treasury by pengxiang

VIEWS: 18 PAGES: 40

									             Fundamental Concepts
     Liquidity: is a summary quality or attribute of a security or asset market.


     A dynamic measure of the many facets of liquidity (depth, breadth, resilience,
     etc.) is the market impact function. Fleming and Mizrach (2007) is the first
     paper to examine market impact in one of the new Treasury ECNs.




February 2, 2007                                                                      1
                        Market Impact
     Liquidity is used in many literatures, without any consistent definition.


     We want a measure that measures both the quantity available for sale at a
     given moment, and the dynamic effect of order flow on prices.


     The market impact function is the impulse response of prices to a trade shock
     initiated on one-side of the market xt

                                    p t 
                                     b s /b
                                            xt



     It will be related to quoted depth, hidden depth, the bid-ask spread, and
     number of market participants (among other things).




February 2, 2007                                                                     2
                   Policy Questions 1/2
     (1) The Federal Reserve and the Treasury obviously have a concern about the
     liquidity of the government bond market.




February 2, 2007                                                                   3
                     Policy Questions
     (2) Measuring execution quality in the fixed income market.
     The SEC approved several amendments to NASD Rule 2320, requiring broker-
     dealers to use “reasonable diligence to ascertain the best market for the
     particular security and to buy or sell in that market so that the price to the
     customer is as favorable as possible under the prevailing market conditions”
     for all securities, including bonds, and for all trading venues.

     The European Union wants to extend its Markets in Financial Instruments
     Directive (MiFID) to bonds.

     MiFID goes a step further than Reg NMS in requiring that systematic
     internalizers--firms that cross trades on their own books--publish firm bid-and-
     offer prices in a "public fashion" during trading hours, much as stock exchanges
     do, in a nondiscriminatory manner and at a reasonable commercial cost.”
     (Securities Industry News January 22, 2007)


     IMHO: the Fed system should enter into this discussion.

February 2, 2007                                                                        4
   Stages of the Treasury Market
     Stage               Factoid             How Traded           Database
     When Issued         Before auction      ECN, Voice           eSpeed, GovPX
     New Issues          Calendar            Auction              Treasury Department
     On The Run          Commoditized        ECNs                 eSpeed, BrokerTec

     Off the Run         Illiquid            Voice                GovPX



     Futures             Liquid              Trading pits CBOT,   Cisco futures
                                             CME

     The when-issued market consists of trading in securities that are about to be
     auctioned.
     The on-the-run Treasury is the most recently auctioned security of a particular
     maturity.
     Previous issues are considered off-the-run. Off-the-run Treasuries are more
     illiquid, have larger spreads, and seem arbitrageable.

February 2, 2007                                                                        5
               Electronic Communication
                       Networks
     Electronic communication networks (ECNs) are alternative trading systems
     (ATS) that match buying and selling interest automatically without dealer
     intermediation.


     They have become the predominant trading mechanism in both equities and
     foreign exchange.

                                                                               70                                               70


                                                                               60                                               60




                                                                                                   Percentage of Transactions
                                                                               50                                               50
                                                                                    Market Share
                                                                               40                                               40


                                                                               30                                               30


                                                                               20                                               20


                                                                               10                                               10


                                                                               0                                                 0
       1996   1997   1998   1999   2000     2001   2002   2003   2004   2005                                                         1989   1992   1995          1998   2001   2004   *
                                                                                                                                                          Year
                                       Year



      Nasdaq ECN share                                                                                           FX market share

February 2, 2007                                                                                                                                                                          6
    On-The Run Treasury Market
      Mizrach/Neely (2006): On-the-run volume nearly 100% electronic, split
      between ICAP BrokerTec and Cantor‟s eSpeed, two ECNs.




                       39%

                                                          BrokerTec
                                                          eSpeed

                                                 61%




      Momentum is with BrokerTec. Cantor had 70% share in 2001.


February 2, 2007                                                              7
     Trends in Trades and Quotes
     Huge increase in trading activity:

                        Trades                          Volume $mn
            GovPX    BrokerTec      %ch       GovPX     BrokerTec %ch
      02Y      388         3,656      841.25%   5,129       30,497 494.58%
      05Y      361         8,552     2271.60%   2,086       25,435 1119.27%
      10Y      134        10,335    7609.46%      984        27,143 2657.91%
      30Y       62         1,706     2645.77%     179         2,836 1483.12%
    Dramatic decline in spreads
                                 Spreads (256ths)
                           GovPX BrokerTec %ch
                     02Y     6.6752     2.0305 -69.58%
                     05Y     9.2576     2.0874 -77.45%
                     10Y    16.7888      3.9778 -76.31%
                     30Y   43.8720      8.1286 -81.47%

     Source: Mizrach and Neely (2006) and Fleming and Mizrach (2006).

February 2, 2007                                                               8
                          BrokerTec 2002-6
     Trades – Daily averages
                          2Y            3Y            5Y            10Y        30Y
                   2002
                   2003
                                1,142
                                1,252           435
                                                            1,590
                                                            2,408
                                                                       1,397
                                                                       2,040
                                                                                   165
                                                                                    99
                                                                                         eSpeed still holds a
                   2004         1,924           895         3,924      3,937       458   significant share in
                   2005
                   2006
                                2,965
                                3,656
                                              1,839
                                              2,704
                                                            6,669
                                                            8,552
                                                                       7,308
                                                                      10,335
                                                                                 1,262
                                                                                 1,706   the 30-year
     Volume -$mn

                          2Y            3Y            5Y            10Y        30Y
                   2002        12,465                       7,536      5,259       432
                   2003        11,665         2,827         9,923      6,691       264
                   2004        18,286         5,032        16,240     13,203     1,069
                   2005        27,194         7,914        23,303     21,876     2,687
                   2006        30.497        12,224        25,435     27,143     2,836




February 2, 2007                                                                                                9
                BrokerTec Order Book
                                         Bid       Ask
     Maturity      Quotes    Spread     Depth     Depth     #Bids     #Asks     Avg. Bid   Avg. Ask
     2-Year         11,897     2.0305    205.37    199.53     20.99     20.75      10.16      10.02
     3-year         10,944     2.3079     55.20     55.08     12.65     12.80       4.56        4.52
     5-Year         27,959     2.0874     43.61     43.91     11.33     11.38       4.00        4.01
     10-Year        31,461     3.9778     44.44     43.91     13.58     13.48       3.35        3.34
     30-Year        13,706     8.1286      5.57      5.52      3.43      3.46       1.67        1.64




     More than 50% of BrokerTec quotes are from computerized trades (Safarik,
     2005) rather than from the primary dealers.




February 2, 2007                                                                                       10
                   Modeling Bid and Ask
         We follow Escribano and Pascual‟s (2006) generalization of
         Hasbrouck's structural model. The model allows bid and ask prices to
,        follow separate stochastic processes, but imposes a vector error
         correction mechanism through the spread,

                                     st  pta  ptb


         Signed order flow xt drives the model. Trade initiation is included in the data
         set, and we use the convention that a buy (sell) trade is signed +1 (-1).

         We interact the order flow indicator with volume to identify buy and sell shocks
         independently
                                     xtb  I ( xt  0)Vt

                                     xta   I ( xt  0)Vt

         Remark: No ambiguity about trade initiation in this market.

    February 2, 2007                                                                        11
                                             VECM
         A vector error correction representation is our baseline model.
,
         The change in bid and ask prices, and signed order flow  trade size on
         the bid and ask


                       ptb            ptb1           ptb1  u pb,t 
                        a               a                a               
                       pt   s   pt 1   ...   pt 1   u pa ,t ,
                        xb      t 1 1
                                          xb            p
                                                             xb   u xb,t 
                        a 
                          t
                                          a 
                                             t 1
                                                             a1  
                                                                t
                                                                                
                        xt 
                                        x t 1 
                                                           x t 1  u xa,t 
                                                                    




    February 2, 2007                                                                 12
    The Dynamics of Market Impact
         The current state of the book is not the measure of the market impact. We need
         to know (at least) four other things
,
         (1) How often are buy (sell) orders followed by other buy (sell)
         orders.

         (2) How quickly do new liquidity providers enter to fill the
         now larger spread.

         (3) When does the macroeconomic environment make the order book
         less resilient.

         (4) Is there hidden depth




    February 2, 2007                                                                      13
              Market Impact Dynamics
         Baseline model:
                                          0.01


,                                         0.01



                                          0.00                                                        Buy Shock
                  Price Impact in 256th




                                          0.00
                                                      Bid _______________
                                                      Ask - - - - - - - - -
                                          0.00
                                                  0         10                20       30        40               50   60

                                          0.00
                                                                                                  Sell Shock

                                          0.00



                                          -0.01



                                          -0.01
                                                                                   Trade Ticks


    February 2, 2007                                                                                                        14
                               Asymmetries
         There are four (minor) asymmetries in the baseline market impact estimates.

         (1) A buy shock has a larger impact on the ask price than the bid
,
                                     pta s /  xb,t  ptb s /  xb,t

         (2) A sell shock reduces the bid price more than the ask price
                                     ptb s /  xa,t  pta s /  xa,t

         (3) Sell shocks have a larger absolute market impact on the bid

                                     ptb s /  xa,t  ptb s /  xb,t ,

         (4) And the ask
                                       pta s /  xa,t  pta s /  xb,t


         The last effect is of concern to policy makers because it shows that bear
         markets are thinner on average.


    February 2, 2007                                                                   15
       Market Impact Magnitudes
                                        Buy Shock                      Sell Shock
                                  Bid        Ask                Bid          Ask
                   Baseline           0.0150     0.0153             -0.0157     -0.0155


     The units are in 256ths of par. All four asymmetries are not particularly
     significant quantitatively.

                              Buy Shock                           Sell Shock
                    Bid            Ask                    Bid           Ask
                                67         65                       64             65

     Amount of buying or selling in millions of dollars required to move the price
     1/256th of par (1 tick).


     These indicate the market is quite deep and can, in normal times, handle
     sizable trade volumes.




February 2, 2007                                                                          16
        Incorporating the Order Book
         Model (a) is the baseline 4-variable VAR

,        Model (b) adds the number of inside market participants     ntb , nta


         Model (c) adds the inside bid and ask depths.               qtb , qta


         Model (d) adds both participants and depth.


         Model (e) in the next section will add hidden depth


         Not (yet) included: liquidity away from the inside quote.




    February 2, 2007                                                             17
                 Buy Side Market Impact
    Excluding the                               0.018

    order book
                                                0.016
    information
,
    overstates the                              0.014
    market
    impact.                                     0.012
                        Price Impact in 256th




                                                0.010


                                                0.008


                                                0.006


                                                0.004                                    Model (d)
                                                                                         Model (c)
                                                                                         Model (b)
                                                0.002                                    Model (a)


                                                0.000
                                                        0   10   20       30        40     50        60
                                                                      Trade Ticks
          Model (b) and (d) are essentially identical, so we work with the more
          parsimonious model (b)
     February 2, 2007                                                                                     18
       Joint Buy and Participant Shocks
         The number of inside market participants impacts the price more than bid and
         ask depths. The overall impact is nearly five times that of model (b)
                                       0.030

,

                                       0.025                                                              ptb s /( xb,t   nb,t )


                                                                            Model (b)
                                       0.020
                                                                            Model (a)
               Price Impact in 256th




                                                                            Model (b) Joint Shocks


                                       0.015




                                       0.010




                                       0.005




                                       0.000
                                               0   10   20       30        40             50         60
                                                             Trade Ticks
    February 2, 2007                                                                                                                      19
                 FOMC Announcements

,                               Date                    Time
                                     February 2, 2005          14:17
                                       March 22, 2005          14:17
                                         May 3, 2005           14:16
                                        June 30, 2005          14:15
                                       August 9, 2005          14:17
                                   September 20, 2005          14:17
                                    November 1, 2005           14:18
                                   December 13, 2005           14:13
                                     January 31, 2006          14:14



         There are 9 FOMC announcements in the sample. The time stamps are from
         Fleming and Piazzessi (2006).


         We will examine market impact functions for the 2-year note



    February 2, 2007                                                              20
                   FOMC Market Impact
        Evidence of overshooting
                                               0.016


,                                              0.014




                                                                               }
                                               0.012
                       Price Impact in 256th




                                               0.010
                                                                                      Market impact 2x larger
                                               0.008


                                               0.006


                                               0.004

                                                                                   Model (b) - FOMC days
                                               0.002                               Model (b)


                                               0.000
                                                       0   10   20       30         40              50     60
                                                                     Trade Ticks




    February 2, 2007                                                                                            21
                                               FOMC Joint Shocks
          On FOMC days, one new bidder and $18 million moves the bid 1-tick.
                                               0.070


,                                              0.060



                                               0.050
                       Price Impact in 256th




                                               0.040
                                                                                   Model (b) FOMC
                                                                                   Model (b) no FOMC
                                               0.030                               Model (b) Joint Shocks+FOMC




                                               0.020



                                               0.010



                                               0.000
                                                       0   10   20       30             40             50        60
                                                                     Trade Ticks



    February 2, 2007                                                                                                  22
                      Quick Summary
     In our preferred 6-variable model, $159 million dollars in 2-year notes moves
     the market one tick (1/256th of 1%).

     The only notable asymmetry is that the market impact of sells is larger in
     absolute value than buys.

     On days of FOMC announcements, the market thins considerably, and only
     $80 million is need for a 1-tick change.


     On FOMC announcement days, one new bidder and $18 million has a 1-tick
     impact.


     Remark: These impacts are linear and easily extrapolated.




February 2, 2007                                                                     23
      Hidden Orders - Definitions
     An iceberg order is a limit order where a portion of the volume is kept hidden
     from the other traders. The visible portion of the iceberg order, called the „peak‟,
     enjoys full price and time priorities as any visible limit order. The hidden portion
     receives only price priority.


     All disclosed volumes are executed first, even if those volumes entered the book
     after the iceberg order submission.


     When a market order hits the hidden portion, a new portion of the iceberg
     order (equal to the peak size) is revealed to the market participants and is
     granted time priority over subsequent order submissions.




February 2, 2007                                                                            24
            Hidden Orders - Theory
     H1: Hidden orders are largest when adverse selection costs are higher.



     H2: Hidden order depth is greatest at the inside.


     H3: Volatility will increase the proportion of hidden orders.



     H4: Hidden depth should reduce market impact.




February 2, 2007                                                              25
              Adverse Selection -1/2
                                     Bid side: hidden depth as proportion of total, hourly means
                                                      data=(hidden depth)/(total depth)
                                                            contract: 2-year, 2005


              0.14



              0.12



              0.10



              0.08



              0.06



              0.04



              0.02



              0.00
                     0   1   2   3    4   5   6   7     8    9   10   11   12   13   14   15   16   17   18   19   20   21   22   23
                                                                       hour




     H1: Hidden order proportions are higher outside of New York trading hours.


February 2, 2007                                                                                                                       26
               Adverse Selection -2/2
                       Average daily means of the proportion of hidden to total depth
                                               contract: 2-year, 2005


             0.2


            0.18


            0.16


            0.14


            0.12


             0.1


            0.08


            0.06


            0.04


            0.02


              0

               Jan   Feb   Mar   Apr   May         Jun         Jul     Aug      Sep   Oct   Nov   Dec

                                             ask         bid    FOMC annoucements




     H1: Hidden order proportions are higher near FOMC announcements.


February 2, 2007                                                                                        27
         Tests of Adverse Selection

                                no FOMC 07:00-15:0014:00-14:30
                      Mean          0.0309  0.0354     0.0122
                      Var           0.0160  0.0177     0.0065
                      t-ratio                 -5.40     18.26
                      Obs          702,195  27,428      6,304




     H1: Partial confirmation. On the overall FOMC announcement days, hidden
     order proportions are higher, but in the critical half-hour window around
     announcements they fall by 2/3.




February 2, 2007                                                                 28
  Where Are the Hidden Orders?
  Bid      Mean      SD        Min       Max           Ask      Mean      SD        Min       Max
  Tier 1       9.647    48.085       0         3,000   Tier 1       8.095    47.543       0         2,000
  Tier 2       7.783    38.223       0         2,800   Tier 2       5.976    42.561       0         1,000
  Tier 3       4.708    28.468       0           900   Tier 3       3.493    31.687       0           903
  Tier 4       3.022    21.992       0           550   Tier 4       2.049    20.275       0           903
  Tier 5       2.183    18.300       0           450   Tier 5       1.547    17.016       0           890


     H2: Confirmed. Hidden depth is highest at the inside.




February 2, 2007                                                                                            29
                               Volatility
    Mat. Constant % Visible(t-1) # Participants Abs. Return
    02Y     0.0065        0.1386          -0.0336  14,469.7513
             (6.75)       (26.13)         -(12.67)     (34.94)



     H3: Confirmed. An 1% increase in the 50-tick moving average of volatility
     increases the proportion of hidden orders by nearly 0.15%.




February 2, 2007                                                                 30
                       Hidden Depth VECM
              We add hidden inside bid and ask depth hb, ha to our VAR.

,
                 p b                   p b 1            p b p
                                                             t        up b
                    t                      t                             t

                 p a                   p a 1            p a p
                                                             t        up a
                    t                      t                             t

                  nb
                   t                    nb 1
                                         t
                                                          nb p
                                                           t          un b
                                                                          t

                  na
                   t                    na 1
                                         t
                                                          na p
                                                           t          un a
                                                                          t
                                  
                          s t  1
                                1                  
                                                  p              
                  xb
                   t     
                         8 1    
                                 8 8   xb 1
                                         t        
                                                   8 8   xb p
                                                           t          ux b
                                                                          t

                  xa
                   t                    xa 1
                                         t               xa p
                                                           t          ux a
                                                                          t

                  hb
                   t                    hb 1
                                         t               hb p
                                                           t
                                                                       uh b
                                                                          t

                  ha
                   t                    ha 1
                                         t               ha p         uh a
                                                                          t
                                                           t




    February 2, 2007                                                          31
 Hidden Orders and Market Impact
                                           0.008


                                           0.007


                                           0.006
                   Price Impact in 256th




                                           0.005


                                           0.004


                                           0.003
                                                                                    Model (b)
                                                                                    Model (e) Hidden
                                           0.002


                                           0.001


                                           0.000
                                                   0   10   20       30        40           50         60
                                                                 Trade Ticks

      H4: There is a very small effect of hidden orders on market impact on a
      typical day. 154.877 v. 154.319 million to move the bid by 1-tick.
February 2, 2007                                                                                            32
                    Hidden Depth Shocks
     Consider then FOMC announcementwith double accompanied by impact
     The shock on the case asymmetric buy shock the impact on the
     Thisshock is also highlyof a $1 milliondays has the largest market a ask,
     of million a much wider spread. depth.
     $1 any joint shock process. It is larger than the display of a new bidder.
     leading to increase in hidden bid
                                       0.060
                                                                      p t t  t t
                                                                       b s /b  p b s /b
                                                                              x            h

                                       0.050




                                       0.040
               Price Impact in 256th




                                                                  Model (e) Hidden
                                                                  Model (e) w/ New Hidden Depth
                                                                  Model (e) Bid New Hidden+FOMC
                                       0.030                      Model (e) Ask New Hidden+FOMC




                                       0.020




                                       0.010




                                       0.000
                                               0   10   20       30            40            50   60
                                                             Trade Ticks
February 2, 2007                                                                                       33
      Conclusions and Extensions
     Market Microstructure Concepts

     Multi-market Structural Model

     Estimation of the Cointegrated Model

     Description of Treasury Market Data and Architecture

     Spot and Futures Market Price Discovery Estimates

     Transition to ECNs

     The Market Impact Function

     Hidden Orders

     Conclusion and Extensions

February 2, 2007                                            34
                            Conclusions
     The futures market plays an important role in price discovery, but its
     contribution does not change appreciably upon release of public information.


     Information shares are functions of observable liquidity, especially the bid-ask
     spread. We can predict information shares out-of-sample.


     ECNs have come to dominate on-the-run trading. ICAP PLC, which owns
     Brokertec, has become the world‟s largest interdealer broker.


     Market impact is a useful way for understanding liquidity. It is a dynamic
     concept and incorporates all the useful information in the order book.




February 2, 2007                                                                        35
                      Conclusions 2/2
     On FOMC announcement days, one new bidder and $18 million has a 1-tick
     impact. These impacts are almost 9x the baseline.


     Hidden orders are an important feature of new electronic markets. They are
     more prominent (a) when adverse selection risk is high (after hours, macro
     announcement days; (b) at the inside market; (c) when volatility is higher.


     A new result in the literature is the magnitude and asymmetry of the hidden
     depth shocks. They matter as much as displayed liquidity.




February 2, 2007                                                                   36
                             Extensions
     Extend information share regression to include level, slope and curvature of
     the yield curve. Link to affine term structure literature.

     Incorporate variables that explain information share into the structural model.
     Korenok, Mizrach and Radchenko (2006) estimates some new models using
     MCMC.


     Quantify potential losses to liquidity traders of market fragmentation and
     slower adjustment.

     Nonlinear market impact functions.




February 2, 2007                                                                       37
             Supplemental
               Material


February 2, 2007            38
                   Microstructure




February 2, 2007                    39
                   Market Architecture
     Securities also trade in a hybrid environment of market designs or
     architectures.

     NYSE: Floor based auction organized by a specialist


     Nasdaq: Interdealer electronic network


     ECNs (ATS): Electronic networks with no dealer intermediaries (e.g.
     Archipelago, Instinet)

     Open outcry: CME, CBOT futures pits, Treasury phone based market


     Mizrach and Neely (2006) compares two open outcry markets: the GovPX voice
     market and the CBOT futures pits.




February 2, 2007                                                                  40

								
To top