[noahc]

Can anyone with more experience and knowledge explain the effects of #3. I've read that HFT is actually a good thing because it basically acts as a market maker. This is just a back of the napkin opinion, but it seems that #3 would reduce liquidity in the markets.

We know that markets aren't perfect or else Buffet and other value investors couldn't survive. Does HFT make markets more or less perfect at pricing? Does this matter for making markets more efficient in reality, not some hypothetical perfect market?

[thwest]

Market liquidity is not the same as creating value. If you aren't investing on the same timeframe that it takes to create a business (3-5 years), you are simply exchanging money for money, not creating value. There is some minimum amount of liquidity needed for things to work at all, but massive amounts of liquidity is not good in and of itself. After a point market liquidity is a force for creating bubbles, not sustainable businesses.

[jarek]

The standard response to claiming HFT as a market maker is that it provides minute amounts of additional liquidity and its costs and drawbacks, or potential costs and drawbacks, outweigh the benefits. Of course some amount of market making is necessary, but you could claim that without HFT you'd just settle your trades in 0.1 s rather than 0.02 s and with 0.5% spread rather than 0.49% spread.

I don't know enough to form a proper opinion whether or not this is the case.

Of course part of the problem is defining HFT or algorithmic trading. If you build a robot to press an appropriate button at a trading terminal really fast, is that algorithmic, HFT, and should that be banned? Do you fix the tax at 10 cents, and if so how do you react if companies just create massive securities worth millions of dollars and trade those?

[mjtokelly]

According to some recent research, HFT appears responsible for lowering spreads by more than an order of magnitude: http://marginalrevolution.com/marginalrevolution/2011/10/mor... That's a definite social good that would go away under proposal #3.

[jarek]

Right, we'd have to get some concrete numbers for the drawbacks of HFT (if any) and then compare with the benefits. What's the societal value of having a $20 spread vs a $230 spread on a million dollar transaction? What is an acceptable trade-off spread value, how do we define it in the first place?

[mikeash]

You don't need to define HFT, nor do you need to play games with values. Set the tax at a raw percentage of the transaction's value. This percentage can be ludicrously low, like one hundredth of one percent, and still achieve the desired outcome. Such a tax will have no appreciable effect on "normal" trading, but would seriously diminish HFT.

I have no idea whether this would actually be a good thing or not, though.

[jacoblyles]

I haven't seen any thoughtful reasoning behind calls for a Tobin tax. It's mostly populists who are upset that some people make money by trading.

[prostoalex]

People usually put an equality sign between high-frequency trading and algorithmic trading. One is a subset of another.

There's also a subset of algorithmic trading that inserts large stop-loss orders (not much of an algorithm there, it's just done by computer program instead of an individual), which have been known to cause flash crashes.

Hence the confusion and belief that by eliminating HFT we'll eliminate automatic stop-loss orders.

[nhaehnle]

My totally abstract reasoning is like this: For markets to function well, the barriers of entry must be as low as possible.

The possibility of HFT raises the barriers of entry. Therefore it causes at least some bad.

Personally, I think the markets should just operate at, say, 15 minute heartbeats. Anybody is open to place orders at any point in time in secret. Then, at a predetermined heartbeat time, all orders are revealed and some standard auctioning mechanism is used to match them up against each other.

Sure, spreads would be larger than they are now, but at the same time there would also be less volatility by definition. More importantly, barriers of entry would be much lower, which will cause the market to function closer to a "perfect market" over time.

[bravura]

Why does a market need millisecond-resolution liquidity?

[noahc]

I'm not sure it does need that much liquidity. But, at what point are we not liquid enough? Is hourly liquidity enough? It's a good question, but I'm not sure how you measure what 'enough' is.

[argv_empty]

At some point you'll have to decide: "enough" for whom? For someone just trying to accumulate some money for retirement, even hour-level liquidity probably looks like more than is useful (and will his broker even act on a request he makes in less than an hour? maybe someone else can answer that one).

[ArchD]

It depends on how you draw the boundaries for an 'entity'. A sufficiently large company may have many departments trading in different ways fairly independently. If you are going to tell people, "You can't sell/buy sooner than X seconds after you buy/sell.", then large organizations will have a problem being nimble being hindered by regulations about how soon you can flatten a position. Even brokers like Ameritrade and e*trade will have problems acting on behalf of their clients because different individual client will want to trade differently, if you treat them as single organizations. What about dark pools? Are you going to regulate those as well? The trading venues are about people trading stuff. It's people acting independently out of free will.

[ArchD]

Down-vote and run?

[tsunamifury]

To allow for volume.

[guimarin]

Instituting #3 will lower liquidity. I think the trade-off is worth it.

This is a topic where the critics, who argue in favor of the status quo are at an advantage, because the realities of HFT are not easily discernable to the layman, and require a much deeper understanding of the entire modern financial ecosystem. Suffice to say, there were/are pros and cons to having a specialist.

[veyron]

"Suffice to say, there were/are pros and cons to having a specialist."

Let's not go there. Specialists were accused of favoritism: filling orders that certain traders submitted and ignoring requests from other traders.

More generally, the advancements, both in technology and in pricing, actually helped everyone. SOES (small order entry system) helped smaller traders to quickly trade. Decimalization really helped reduce transactions costs (spread between bid and offer is a sort of transaction cost, when we were dealing with 1/16th quoting, the spread cost was over 6 cents)

On another front, brokers used to screw people over by telling people they bought at one price while actually buying at a lower price. This is why Regulation NMS came about. And HFTs nowadays help keep all of the exchange prices in line, minimizing the potential damage of shady broker behavior.

[guimarin]

yes and sophisticated HFT systems screw large funds from moving into and out of positions with relative ease/fairness. This cuts both ways, thus pros and cons. At least with a specialist, everyone was relatively even in that regard. Now, measuring the distance to the server in micro-seconds of fiberoptic wires, I can literally create an unfair system by owning the closest supercomputer. I disagree with your point and think that my 'compromise' tax is pretty damn fair considering. Also, I'm not arguing about the merits of digitizing the exchanges, I'm arguing that there was a cost to removing the specialists, it's obvious now, and we should fix it. The specialists were bad in some cases, enough to get people to change, but the system is just as corrupt now as it was then, just differently.

[veyron]

"yes and sophisticated HFT systems screw large funds from moving into and out of positions with relative ease/fairness." <-- that's completely false. It is precisely because of HFT systems that the capital markets were reasonably resilient, and it is because of the lack of HFT in the current environment that prices swing on the order of 5% during the day

"At least with a specialist, everyone was relatively even in that regard." <-- again, specialists were accused of playing favorites. That's manifestly unfair.

"I can literally create an unfair system by owning the closest supercomputer." <-- That's also not true. It's not good enough to have the lowest latency connection. I can have a slightly slower connection but still have an edge. To put numbers to this, lets say that I was 100 microseconds slower than the fastest person, but I could predict the next price movement 150 microseconds faster (this is due to many factors, including the particular model I use and other implementation details). Then, I would still have a 50 microsecond edge.

"my 'compromise' tax is pretty damn fair considering" <-- this would absolutely screw large funds. As was mentioned by many others, if you tax transactions, market makers will quote wider spreads (and yes, spreads are transactions costs to the large funds moving in and out of positions)

"I'm arguing that there was a cost to removing the specialists, it's obvious now, and we should fix it." <-- NYSE parity (which is a vestige of the specialists) is broken and unfair. Why should someone who sent an order after me get filled before me? Most exchanges have a price-time priority concept, so that if we are sending at the same price but I sent first, then I should get my fill first. The specialists wrecked the idea back then and the current implementation is still unfair.

"the system is just as corrupt now as it was then, just differently." <-- it's actually much less corrupt. I recommend you go and try to build your own HFT, and then you will see that the system now is much more fair than you think. (shameless plug: I started blogging about my experience: http://news.ycombinator.com/item?id=2835656 )

[yummyfajitas]

yes and sophisticated HFT systems screw large funds from moving into and out of positions with relative ease/fairness.

Fact: when you make a large trade, there is a price impact. I.e., if you sell lots of GOOG, the price goes down.

Once upon a time, large funds could use their own sophisticated algorithms to make sure their unsophisticated counterparties feel the price impact. I.e., the big fund disguises their intent, Joe IRA buys from the large fund, and then the price goes down after Joe already owns the stock.

Nowadays, HFT makes the price impact happen immediately, typically splitting the difference with Joe [1].

Why is this worse?

[1] The HFT's try to outbid the large fund in order get ahead of it, resulting in Joe getting a better price.