Side Lesson 26: Market Liquidity — Depth, Spread, and What Really Happens in a Crisis

---

1. Why This Is Important

Liquidity is the single most under-appreciated risk in retail portfolios. Returns get all the headlines; liquidity gets all the obituaries. Every major blow-up in the last forty years — 1987, 1998, 2008, 2010, 2020, 2022 — had a liquidity component bolted onto whatever fundamental trigger started it. Vol-tail-wags-dog is really a liquidity statement: when volatility doubles, the cost of getting out of a position quintuples, and it is that second leg that does the damage.

There are four reasons a retail investor needs a working model of liquidity.

Implicit cost compounds. A 2 bp average round-trip on AAPL is

noise. A 50 bp round-trip on a $300M small-cap, paid four times a year through rebalancing, eats 200 bps off your return — bigger than most factor premia (week50). You cannot decide what to own until you can price the cost of getting in and out.

Crises are liquidity events first, fundamental events second.

In March 2020 the cleanest, deepest market on Earth — the 30-year on-the-run US Treasury — went from a half-basis-point bid-ask to forty basis points in eight trading days, and the Fed had to intervene to put it back together. If that can happen to the benchmark risk-free asset, your small-cap value sleeve is not special.

**Asset-allocation matches liability liquidity, or it punishes

you.** A 401(k) that you cannot draw for thirty years can hold illiquid private credit or non-traded REITs (side14). A taxable account that funds tuition in two years cannot. The four-tranche framework is, at heart, a liquidity tiering exercise.

The Fed's playbook is a liquidity script. PDCF, MMLF, BCFP,

BTFP — every emergency facility named in the last two cycles is a targeted plug into a specific drained pool. Knowing which facility addresses which pool tells you who is solvent and who is not when the next round comes.

This side lesson installs the vocabulary, the proxies, the historical case studies, and a working calculator so that "is this thing liquid?" becomes a question you can answer with numbers, not adjectives.

---

2. What You Need to Know

2.1 The Three Dimensions of Liquidity — Tightness, Depth, Resilience

Academics divide liquidity into three dimensions that are usually correlated but can decouple violently in stress.

Tightness is the bid-ask spread — the round-trip cost of trading a single share at the touch. On AAPL during a normal session you will see $0.01 on a $200 stock — half a basis point. On a $5 micro-cap you might see $0.05 — 100 basis points. Tightness is what you see; it is the most visible dimension and the only one most retail traders track.

Depth is the size resting at the touch. AAPL might show 50,000 shares offered at the inside. A small-cap might show 200. Depth tells you how big you can trade before you start walking the book — eating through successively worse-priced limit orders. The Almgren-Chriss square-root impact model (week44) says cost in bps grows like the square root of trade-size-as-percent-of-ADV: a 1%-of-ADV order is about 10 bps of impact in normal vol, a 10% order is about 32 bps, a 100% order is about 100 bps.

Resilience is how fast the book replenishes after you trade. A deep, resilient market lets a market-maker walk back to within 1 bp of the prior mid in seconds. A fragile market — one where the displayed depth is spoofed by HFTs that vanish when stress hits — gaps and stays gapped. The classic resilience failure was the May 6, 2010 flash crash, when bona-fide market-makers withdrew quotes and SP500 e-minis fell 9% in five minutes against a thin book that took twenty minutes to rebuild.

2.2 Liquidity Proxies — What to Measure When You Can't See the Book

You will rarely have direct access to the limit order book outside of exchange-paid data feeds. Four proxies that approximate the three dimensions, all available in any retail data package:

Quoted spread. Bid-ask divided by mid, in bps. Direct measure

of tightness. Apr 2026 typical print: SPY 0.5 bps, AAPL 1 bp, median S&P 500 name 4 bps, median Russell 2000 name 25 bps, median micro-cap 150 bps.

Average daily volume (ADV). Dollar volume per session. Direct

measure of capacity. Apr 2026: SPY $30B/day, AAPL $14B, median S&P 500 name $400M, median Russell 2000 name $25M, median micro-cap under $500k.

Amihud illiquidity ratio = mean of |daily return| / dollar

volume. Higher = more price reaction per dollar traded = illiquid. Empirically the cleanest single-number cross-sectional liquidity score.

Turnover = ADV / market cap, annualised. Mega-caps run

100-200% turnover; micro-caps often under 20%. Low turnover means a name is available but not flowing — a thin holder base that does not absorb a forced seller without price impact.

For ETFs there is a fifth proxy worth a paragraph on its own: the creation/redemption arbitrage keeps the ETF's premium-to-NAV tight because authorised participants (APs) profit from any meaningful gap (side03). For SPY, premium rarely strays more than 2 bps from NAV during a normal session. But the AP arbitrage is a capacity-limited mechanic — when underlying basket liquidity dries up, AP balance sheets can't absorb creation flow, and the ETF can detach from NAV by 100+ bps (March 2020 LQD: NAV-discount of 5%).

2.3 The Stylised Liquidity Curve — Why Small-Caps Cost 50x More

Across the US-listed universe, bid-ask spreads scale roughly with the inverse square root of market cap, with a sharp non-linear blow-out below $1B.

Bucket	Cap range	Typical spread	Typical ADV	Round-trip cost
Mega	> $500B	1 bp	$5-30B	<5 bps
Large	$10B - $500B	2-5 bps	$200M-$5B	5-10 bps
Mid	$2B - $10B	5-15 bps	$30-200M	15-40 bps
Small	$300M - $2B	25-60 bps	$3-30M	50-150 bps
Micro	< $300M	100-300 bps	< $3M	200-600 bps

Round-trip cost includes spread plus square-root impact for a 1% ADV order. A retail investor who buys AAPL pays ~5 bps to get in and out. A retail investor who tries to build a $50,000 position in a $200M micro-cap pays 400-600 bps in one direction — and may not be able to exit at all in stress.

2.4 ETFs vs Underlying Basket Liquidity — The AP Backstop

The standard argument for owning fixed-income via ETF (LQD, HYG, MUB, TLT) rather than the underlying bonds is that the ETF wrapper trades 1000x more liquidly than the basket. LQD averages $2-3B/day in on-screen volume; the median bond inside it changes hands on TRACE maybe twice a week. The wrapper is the liquidity.

This works in normal regimes because APs arbitrage any wrapper-vs-NAV gap. It can break in stress. March 2020 LQD: ETF traded at NAV-minus- 4.6% on March 12; HYG hit NAV-minus-7.0%. The Fed's announcement of SMCCF/PMCCF (corporate-credit facilities) on March 23 closed those discounts inside 48 hours and they have not reopened. The lesson is not "don't own bond ETFs" — they are the only practical way for a retail investor to hold credit. The lesson is that *during a once-a- decade liquidity event the wrapper can detach by 5%, and you should not need to sell it that day.*

2.5 Liquidity Vanishes When Needed — The Three Canonical Crises

Three case studies, each illustrating how liquidity disappears in exactly the regime when an investor most wants to use it. They are the empirical foundation for the vol-tail-wags-dog observation.

March 2020: Treasury market dysfunction. The on-the-run 30-year US Treasury — the deepest, most liquid asset on the planet — saw bid-ask spreads explode from sub-1 bp to over 40 bps between March 9 and March 17, 2020. Off-the-run spreads were worse. Cause: foreign central banks selling Treasuries to fund USD redemptions; hedge funds unwinding relative-value Treasury basis trades amid margin calls; primary dealer balance sheets full from Treasury issuance and unable to absorb more. The Fed responded with $1.5T in repo operations on March 12, then unlimited QE on March 23, restoring spreads to near-normal by April.

May 2010: Flash crash. On May 6 at 14:42 ET, a single $4.1B e-mini S&P sell algorithm executed without price/time discretion. Liquidity providers withdrew. The S&P fell 9% in five minutes. Hundreds of stocks printed at "stub quotes" of $0.01. Recovery took 20 minutes. Cause: a thin order book at lunch, fragmented routing across 13 exchanges, and HFTs whose quotes were not obligated to provide depth. SEC response: limit-up/limit-down (LULD) bands, single-stock circuit breakers, and the IEX 350μs speed bump (week44).

August 2007: The quant unwind. Long-short equity-market-neutral funds running similar value/momentum factor portfolios suffered simultaneous losses August 6-9, 2007. The mechanism: one or two levered shops were forced to reduce exposure (subprime margin calls elsewhere); their selling moved factor returns; everyone running the same factors saw losses; the strongest funds also had to delever; the selling fed itself for three days. Renaissance lost ~6% in 48 hours. Goldman's Global Equity Opportunities lost 30% over a week. By the end of the second week, factors fully recovered. The lesson: in crowded low-vol strategies, liquidity withdrawal manifests as a correlation spike — every "uncorrelated" trade becomes the same trade for as long as the unwind takes.

2.6 Fed Liquidity Toolkit — The Modern Backstop

Since 2008 the Fed has assembled a menu of liquidity facilities, each sized to a specific failure mode. They map directly onto the case studies above.

• Discount window (always-on): banks borrow against collateral.

Stigmatised, low usage.

• PDCF — Primary Dealer Credit Facility (2008, 2020, 2023): repo

for non-bank primary dealers. Plugs broker-dealer balance sheets.

• MMLF — Money Market Mutual Fund Liquidity Facility (2008,

2020): backstop for MMF redemption runs.

• CPFF — Commercial Paper Funding Facility (2008, 2020):

buys CP from corporate issuers when private buyers vanish.

• BCFP / SMCCF / PMCCF — Corporate Credit Facilities (2020): Fed

buys IG corporate bonds and ETFs. Closed the LQD/HYG NAV discount in 48 hours.

• BTFP — Bank Term Funding Program (2023): one-year loans against

Treasuries at par (not market value), specifically designed to prevent forced sales of underwater long bonds during the SVB run.

The market lesson from 2008/2020/2023: the Fed will eventually restore liquidity to systemic markets, but the announcement effect matters more than the dollar size. Investors who waited for the announcement (March 23 in 2020) caught most of the recovery; investors who tried to time the bottom before the policy response mostly missed it. The market can stay irrational longer than you can stay solvent — but the Fed can keep you solvent if you survive long enough for the facility to arrive.

---

3. Common Misconceptions

"If a stock has a quote, it's liquid." A quote is a one-share

indicative price. Depth at the touch can be $200 even when the spread is $0.01.

"Bid-ask is the only cost." Spread is half the cost of trading;

market impact is the other half. For 1%-of-ADV orders they are roughly equal. For larger orders, impact dominates.

"ETFs are always more liquid than their underlying." True in

normal regimes via the AP arbitrage. False in stress when AP balance sheets are full — see LQD March 2020.

"Trading volume = liquidity." Volume is demonstrated

activity. It does not tell you what the next million shares costs. A name can have $50M average volume and still cost 200 bps to move $5M in five minutes.

"Treasuries are always liquid." They are almost always the

most liquid market on Earth. They were not in March 2020. The Fed had to intervene to make it true again.

"The flash crash was an HFT problem." The proximate cause was

a single sell algorithm without price discretion. HFTs amplified the move by withdrawing liquidity, but they did not cause it. The policy fix (LULD, circuit breakers) targeted the amplification.

"Big-name index funds can absorb anything." SPY can. A $400M

thematic ARK product cannot. Fund AUM and underlying basket liquidity must be separately checked.

"Liquidity premia are arbitrage-free returns." No. They are

compensation for being illiquid in stress. The premium is real but you are paid to not be able to sell when everyone else is.

---

4. Q&A Section

Q1: How do I check the bid-ask spread on a stock I'm considering? A: Any retail platform shows top-of-book. Look at the displayed bid and ask, take the spread, divide by the midpoint, multiply by 10000 to get bps. Do this during regular hours — pre-market and after- hours spreads can be 5-10x wider. For meaningful comparison use the 3 PM ET print, when both market-makers and institutional order flow are most active.

Q2: What's the smallest market cap I should consider holding? A: For taxable accounts with rebalancing pressure, $2B is a reasonable floor — that's the small-cap line. Below $300M (micro-cap) the combined spread + impact + occasional gap risk usually exceeds the expected alpha. The Russell 2000 itself is ~$3B median; most retail factor ETFs (AVUV, VBR) screen out the bottom 10-20% of micro-caps explicitly.

Q3: Is bond ETF liquidity real or fake? A: Both. The wrapper liquidity is real and tradeable in any normal market. But it is backstopped by AP arbitrage on the underlying basket, and the basket itself trades by appointment. In March 2020 LQD detached from NAV by 4.6% and HYG by 7%. Owning bond ETFs is fine — sizing them as if they have stock-grade liquidity in a crisis is not.

Q4: Why does the bid-ask spread widen in volatile markets? A: Market makers price spread as compensation for adverse-selection risk plus inventory risk. Both rise with volatility. Empirically spreads are roughly proportional to realised volatility — VIX 30 implies spreads 2-3x the calm-market level; VIX 60 implies 5-10x. Crisis = spread 10x. This is deterministic, not surprising.

Q5: What is the "liquidity premium" I see in academic papers? A: The empirical compensation for holding illiquid assets, ~1-3%/yr across studies. It is real but it is not free money. You earn it by being unable to sell during the worst quarters — exactly when most investors most want to. Lock-up funds (PE, VC) capture it mechanically. Daily-priced retail products (small-cap value ETFs) capture maybe a third of it because the daily liquidity erodes the illiquidity advantage.

Q6: Can I use the Amihud ratio at home? A: Yes. Take 60 days of (|daily return|, daily dollar volume). Compute mean of |return|/$volume. Multiply by 10^6 to get a usable scale. Lower = more liquid. SPY scores ~0.001; a typical small-cap scores 1-10; a micro-cap can score 100+. It is a one-number proxy that ranks names well across a peer group.

Q7: How do I size a position in an illiquid name? A: Work backwards from the cost of exit. Pick a stress regime — say, 3x normal spread + 30% of ADV in one day (forced sell by Friday). Compute the round-trip cost. If it exceeds your one-year expected return, the name is too illiquid for the position size. If you cannot exit in five days at 2x normal spread, halve the size.

Q8: What was the market lesson from August 2007? A: Crowded trades are the same trade. Five "uncorrelated" hedge funds running similar value/momentum/quality factors all sold the same positions at the same time. Diversification across managers does not help when the managers diversify across the same factors. Quant shops now run capacity dashboards explicitly tracking how much of a factor's ADV they own.

Q9: Should I trust the displayed depth at the touch? A: Partly. About 60-70% of displayed shares actually fill at the quoted price; the rest is fading liquidity — quotes that vanish faster than your order can hit them. In stress that ratio can fall below 30%. Treat displayed depth as an upper bound, not a contract.

Q10: When the Fed announces a new facility, should I buy? A: Empirically, yes — but only the systemic targets. After March 23, 2020 (announcement of unlimited QE + corporate credit facility), LQD returned 14% by year-end and SP500 returned 67% from the low. The trade was: announcement = bottom on the liquidity-stressed assets. But the Fed only backstops things it considers systemic. SVB (2023) backstopped Treasury collateral but left equity holders to zero. Read the press release.