And what is the big-picture economic difference between getting rid of arbitrage opportunities with a millisecond (in the old times) or within a nanosecond? Because that's what we are talking about, not about inefficiencies remaining for hours.

There's a hidden assumption in your statement.

AIUI, you think that HFT has driven various numbers far down, the rest of us benefit from them being driven down, but don't benefit significantly from them being driven as far down as HFT actually has done. Right? I can agree with that.

The hidden assumption is that if HFT hadn't done that, then something else would have come along to drive those numbers down, but not as far. If so, then the benefit to the rest of us would have been much the same as it actually is.

You may be right, but I don't see what that "something else" could be. It would have to be something where some actor benefits from quick reaction and low spread, right? AFAICT, that must be either HFT or something indistinguishable from HFT. Something with the same traits, blamed by the same people for the same reasons.

You can impose a floor beyond which lower latencies no longer provide an advantage by replacing the current continuous auction market with a periodic, blind auction.

That is, have a market where orders accummulate in the order book and every, say, 15 seconds, the order book is crossed and trades are made.

You can do a lot of things with norms and regulations but, in the end, why not let the market organize itself? Also, timing (as per your example) is very very very difficult to keep exactly.

On the other hand, I think a kind of relation between prize of an operation and its energetic cost would be beneficial (like ---as an example--- forcing an operation to compute a hash of proportional difficulty, or whatever). But this is a different problem.

A free market solution doesn't really work when it comes to exchanges, as they are natural monopolies. You've had some attempts from newcomers to break into the space, but with extremely limited success.

Actually prior to governmental action the markets were quite fragmented. In the 70s in response to securities law they started consolidating. Then in response to further law they started fragmenting again in the 2000s.

I’m not largely a free market is the only solution person but this is a very bad example of ‘natural’ monopolies.

Liquidity begets liquidty, so it makes economic sense for exchanges to consolidate in a completely free market.

It makes sense all things being equal, but that isn’t the case.

Even in the old consolidated system you had differentiation between the exchanges in the form of exchange rules which led to Nasdaq specializing in different listings than NYSE.

If you look at the commodities exchanges they were even more specialized and splintered until legal interventions made those specializations less valuable.

It was governmental intervention that made liquidity the only factor in exchange choice.

Yes, so you can, at least if you explain persuasively that this is better than the status quo. Without considering it carefully, it seems to be tilted towards whover has the lowest latency, except that in this scheme what matters is being able to get your offers in last before the deadline, based on the latest information from other exchanges.

A sibling comment links to the research, if you're interested in knowing more, but there are ways of making latency matter less. E.g. with a blind auction, if you hide the order book as it's being built (so you get no indication of built up demand), you can considerably reduce the impact of low latency

Imagine a market that traded every hour. So everyone lined up their orders and what they were selling and at the top of the hour all the trades were processed. Now imagine this on 9/11, or a fed release. Prices from an hour ago are just really not enough information to understand what the prices should be at now, so it ends up being very confusing (and not to mention bursty!). You could put in an order only to realize your price was too low, and you don't get filled, and have to wait another whole hour! Tons of inefficiencies there.

Those same inefficiencies in the market would be seen at 59 minutes, at 30 minutes, at 10 minutes, but get less and less impactful each time. HFTs basically provide a service that gives you almost instant liquidity and market making, which is valuable to anyone who's looking to act quickly buying or selling.

You are talking about a relatively "new" concept, called Frequent Batch Auction, or Periodic Auction[0].

I remember first hearing about these ideas in the mid-noughties (around 2004..2007), but then again, I was in a technical university with an economics department. The FCA link refers to a paper from 2013, so by that time the topic should have been pretty well researched.

0: https://www.fca.org.uk/publications/research/periodic-auctio...

That is just a property of competition, nothing to do with HFT per se.