There are many liquidity measures in literature, some based on trading volume, others based on bid-ask spread, etc. From a practitioner's point of view, what characteristics do you feel a good market liquidity measure should have? thanks.

A couple of nice ones are depth at NBBO and market impact (e.g., Kyle's Lambda).But you might want to think about measures of the stochasiticity of traditional liquidity measures if the real definition of liquidity is the optionality of transactions. That is one might measure liquidity in terms of whether one can buy and sell as much as one wants any time one wants. Which would you prefer: a market that has constant, modest levels of traditional liquidity (constant modest trading volume or constant modest bid-ask spread) or one in which the market has unknown, time-varying levels of traditional liquidity that happen to have a modest average over time? One might be tempted to say that illiquidity is a measure of the % of time that one cannot readily complete a non-trivial transaction at a fair price rather than the average discrepancy (bid-ask spread) over time. That implies liquidity is a measure of the asymmetric excursions in volume or bid-ask spread (i.e. volatility of volume or volatility of spread) rather than the averages of these values.Of course, historical liquidity or stability of liquidity is no guarantee of future liquidity. One might argue that all markets are potentially instantaneously illiquid for non-epsilon-sized orders until proven otherwise by the successful execution of the order.

One thing to include might be some measure of how long it takes (a non-dealer) to execute (any part of a) limit order that improves the market.The first issue is: are such orders even allowed? They weren't for many years with NASDAQ -- it was aso-called dealer market until 1996-1997. The effect was quite dramatic in the 1987 crash. At least the NYSEwas open for business. The NASDAQ dealers simply stopped answerng the phone. So, if your market isa dealer market, you can probably assume that, under times of great stress (like the financial crisis), it will simply disappear. The US listed options markets is an order-book market. But, for many underlyings, it is *not* liquid, and thistime-to-execute measure would be quite telling. So, the 'characteristic' of this one would be that it can distinguish if liquidity is being provided by (i) just market-makers by rule, (ii) mostly just market-makers by practice, or (iii) a good mix of market-makers and others (as for many equities) Just thinking out loud, taking active US equities as the "gold standard" of liquidity, maybe everything should be measured relative to that.

QuoteOriginally posted by: AlanOne thing to include might be some measure of how long it takes (a non-dealer) to execute (any part of a) limit order that improves the market.The first issue is: are such orders even allowed? They weren't for many years with NASDAQ -- it was aso-called dealer market until 1996-1997. The effect was quite dramatic in the 1987 crash. At least the NYSEwas open for business. The NASDAQ dealers simply stopped answerng the phone. So, if your market isa dealer market, you can probably assume that, under times of great stress (like the financial crisis), it will simply disappear.Interesting! Isn't this pathological condition exactly like what happened in the flash crash when HF traders pulled out?If a market seems to be very liquid until, suddenly, it isn't, then was it ever liquid?

QuoteOriginally posted by: Traden4AlphaQuoteOriginally posted by: AlanOne thing to include might be some measure of how long it takes (a non-dealer) to execute (any part of a) limit order that improves the market.The first issue is: are such orders even allowed? They weren't for many years with NASDAQ -- it was aso-called dealer market until 1996-1997. The effect was quite dramatic in the 1987 crash. At least the NYSEwas open for business. The NASDAQ dealers simply stopped answerng the phone. So, if your market isa dealer market, you can probably assume that, under times of great stress (like the financial crisis), it will simply disappear.1. Interesting! Isn't this pathological condition exactly like what happened in the flash crash when HF traders pulled out?You could be right. I'm not sure. Trackstar may have some thoughts on that.2. If a market seems to be very liquid until, suddenly, it isn't, then was it ever liquid?Well, liquidity is a random variable -- let's say non-negative, where 0 means the market is essentially gone.You could ask: what is the probability it can reach 0, especially because of structural issues like being a dealer market with just a few dealers? If it can reach 0, can it jump to 0 or will there be some warnings?

QuoteOriginally posted by: AlanQuoteOriginally posted by: Traden4AlphaQuoteOriginally posted by: AlanOne thing to include might be some measure of how long it takes (a non-dealer) to execute (any part of a) limit order that improves the market.The first issue is: are such orders even allowed? They weren't for many years with NASDAQ -- it was aso-called dealer market until 1996-1997. The effect was quite dramatic in the 1987 crash. At least the NYSEwas open for business. The NASDAQ dealers simply stopped answerng the phone. So, if your market isa dealer market, you can probably assume that, under times of great stress (like the financial crisis), it will simply disappear.1. Interesting! Isn't this pathological condition exactly like what happened in the flash crash when HF traders pulled out?You could be right. I'm not sure. Trackstar may have some thoughts on that.2. If a market seems to be very liquid until, suddenly, it isn't, then was it ever liquid?Well, liquidity is a random variable -- let's say non-negative, where 0 means the market is essentially gone.You could ask: what is the probability it can reach 0, especially because of structural issues like being a dealer market with just a few dealers? If it can reach 0, can it jump to 0 or will there be some warnings?Exactly.Perhaps the key here is that there are at least two different definitions of liquidity that depend on the user's context:1) If one is running daily operations on a portfolio and wants to estimate liquidity associated with routine variations in AUM, rebalancing, or gradual changes in asset allocation, then "average liquidity" probably does a good job. After all, those operations permit some flexibility on the exact timing of trades.2) But if one is handling risk management, margin calls, unwinding rogue trades, and the like, then the "volatility of liquidity" seems more germane because there's much less flexibility on the exact timing of trades. Moreover, to the extent that liquidity is most likely to drop during the times when the participant most needs it, then "worst case liquidity" seems like the way to think about it.

Volume * transaction cost = liquidity.

QuoteOriginally posted by: tigerbillThere are many liquidity measures in literature, some based on trading volume, others based on bid-ask spread, etc. From a practitioner's point of view, what characteristics do you feel a good market liquidity measure should have? thanks.Without reference to the transaction size and execution horizon, any liquidity measure is partial and incomplete, since it refers to the marginal investor who will transact one share, or one round lot. Implementation shortfall if measured right seems to be the right metric. Moreover, when you split your order in a sequence of several smaller ones, then market resiliency is the dominant factor. It drops below 90% of cost attribution only when resiliency is nearly perfect in the interval between two consecutive orders.

QuoteOriginally posted by: tigerbillThere are many liquidity measures in literature, some based on trading volume, others based on bid-ask spread, etc. From a practitioner's point of view, what characteristics do you feel a good market liquidity measure should have? thanks.If liquidity means the ability to trade, then bid/ask trade sizes are the obvious primary measure, supplemented by orders waiting in or around the price range the interested party has in mind. But , of course, if bid/ask spread is too wide, then this implies a lack of willingness to trade around current anticipated prices. So here is a simple suggestionBuyLiquidity = askSize *(askPrice + bidPrice)/(2*(askPrice - bidPrice)) + nextAskSize *(nextAskPrice + bidPrice)/(2*(nextAskPrice - bidPrice)) + .......SellLiquidity = bidSize *(askPrice + bidPrice)/(2*(askPrice - bidPrice)) + nextBidSize *(askPrice + nextBidPrice)/(2*(askPrice - nextBidPrice)) + .......

QuoteOriginally posted by: FermionBuyLiquidity = askSize *(askPrice + bidPrice)/(2*(askPrice - bidPrice)) + nextAskSize *(nextAskPrice + bidPrice)/(2*(nextAskPrice - bidPrice)) + .......SellLiquidity = bidSize *(askPrice + bidPrice)/(2*(askPrice - bidPrice)) + nextBidSize *(askPrice + nextBidPrice)/(2*(askPrice - nextBidPrice)) + .......I wonder if I could teach my dog to count the number of bids and offers. You think he could get a job on Wall Street?

QuoteOriginally posted by: farmerQuoteOriginally posted by: FermionBuyLiquidity = askSize *(askPrice + bidPrice)/(2*(askPrice - bidPrice)) + nextAskSize *(nextAskPrice + bidPrice)/(2*(nextAskPrice - bidPrice)) + .......SellLiquidity = bidSize *(askPrice + bidPrice)/(2*(askPrice - bidPrice)) + nextBidSize *(askPrice + nextBidPrice)/(2*(askPrice - nextBidPrice)) + .......I wonder if I could teach my dog to count the number of bids and offers. You think he could get a job on Wall Street?I am sure your dog could get one if you could.

I like the point expressed in section 1.1 "Credit default Swaps Liquidity modeling: A survey" Brigo, Capponi Predescu, 2010.Though paper contains a fresh idea it is not look completed technically.