Mechanics Behind the “FLASH CRASH”

Presented by Dennis Dick, CFA

 May 6th Flash Crash

Contributing Factors:

 Market Fragmentation
 Lack of Uniform Circuit Breakers
 Lack of Displayed Liquidity
• due to discouragement of displayed liquidity providers
 Dependence on High Frequency Liquidity
 No Affirmative Obligations
 NYSE LRP Circuit Breaker

NYSE Liquidity Replenishment Point (LRP)

 The NYSE has a circuit breaker system called the LRP

 Reason – to curb excessive volatility
 Each stock has it’s own individual LRP
 Typically a few percentage points away from current price
 LRP adjusts as price of stock moves
 Adjusts every few seconds
 Examples

Examples of LRP’s

Ticker Last LRPBid LRPAsk

C $4.05 $3.85 $4.25
 Citigroup, LRP Bid is 20 cents below last, LRP Ask is 20 cents above last.

Ticker Last LRPBid LRPAsk

GS $146.50 $144.50 $148.50

Ticker Last LRPBid LRPAsk

F $13.00 $12.60 $13.40
 Price rises/falls to LRP

LRP circuit breaker is reached:

 Individual stock converts from an automated market to a manual

auction market
 Allows the designated market maker to step in and manually trade
the order flow
 Will manually re-open price of stock at price where supply meets
demand (like a typical NYSE open at 9:30 ET).
 Stock will then resume trading in automated market
 LRPs and May 6th

How did the LRP system affect May 6th:

 Trending down most of the day

 At 2:45 ET, selling pressure increased causing number of LRPs to be
reached
 NYSE went to “slow market” on these stocks
 Unable to access NYSE liquidity during this time
 Smart routers seek out best available liquidity
 Other ECNs are thinner
 Smart routers swept out ECN limit books, in some cases down to as
little as 1 cent
 PG sample of Trades on May 6th

Time & Sales Ticker: PG Date: May 6, 2010

Time Bid Size Ask Size Last Size

14:43:00 61.23 7 61.24 3 61.23 2
14:44:00 60.86 7 60.87 3 60.86 4
14:45:00 59.60 1 59.73 1 59.61 15
14:45:18 59.37 1 59.41 144 59.39 1 Reaching LRP
14:46:00 57.26 15 57.36 2 57.26 7
14:46:30 54.62 1 55.51 12 55.00 3
14:46:35 52.80 2 53.99 3 53.00 10
14:46:40 50.94 1 52.93 6 51.00 1 Trading at bid and ask rapidly
14:46:55 50.00 4 51.00 1 50.00 1
14:47:00 47.63 3 48.96 2 47.63 3
14:47:04 46.01 119 49.01 2 46.01 7 46.01 size bid is taken out
14:47:07 43.51 1 48.71 2 43.90 1
14:47:10 41.58 1 48.99 7 41.89 1 Trading at bid/ask back and forth
14:47:17 39.37 9 49.06 2 39.37 1 Low Print
14:47:21 46.20 6 48.65 6 48.65 2 Bids coming in
14:47:42 52.89 22 52.84 14 56.27 1084 NYSE re-opens
14:48:08 57.95 7 60.00 2 59.04 1
14:50:59 62.00 3 62.19 2 62.09 1 Starting to trade normal again
 ACN sample of Trades on May 6th

Time & Sales Ticker: ACN Date: May 6, 2010

Time Bid Size Ask Size Last Size

14:45:01 40.19 4 40.21 1 40.31 4
14:46:02 39.89 2 39.90 6 39.73 2
14:46:31 39.66 2 39.40 1 39.68 2 crossed market
14:47:41 38.00 99 39.04 2 38.00 4 size at 38 is taken out
14:47:46 32.62 39 36.59 3 34.61 1
14:47:49 27.70 1 33.24 3 32.12 1
14:47:50 24.02 1 33.24 4 24.09 1
14:47:51 1.88 1 33.24 4 17.74 1 lowest trade not busted
14:47:55 0.01 4 33.23 3 27.00 1 note- busted trades not showing
14:48:00 28.34 3 33.24 2 33.24 2
14:48:00 0.01 4 33.49 1 28.34 3 somebody bids, immediately hit
14:48:02 38.88 22 36.69 1 39.00 54 stock re-opens NYSE, crossmkt
14:48:03 39.01 3 39.02 10 39.01 4 stock starts to trade normally
IWF sample of Trades on May 6th
Ticker: IWF (Russell 1000 Growth Index fund) Date: May 6, 2010
Time & Sales
Time Bid Size Ask Size Last Size
14:45:00 48.26 14 48.32 1 48.28 1
14:45:30 47.25 20 47.43 20 47.50 10
14:46:00 41.48 3 44.28 3 42.14 1
14:46:08 35.28 2 42.64 4 36.58 1
14:46:15 27.56 1 34.98 2 27.56 1
14:46:17 18.58 1 32.53 4 21.58 1
14:46:19 0.58 1 20.89 8 3.58 1
14:46:35 0.10 984 18.24 11 0.10 23
14:46:38 19.96 40 18.16 2 19.97 16 crossed market
14:46:39 0.10 906 19.87 1 14.65 1
14:47:27 0.01 22 0.03 97 0.10 2
14:47:28 0.0001 10 19.97 4 0.0001 1 trading below a penny
14:49:57 11.20 12 18.24 19 17.28 1
14:54:44 38.00 2 38.39 5 38.19 5
 NYSE LRP to blame?

Who’s at fault?

 Is NYSE at fault for going to a “slow” market

 Some critics say Yes, but I disagree
 PG traded no lower that $56 on NYSE
 NYSE busted zero trades, although some were busted on their ARCA
exchange
 Fault is that other exchanges didn’t have similar volatility control
systems in place
 Uniform Circuit Breakers

SEC Solution: Uniform Circuit Breakers

 Pilot program began in mid-June on uniform circuit breakers for S&P

500 stocks
 Any stock moving more than 10%, in a five minute period, is halted
for 5 minutes on ALL exchanges
 Idea – give the affected security time to attract new trading interest
 Been a few incidents when circuit breakers have been triggered.
- Citigroup (C) – June 29th, trades at 3.3174 for 8800 shares outside
of current market of 3.79 – 3.80 and stock halts for 5 minutes
- Washington Post (WPO) – June 16th
WPO Trades on June 16th

Time & Sales Ticker: WPO Date: June 16, 2010

Time Last Size Time Last Size

15:07:30 452.35 3 15:07:30 453.67 1
15:07:30 452.79 1 15:07:30 453.67 1
15:07:30 452.77 1 15:07:30 455.12 2
15:07:30 452.52 3 15:07:30 455.14 2
15:07:30 452.79 1 15:07:30 456.91 1
15:07:30 453.50 1 15:07:30 457.76 1
15:07:30 452.77 1 15:07:30 457.99 2
15:07:30 453.68 1 15:07:30 458.72 1
15:07:30 453.50 1 15:07:30 457.99 5
15:07:30 453.52 1 15:07:30 456.91 1
15:07:30 453.69 1 15:07:30 462.84 1
15:07:30 453.67 1 15:07:30 919.18 4
15:07:30 454.43 1 15:07:30 919.18 2
15:07:30 454.51 1 15:07:30 462.85 1
15:07:30 454.51 1 15:07:30 459.11 1
15:07:30 455.12 1 15:07:30 929.18 1
 Will uniform circuit breakers stop
future flash crashes?

With the uniform circuit breakers in place, will

future flash crashes be avoided?

 It will help, but in a real impact event may just slow impending crash
 LRP system of NYSE and lack of similar circuit breakers on other
exchanges helps to explain problems with NYSE stocks, BUT
 Does little to explain why Nasdaq listed issues fell
 Eg. AAPL fell 50 points in a 15 min span
 All liquidity was accessible
 Lack of Displayed Liquidity

Another Contributing Factor: Lack of Displayed

Liquidity

 Internalization practices where Tier 1 participants internalize

uninformed flow and “sub-penny” displayed orders
 Increases “toxicity” of public order flow
 Discourages displayed market making activities
 Pushes market makers to undisplayed venues, leaving us with less
liquidity in “Lit” markets
 Broker-Dealer Internalization

What is Broker-dealer Internalization?

 When a broker-dealer executes directly against it’s customers orders,
or alternatively routes it’s customer’s order to an internalization pool
where other market participants will execute against the order
- done off exchange – reported to a TRF – Trade Reporting
Facility

Reasons:
1. To avoid access fees.
2. To receive payment for order flow, from internalizing participant.
3. To jump the displayed order queue.
 Informed vs Uninformed orders

Informed Orders
 Those orders on the right side of the market in the short-term, with
regards to the bid-ask spread and basic market making mechanics
 Internalizers typically do not trade against informed orders

Uninformed Orders
 Those orders on the wrong side of the market in the short-term, with
regards to the bid-ask spread and basic market making mechanics
 Internalizers typically execute against uninformed orders
 Most common type of uninformed order: the Market order
 Profit by queue jumping

Consider the following example:

Ticker: C

Bid Size Ask Size

4.18 30287 4.19 48298

An internalizer can take the opposite side of their customer’s

market buy order and sell the stock at 4.19, jumping ahead
of the 4.8M shares offered there.
Similarly, take opposite side of marketable sell orders and
buy at 4.18 ahead of queue.
 Sub-Pennying

Sub-pennying to improve 605 stats:

The SEC keeps track of price improvement stats in their rule 605 reports:

Internalizers will offer a few sub-pennies of price improvement to improve

their 605 stats, and give them justification for jumping the queue (price
improvement, and saving access fee).

Eg. Sell C at 4.1899 or buy at 4.1801 in front of displayed NBBO.

 Toxic Order Flow on Exchanges

“Toxicity” of Public Order Flow

With the majority of uniformed orders being internalized, order flow on

exchanges becomes more toxic.

 Discourages displayed market makers

 Pushes displayed MMs to undisplayed centers
 Less displayed liquidity
 Less buyers to absorb selling pressure in market impact event
 HFT Dominance of Public
Exchanges

Additional Problem, HFT dominance of

exchanges:

HFT enjoys specific advantages over other market participants:

1. Co-location – reduces latency

2. Flash Orders – glimpse of incoming orders
3. Queue jumping – using ISO orders, due to SIP slowness
4. Participation in some internalization pools
 Dependence on HFT Liquidity

HFT Dominance, and internalization practices

have pressed out traditional displayed
liquidity providers:
Leaving market with a dependence on HFT Liquidity.
Problems:

NO Affirmative Obligations!!
When going gets tough….they step away.
 Summary

Summary of market structural problems:

1. Lack of Uniform Circuit Breakers
2. Lack of Displayed Liquidity
3. Lack of regulation on B/D Internalization
4. HFT Dominance of Public Exchanges, and lack of competition
5. Lack of Affirmative Obligations for current displayed market makers

All these factors led to “Flash Crash”.

 Possible Solutions

Possible Solutions:

1) Uniform circuit breakers – Pilot is in place.

2) Internalization regulation – Trade At Rule, or minimum amount of
price improvement
3) HFT dominance – Level the Playing Field, re-attract traditional market
makers.
4) Lack of affirmative obligations – need better than “stub” quotes
- more affirmative obligations for HFT and market making
participants.