[dbrower]

Can someone explain why any of these is a bad idea? (a) completed transaction tax; (b) regulatory fee on offers/cancellations; (c) insertion of random delays into offers/cancellations;

All could increase friction and reduce the speculative/arbitrage opportunities, while having little effect on those wanting to trade to hold for periods exceeding seconds.

There is a belief that the churn of HFTs/Arbs is enhancing liquidity for "real" investors. There ought to be reasonable questions what amount of churn is useful, adequate, and whether some frothy levels should be constrained in some way.

Is there any way to decide when things are excessivly liquid, in ways that lead to undesirable effects?

[kasey_junk]

Any tax you propose will reduce the amount of the taxed thing and introduce unintended consequences. So we have to ask why do you want less transactions or orders? It won't reduce HFT activity for instance. It will just mean that HFT systems will only make more profitable trades. That means higher bid/ask spreads and higher risk limits leading to higher volatility. It also may have the unintended consequence of consolidating more volume into smaller firms.

That doesn't seem to be in anyone's best interest.

As for latency games the issue is not the overall latency it is fifo priority matching. Without changing that bouncing trading signals off of mars won't help.

[dbrower]

I think a reasonable question is why we consider /more/ transactions a good thing, if a large fraction of them are for holding periods in small number of seconds.

I think I am questioning the fifo paradigm, which creates these arbitrage opportunities, especially when there are multiple fifo queues representing multiple markets. It is not clear to me that batching things in 1 sec increments, and randomizing the our ordering would be bad or unfair.

I also don't see why a modest fee that would make short-hold transactions for tiny gains is a bad thing.

Structuring the system to reward HFT latency advantages seems opposed to stability, if one believes that the market is for actual investments.

HFT seems to be a second order phenomenon that games the system, and may have come to dwarf what could be called legitimate investment.

At what point is there "enough" liquidity, and when is "too much"? I suspect the people who do HFT and other arb techniques think there is no such thing as too much, because they profit on the churn. Others see this as producing nothing of societal value, extracting real money from the system that could be used for other purposes.

[kasey_junk]

"I think a reasonable question is why we consider /more/ transactions a good thing, if a large fraction of them are for holding periods in small number of seconds"

I don't want to put words in your mouth, but can we at least say that transaction count is not what we are actually interested in? In a vacuum we (larger society) don't care a whit about the number of transactions that occur?

"It is not clear to me that batching things in 1 sec increments, and randomizing the our ordering would be bad or unfair."

It is clear to me that batching things in 1 sec increments would be as unfair to someone as the current system (I'm sorry to not back this up, but I've been doing a ton of commenting about it in the last few days and don't have the patience I did). It's possible that this unfairness is "better" for society overall. I don't know how to quantify that.

"I also don't see why a modest fee that would make short-hold transactions for tiny gains is a bad thing."

My "expert" opinion on market making strategies is that this would increase the bid/ask spread and volatility in the market. Further my opinion is that this is a bad thing for retail investors and market makers, to the advantage of large institutional investors. I don't have the stamina to prove this assertion.

"Structuring the system to reward HFT latency advantages seems opposed to stability, if one believes that the market is for actual investments."

This is a common simplification that I think deserves a lot of thought. Many folks on Hacker News think that the market is for "investment". Usually they equate this to "bootstrapping enterprises that aren't viable without external money". This is an obvious bias for an internet forum dedicated to startups to have. In reality, the vast majority of market forces are not about that. They are about risk mitigation. So when you say that HFT "dwarfs" legitimate investment, are you saying it doesn't provide a valid mode for bootstraping enterprises, or are you saying it doesn't help with risk mitigation?

Your answer to that question means a lot for how we debate the topic. A similar answer is available for your assertion that HFT is a second order phenomenon.

As for when is there enough liquidity? I'd say when people stop paying for it. Which hasn't happened yet. Paying for liquidity is reaching a saturation point, because it has gotten so cheap. This is most obviously demonstrated by the massive loss in value that HFT groups have had in the last 5 years. If anything we have too much liquidity, and that comes from someone who's paycheck comes from an HFT.

[dbrower]

Thanks, I appreciate the answer, and am trying to be more informed.

On transaction count, I think "we" /may/ care about transaction count, if the majority of transactions are shams or gaming the trading system. To the extent that the value of the transactions turns up in statistics like GDP, then they distort "our" view of economic activity. Perhaps that is a problem with metrics that include such transactions rather than their existence.

Let's table discussion of batching or randomization.

In terms of volatility, I will proffer that HFT and program trading go hand-in-hand in my mind, and that program trading at high frequency seems to be a volatility amplifier. It's not yet clear to me why added damping friction automatically leads to wider spreads and volatility as you indicate. I accept I may be uninformed on the matter.

When I speak to "investment" I think more of capital gain and dividend than of capitalization from public offering. Over not long periods, the value of shares traded "long" usually greatly exceeds the value of the offering. That is the majority of "investment", and doesn't really relate to bootstrapping startup-ness.

I think there is a fair discussion to be had about amounts of liquidity and risk mitigation. Some reasonable questions are about whose risk is being mitigated, at what cost to whom else, whether all affected parties are willing participants or if they are left no viable alternatives.

The argument is that the "risk" being mitigated is a derivitive of "true" investment, and should be of a lower aggregate value than the underlying security. When the mitigation is exceeding the value of the asset, then that suggests the system is out of balance. Relate to aggregate CDO valuations exceeding the value of the underlying securities.

I'm not saying I accept that argument, or reject your point about risk-mitigation. I'm trying to understand the forces that should balance, and how one might judge them.

I'm not sanguine about the claim that payments for liquidity are proxies for the will of an informed market, since I'm not a Chicagoan who thinks the market is always right by definition. There can be market failures, and it seems possible that "over liquidity" may represent a market failure.

thanks.

[kasey_junk]

On the transaction count issue, I agree that there may be some societal good to removing sham or gaming trades from the markets, but that is not what a transaction tax targets, it targets all transactions whether they are shams or not. Further a transaction tax raises the cost of entry for all market participants, without providing any disincentive to those making sham trades.

It's possible that HFT is a volatility amplifier, it's hard to judge though because HFT rose when other macro market forces were driving huge amounts of volatility into the markets. One thing I think most people will agree on is that market segmentation and electronic access drove down the price of trading dramatically for everyone and enabled a variety of extremely useful investment vehicles for the "retail" investor (I'm thinking of ETFs especially). Any system that has market segmentation and electronic access will also enable algorithmic trading. My opinion is that it is foolish to throw out all the positive benefits in the name of some artificial "fairness" between human and computer traders.

Chris Stucchio has a good explanation about the mechanics of market making at http://www.chrisstucchio.com/blog/2012/hft_apology.html

Understanding those mechanics we can see that any added costs that the market maker bears must be reflected in their trades and will eventually drive them to increase their spread on the order books. If enough of the market makers do this it will increase the spread which is the biggest driving cost in trading after operational costs.

Capital gains and dividends are a mechanism to make capital aquisition possible. They happen to be the attributes of the capital markets that you are most interested in but there is nothing sacred about them (in fact lots shares lose capital and many companies don't pay dividends). For many other market participants hedging against inflation risk is vastly more important than dividends for instance. My opinion is that we should enable a system that allows for as open of access as possible to these markets for as cheap a cost without bias towards whats most important to any class of market participant. It is also my opinion that HFT actually does these things and that is why they are so scary to hedge funds and big banks.

I am also not a Chicagoan (well in an economic sense) and you are right over liquidity may be some sort of market failure. As someone who sells liquidity I will tell you that the price of it is currently being propped up with legislation. If it weren't for the sub-penny rule the spreads on some commonly traded instruments would be much smaller than they currently are. I'd also say if you figured out a way to remove some of the liquidity in the system without increasing my costs I'd appreciate it as this whole competing in an efficient market thing is for the birds.

[dbrower]

Thanks, I appreciate the discussion.

[harryh]

Your stamina on this topic is impressive kasey_junk. I've enjoyed reading your comments in the face of confusion.

[kasey_junk]

Thanks for the vote of confidence.

To be honest I'd much rather talk about the pro's and con's of strong type systems or whether you can model semi real time systems on the JVM than I would the relatively uninteresting questions about FIFO matching in modern exchanges. But man people say some really ignorant things as soon as equities trading gets involved.