Docstoc

Industry Momentum and Behaviroal Limits to Arbitrage

Document Sample
Industry Momentum and Behaviroal Limits to Arbitrage Powered By Docstoc
					Informed Trading in Limit Order Markets:
    Evidence on Trinary Order Choice



 Norges Bank/BI Conference on Microstructure,
                 Oslo 2005

               Lukas Menkhoff
               Maik Schmeling
                         Motivation

• Structural change: electronic limit order markets have
  become increasingly successful in the last ten to fifteen
  years.
→ Changing market structure may determine changing
  market outcome. However, little is known about order
  choice of differently informed traders in these markets.

• The traditional paradigm about information and order
  choice is:
   – Informed traders use market orders.
   – Uninformed traders use limit orders.
→ This paradigm has been challenged.

                                                              Slide 2
   Challenges to the traditional paradigm
             Trader    Order      Analysis        Economic       Market
              type     type                        issue        conditions

Traditional informed   market    price impact     information       n.a.
                                                  aggregation


• Traders use limit orders for informed trading
  (Kaniel and Liu, JoB 2005).

• All traders use market and limit orders
 (Bloomfield, O‘Hara and Saar, JFE 2005).

•There are different kinds of limit orders (Hasbrouck and Saar, 2004).

→ New roles for different order types?

                                                                             Slide 3
     Findings according to economic issues

1.    Information aggregation (price impact analyses, Section 4)
      1.1 Only orders from financial centers (=informed traders) have price
          impact
      1.2 Market and some limit orders (=screen orders) have price impact
      1.3 The degree of price impact found varies with market conditions


2.    Liquidity provision (analyzing orderbook changes, Section 5)
      2.1 Informed traders provide a significant share of liquidity which
          follows an intuitive time pattern
      2.2 Liquidity is mainly provided via ordinary limit orders (almost 90%)
      2.3 Informed traders increase liquidity provision with higher volatility



                                                                                 Slide 4
                          Data

• Data set for 15 trading days from the Russian UTS,
  March 2002.
• The UTS channels all other RUR/USD interbank spot
  trading in Russia, which takes place on 8 regional
  exchanges.
• The market we analyze is small in volume but serves to
  fix Russia‘s official exchange rate.
• The data mirror the complete limit order book.
• We can recover every single event and have information
  on the geographical location of traders.


                                                           Slide 5
          Trading technology: SELT

• Trading operates on SELT one hour each day.

• SELT is an electronic trading system similar to
  the systems of Reuters or EBS.

• Entries into SELT are
  – limit orders
  – cancellations



                                                    Slide 6
                             Spot FX rate and
                         central bank interventions
                 31.16
                 31.12
Rouble per USD




                 31.08
                 31.04
                 31.00
                                              new sample
                 30.96
                 30.92
                         1 4 5 6 7 11 12 13 14 15 18 19 20 21 22
                                      Day of March 2002

                                                                   Slide 7
Information about the Russian FX market

  Total trading volume (9 days)     697 mn. USD

  Trading volume per day (1 hour)    77 mn. USD
  Average volume
                                     49,395 USD
  per market order
  Average volume
                                    102,048 USD
  per limit order
  Number of trades (9 days)          14,109

  Number of limit orders (9 days)    15,959



                                                  Slide 8
     Informed and uninformed traders

• Traders from the financial, economic and
  political centers Moscow and St. Petersburg are
  regarded as "informed traders“.

• Traders from other regions are regarded as
  "uninformed traders“.

• Traders from financial centers should have
  superior information on average.


                                                    Slide 9
               Moscow and St. Petersburg:
                   financial centers
                 FCAV/      FCAV/               Equity       Profits/   Equity
Region           # Banks     IP      Profits     Inv.        # Banks    Inv. / IP

Moscow             118.50   45.30%    24,937    358,022         41.91   230.01%

St. Petersburg    106.54     5.09%       531      9,783         13.28    11.68%

Ekaterinburg        10.53    0.26%         95     1,415          3.40     1.26%

Rostov               2.22    0.13%         25           57       1.06     0.13%

Samara              25.27    0.58%        119       719          0.54     0.75%

N. Novgorod         11.63    0.35%         73       423          3.69     0.64%

Novosibirsk          5.74    0.29%         39       251          3.02     0.97%

Vladivostock        14.80    0.25%         37           74       6.30     0.21%




                                                                                    Slide 10
                      Some basic statistics
                                                    St.
                     Informed Uninformed Moscow Petersburg
 Market
                            0.14             0.03      0.16     0.08
 order vol.(*)
 Market
                            2.52             0.88      2.64     2.21
 order #(**)
 Limit
                            0.32             0.07      0.36     0.20
 order vol. (*)
 Limit
                            2.33             1.51      2.32     2.36
 order #(**)
 Profits(*)             0.0043             -0.0076   0.0050   0.0022
(*)    per day and trader in million USD
(**)   per day and trader
                                                                  Slide 11
                 Different order types?

We differentiate between

 market orders
                             execute immediately
 (marketable limit orders)

                             aggressively priced limit orders
 screen orders
                             appear on every dealers’ trading screen

                             line up in the book
 ordinary limit orders
                             invisible to other traders




                                                                       Slide 12
             Why screen orders?

• Hasbrouck and Saar (2004) find that there may
  be different kinds of limit orders which serve
  different tasks (fleeting orders on Instinet).

• We find that more than 10% of all limit orders
  are screen orders.

• Why are they used?
  – Competition in market making (BOS, JFE 2005)?
  – Informed trading (Kaniel and Liu, JoB 2005)?


                                                    Slide 13
           e.g. 31 Rouble per USD   new bid at 31
Best ask                             market order
                                    new bid at 30.5
                                    screen order

Best bid
           e.g. 30 Rouble per USD
                                    new bid at 29.5
                                    ordinary limit order




                                                           Slide 14
 Performance of limit order types I

                  Unconditional fill rates

              ordinary    screen        both
 all            44.98%     74.87%       51.21%
 informed       46.02%     75.67%       52.53%
 uninformed     42.87%     73.66%       49.68%

 Screen orders are filled to a high degree


                                                 Slide 15
Performance of limit order types II
                                                            Informed traders

                          100%
                          90%
                          80%
probability of survival




                          70%
                          60%
                          50%
                          40%                  executed ordinary
                          30%
                          20%
                                     executed screen
                          10%
                           0%
                                 0      10    20       30   40    50    60     70   80       90   100   110
                                                        time after submission (in seconds)



                             Screen orders are filled very fast

                                                                                                              Slide 16
          Informed trading: Price impacts

• VAR approach to measure price impact of different order types and
  traders.
• Variables:
         •   xi – informed traders' market order flow
         •   si – informed traders' screen order flow
         •   xu – uninformed traders' market order flow
         •   su – uninformed traders' screen order flow
         •   r – midquote returns
•   Order Flow measures:
         • Market orders: Ask +1, Bid -1
         • Screen orders: Bid +1, Ask -1

•   Aggregation into one-minute intervals.
•   Assumption: Order Flow causes returns (no feedback from returns).


                                                                        Slide 17
                            Results I
• Market and screen orders of informed traders have
  similar price impacts.




       Response of midquotes to a one s.d. shock in order flow.


                                                                  Slide 18
                                       Results II

• Uninformed traders‘ signed flows follow informed
  traders‘ signed flows over short horizons.
               market order flow                           screen order flow
2.0                                          1.2

1.5                                          1.0

1.0                                          0.8

0.5                                          0.6

0.0                                          0.4

-0.5                                         0.2

-1.0                                         0.0
       1   2          3            4     5         1   2          3            4   5

                    minute                                     minute




                                                                                       Slide 19
                                       Results III
               0.0080

               0.0070

               0.0060

               0.0050
Price Impact




               0.0040


               0.0030

               0.0020

               0.0010

               0.0000
                        all   TV low    TV high   BV low   BV high   Spread low Spread high


                    Price impacts under different market conditions
                                                                                              Slide 20
   Informed trading: price impacts again

• Danielsson and Love (2004) find evidence for feedback
  trading in a sample of USD/EUR trades.

• Structural VAR to account for feedback trading in all four flow
  variables.


• Variables: (xi si xu su r)
   – 4 price impact coefficients
   – 4 feedback coefficients
   – 5 structural shocks

   → SVAR is overidentified (2 DF)

                                                                Slide 21
            Results for informed traders




         Response of midquotes to a one s.d. shock in order flow.

• Informed traders‘ flows still have significant long-run price impact.
• We find evidence for feedback from returns to both order flow measures.


                                                                            Slide 22
   Results for uninformed traders

• Uninformed traders' market order flow does not
  significantly impact prices.

• Uninformed traders' screen order flow is negatively
  related to future returns (but the long run response
  is insignificant).

• We find no evidence for feedback from returns to
  both order flow measures for this group of traders.




                                                         Slide 23
         Order flow, feedback trading
               and public news
• Evans and Lyons (2003) find evidence for common
  news shocks that drive both returns and order flow.

• We use the „identification through heteroscedasticity“
  approach of Rigobon (REStat, 2003) to analyze the
  influence of common news shocks in our data set.

• We concentrate on informed traders only since
  uninformed traders' flows do not impact prices.




                                                           Slide 24
                           Empirical setup
           1         α       α        r         1        e 
                                                                     r


                          1        2
                                         t
                                                                t


           α                           x           γ z   e 
                                               i                     x
                 3
                       1        0               t           1   t    t

           α                           s         γ        e 
                                                            
                                            i                        s
                 4
                       0        1           t               2        t




                       A                   yt                      et
xi – informed market order flow, si - informed screen order flow,        r –returns
zt – common news shock,          et – structural shocks

The matrix A contains price impact and return feedback coefficients.
The coefficients in Γ measure the contemporaneous effect of public news on
midquote returns and flow variables.
Identification is achieved through the unconditional heteroskedasticity of the news
shocks zt.and et (3 regimes of return volatility).
A and Γ are fixed over the three volatility regimes.
Results are very similar for different choices of regimes.
                                                                                      Slide 25
    Estimation results: price impacts and
              feedback trading
1 – Price impact of informed market order flow
2 – Price impact of informed screen order flow
3 – Feedback from returns to informed market order flow
4 – Feedback from returns to informed screen order flow
1 – Impact of common news shocks on informed market order flow
2 – Impact of common news shocks on informed screen order flow

                              p-value                      p-value

         1         0.108     (0.000)      1     0.801    (0.000)
         2         0.120     (0.002)      2     0.259    (0.001)
         3         0.221     (0.000)
         4         0.106     (0.000)

                                                                     Slide 26
    General results: liquidity provision
• Informed traders
   – Consume liquidity at the beginning of a trading session.
   – Provide more liquidity towards the end of the trading session.
   – Increase the submission of screen and ordinary limit orders in
     times of rising volatility.

• Uninformed traders
   – Provide liquidity at the beginning of the trading session.
   – Consume more liquidity towards the end of the trading session.
   – Rather cut back on liquidity provision in times of increasing
     volatility.



                                                                      Slide 27
Findings according to economic issue

1.   Information aggregation (price impact analyses, Section 4)
     1.1 Only orders from financial centers (=informed traders) have price
         impact
     1.2 Market and some limit orders (=screen orders) have price impact
     1.3 The degree of price impact found varies with market conditions


2.   Liquidity provision (analyzing orderbook changes, Section 5)
     2.1 Informed traders provide a significant share of liquidity which
         follows an intuitive time pattern
     2.2 Liquidity is mainly provided via ordinary limit orders (almost 90%)
     2.3 Informed traders increase liquidity provision with higher volatility



                                                                                Slide 28

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:8
posted:9/9/2011
language:English
pages:28