There are a lot of really bad ideas about there about how banks and financial markets work that seem to be working their way into financial reform. The notion that "Too Big To Fail" is preventable by breaking up large firms is one reasonably stupid set of ideas, as outlined in a prior post (here: http://tfideas.blogspot.com/2009/11/too-big-to-fail-stupid-concept.html). This post is about high frequency trading.
There is a large contingent of people, Wall Streeters and otherwise, who claim that high frequency trading is a good thing for the market because it increases market liquidity. They use this argument to claim that HFT shouldn't be regulated or taxed.
The problem with this claim is that high frequency traders buy and sell stocks that were posted for trading anyway.
Think of it this way:
Say I have 10 shares of Microsoft, and I'm willing to sell it for $80 (my "ask"). If high frequency traders came in and bought my $80 shares, set them at $80.02, and then flip them moments later to someone who came into the market at $80.02, they didn't add liquidity to the market. All they did was take $.02 from someone who would have been able to buy it at $80 anyway - my shares were already there. This has big implications.
For liquidity to truly be added, there must be a deep pool of equity from which I can transact. Microsoft is a large company, so there is lots of equity to transact - If I'm in at $80, there are probably a thousand other sellers willing to come in for $80, $80.01, $80.02, $80.03, etc. at any given time, and a thousand buyers on the other side of that willing to buy for $79.98, $79.97, etc. A small company, (call it Mini-Microsoft) however, probably won't have that many people or shares selling at a given time - if I'm willing to sell my 10 shares for $80, there may not be anyone else willing to sell til 10 more shares at $82. To buy more shares than I have to offer, you have to pay a lot more ($2) for the next batch of shares. They have an illiquid market for their shares.
Thus, more liquidity requires more people willing to participate in the markets to buy or sell the shares and stick with the decision - in other words, something that increases liquidity actively adds shares to the market that wouldn't have been sold otherwise. The small company needs new shares at $80.01, $80.02, etc.
The problem is that high frequency traders don't add any additional shares to the market - they serve as a pool of shares that had already been posted. So for the small company above, a high frequency trader comes in and buys my 10 at $80, sells at $80.02 and pockets profit. Someone else comes in and buys that 10 at $80.02 and sells for $80.05. So the process goes. The buy price follows from $79.98 upward because high frequency traders know they can buy for that price and then repost a few cents higher. The whole process stops whenever you hit enough liquidity that you can't keep marking your prices up a few cents and reselling, so in this case, if there were more shares at $82, then you can only keep going until my 10 shares hit $82. If you try and post at $82.01, you won't sell because $82 is available. In order to amass a position larger than the original $80 amount, you'd still need to jump up to the $82 price - you just pay more for the first batch of shares.*
So in the small company example, if you're a buyer of Mini-Microsoft and you want to buy 20 shares, if there's no high frequency trading, you can buy 10 at $80 and then you buy 10 at $82. You paid an average of $81 per share and got your shares. If there is high frequency trading, and you can't act as quickly, then you have to buy all 20 at $82 - you paid an average of $82. All 20 shares are gone in both cases. The HFT algorithm pocketed $20 for the bankers that came out of the pocket of real investors. There are no more shares out there than there were... HFT algorithms just vultured what you would have had.
*Two points. Firstly, this exact process can work on the short end too - I can short a stock downwards to the buy price. Either way, the original spread is artificially narrowed without any additional shares. The other point is that if new shares come in below my original 10, say at $79.99, because something bad happened and the value of the company dropped, then the HFT people would have to mark down to the new price... but the HFT people haven't added any shares, they've only flipped them at a loss that I would have taken. Empirically, losses for good HFT algorithms happen less often than gains, which is why they do it at all. The net gains made by HFT come straight out of the pockets of investors.)