Is Eliminating Tracking Error Hazardous to your Wealth?

Embed Size (px)

Citation preview

  • 8/14/2019 Is Eliminating Tracking Error Hazardous to your Wealth?

    1/3

    ItlexrrslexlrGClrouIDCOMMENIARY 64 JANUARY 2OO1

    IS ELIMINATING TRACKING ERROR

    HAZARDOUS TO YOUR CLIENT'S WEALTH?As indexing grows, index changes are having an increasing impact on short-term pricemovement in the underlying stocks. Many Plexus clients have seen costs skewed in the last yeardue to changes in both the S&P 500 and the RU2000. Some have attempted to strategize tradingto reduce the impact of Effective date activity while others have stuck with prices near the closeto minimize tracking error.Academic papers on the S&P 'game' have used mostly closing prices; only recently have papersused intraday pricing.

    SummaryUsing our 10 minute database, we find that:. Stocks deleted from the indices show little

    price change.. Stocks added to the indexes show large

    jumps between the close and subsequentopen after the change is announced.

    . NASDAQ stocks show the most impact afterboth the announcement and on the Effectivedates - with notable reversion after the fact.

    . Broker intermediated stocks show thegreatest impact and reversion; this has a bigimpact on the S&P400 and RU2000.

    . lt is cheaper to be early, or even to be late,rather than to trade at or close to the closefor NASDAQ and small cap NYSE stocks.

    Late last fall we read a paper by David Cushing andAnanth Madhavan on the behavior of stock returns at theclose, with particular emphasis placed on Market on Close(MOC) orders.* The primary conclusions were:. End of the day volatility has increased significantly.. The large market caps show unusually higher levels of

    volatility - partially due to more NASDAQ names.. A higher percentage of liquidity imbalances has led toincreased close to ooen reversion.

    The conclusions were consistent with observations thatwe had made in the last year. Clients buying stocks on theday they were added to the S&P showed increasing costsover the course of the day. Similarly, clients rebalancingthe Russell 2000 also had increasing costs as tradingclosed in on the Effective date (June 30). The rationale

    provided for trading toward the close was to minimizetracking error: and the risk of negative tracking due topremature trading exceeded the potential gains.

    Not all index funds trade at the close, however. Many willstart to trade days earlier, basing their strategies onhistoric observations consistent with both our findinos andM ad havan/Cushino.

    Price Adjustment onAnnouncement Date

    The NYSE has taken steps to reduce volatility for dailyimbalances by requiring MOC entries by 3:20 and thesubsequent announcement of order imbalances toencourage public arbitrage.'

    Despite changes. when Yahoo entered the S&P, theEffective day mispricing and subsequent reversalsuggested that problems still exist. To test, we used our 10minute orice database io iook at 40 additions and thelinked deletions from the S&P500 during 1999 to see ifthere were consistent return patterns both prior to andfollowing the Effective date. [Yahoo was dropped due toits skew on results.l We then isolated the changes byexchange to see if the NYSE changes were helpingrelative to the NASDAQ listings. Finally, we looked atchanges in the S&P 400 to see if the same patternappears in a lesser followed but smaller cap index.

    Following the Safeway and AOL debacles in 1998, theS&P stated that it would delay announcing additions anddeletions until after the close. Chart 1 shows eoualweighted 10 minute returns relative to the open on theAnnouncement date and the following day for additionsand deletions. Prices show no information leak: deleted

  • 8/14/2019 Is Eliminating Tracking Error Hazardous to your Wealth?

    2/3

    stocks rose 0.5% on the Announcement date. There wasno difference between NASDAQ and NYSE results,

    For additions, all the movement is between the Close onAnnouncement day and the subsequent Open. NYSEstocks rose 5.3% and NASDAQ's rose 7.8%. Butfollowing the announcement, stocks often reverted;versus the open. NYSE shares fell 0.3% and NASDAQ'sslide back 3.0%.

    Price Adjustment on Effective DateThe next analysis focuses on the 2 day window when thestock is actually added or deleted f rom the index (Effectivedate). We look at returns over the day of the change, thenlook at the next day for signs of reversion.

    Chart 2 again shows 10 minute equal weighted returnsrelative to the ooen on the Effective date. The contrast

    between the NYSE and NASDAQstocks is striking. NYSE stocks showlittle movement; there does not seemto be an imbalance of buy ordersforcing prices higher near the close.NASDAQ stocks show a muchdifferent oattern. The first half of theday again appears flat, but as theclose approaches. prices rise.Between 2 and 4 PN4 pric-os t'ose1.1"k, and between 3:30 and 4 PMorices rose 0.5%.

    Contrary to the Announcement datepattern, prices revert between theclose and the next day's open.NYSE stocks fell 0.6% and NASDAQstocks tell0.7%. Only 4 slocks rose,and these were stocks that had fallenduring the last hour of trading on theEffective date. Once the oressurewas off, excess shares held in

    anticipalion of additional buying were quickly sold off.Over the remainder of the day, NYSE stocks recovered0.2"k versus the open, but the NASDAQS lost anadditional 0.3%.

    Deleted stocks do not show as dramatic a pattern.Returns are flat over the actual deletion date with littlesubsequent close-to-open reversion. As with theAnnouncemenl returns, the actual deletion does notappear to create any appreciable imbalance.

    What about other lndices?

    To determine whether the Announcement and Effectivereturn phenomena were isolated events in a few S&P500names, we also looked at 48 additions to the midcapS&P400. A comoarative table is included at the end.

    The same patterns appear:. Minimal price movement prior to the announcement;. A significant overnight price adjustment;. Excessive price change late on the Effective date;. Large overnight reversion once the stock goes into

    the index;. Deletions (there were also 48) show little return change

    on either the Announcement or the Effective dates.

    Prices Show Modest RisePrior to Effective Date

    Prices fell after the announcement but did not stay down.Between the close on the day {ollowing the announcementand the Effective date, NYSE stocks rose 0.5% whileNASDAQ's rose an average oI7.2k. Announcement-to-Effective date can vary from stock to stock, but it wouldseem that buying after the initial news settles in wouldhave been orofitable.

    Deletions from the S&P show little impact. There was

    no discernable Close-to-Open move. The average andmedian returns were flat. The average deletiondeclined 0.5% after the open, bul was skewed by a fewnames. The median price actually rose +0.1"k. Norwas there any additional pressure between theAnnouncement and the Effective dates. Holders ofthese names were not compelled to sell immediately,and there was little to be gained by shorting inanticipation of index selling.

    Price Change on Announcement Date10 minute lntraday views

    r NYSE Addsr NASDAQ Addshanges anounced at

    the Close.

    No indication of pendingannouncements prior to

    the Close. All of the price adiustment is at theOpen - with little follow through. Nosignificant impact on Deletions.

    9

    I7

    6

    o5(E1+

    83o-t

    1

    0

    -1

    -2

  • 8/14/2019 Is Eliminating Tracking Error Hazardous to your Wealth?

    3/3

    nange on Efiectt'10 minute Slices

    DeletionsO NYSE Addsr NASDAQ Adds

    Demand imbalance onFffcclirra dav Clnca ic eaan in

    INASDAQ stocks - but not in

    NYSE stocksClose to Open reversion is signiicantfor both NASDAQ and NYSE stocks.

    Again, no impact for Deletions.

    f,r,

    ,/,!I

    I'l *-

    rJlr-_ - tfr'l*-+L-

    .11 -.I

    Il.- -tajaaa.fa

    E*ectivb aare - Effeeiilire date +1 -

    1.5

    Price Cha Effective DateBut we also f ou nd thatNASDAQ prices gradually rosebetween the Announcementperiod and the subsequentEffective date. Further, thenearer the trade was to theclose, the greater the priceimpact. NYSE stocks appear tobe well behaved, but stocks

    with higher levels ofintermediation show thegreatest impacts andsubseouent reversion.

    The reversion aooears to begreatest for stocks with themost movement atler 2 PM.Stocks that weakened actuallyrose the next day - this is apaitern aiscr noied byMadhavaniOushing. But alsonote that the NYSE stocks were

    not f u ly immune f rom reversion; it appears that eventhese stocks had some excess speculation in them thatwas released once the Effective date oassed.

    The observations suggest a strategy that puts anemphasis on trading early on the Effective date, orwaiting until some time after the fact. Trading late inthe day is, in general, the most expensive route. Butthere may be times when price deterioration prior tothe close due to excessive speculation provides anopportunity for prof itable trading. Note that these arethe exceotions. not the rule,

    .Cushing, D. Madhavan, A., Stock Returns and Tradingat the Close. Working paper Nov. 1999.

    1

    t 0.5'iIt0o\

    -0,5

    The S&P 400 stocCloseD0 Close>Dl OpenDl Open>D1 Close

    Effective:D0 Open>2 PM2 PM +3:303:30 ;CloseD0 Close>Dl OpenDl Open;Dl Close

    +22 bp+441 bp-13 bp

    +83 bp+7? hn-257 bp+44 bp

    -r vP+31 bp+15 bp-58 bprA hn

    +25 bp+71 bp+44 bp-64 bp-25 bp

    -23 bp +5 bF+615 bp +783 bp-124 bp -299 bp