers are increasingly using execution management systems
that break up trades and seek the best market liquidity to reduce market
impact. Hedge funds and proprietary trading desks use algorithms to engage in
macro bets, statistical arbitrage or other forms of technical trading. All these
activities are contributing to the increase in FX turnover.
Market estimates suggest HFT accounts for around 25% of spot FX
activity. While many commentators suggest much of the growth in spot turnover
is due to HFT, the contribution of HFT to the increased FX turnover between
2007 and 2010 is not known with precision (Hughes (2010), Lambert (2010)).
Neither the Triennial data on counterparty types nor the data on execution
methods identify HFT. This estimate therefore cannot be verified.
More discussion in M. King and D. Rime, The $4 trillion question: what explains FX growth since the 2007 survey?” BIS Quarterly Review, December 2010. (Basel, December 2010). For me, one interesting question is whether this will make the forex market more volatile (at high frequencies). One study (cited in King-Rime) says no.
[Update, 8:44am Pacific] On a side note, my paper with Michael Moore (Queen’s U. Belfast) incorporating order flow data into a macro model of exchange rates will be presented at this week’s AEA meetings.