Po-Ka Huang The Polcy Performance of NFSF and Slppage n Futures Markets (Receved Apr 28, 2011; Frst Revson Jun 8, 2012; Second Revson Dec 12, 2013; Accepted Jan 16, 2014) Introducton * Government nterventons n foregn exchange markets, settng of nterest rates, and stock markets are often observed. Although many emprcal artcles have examned the effectveness of these nterventons, few studes have focused on government nterventons n the futures market. Investgaton of the October 2000 nterventon n the futures market by the Natonal Fnancal Stablzaton Fund (NFSF) provdes a good opportunty to fll ths research gap regardng the effectveness of nterventon n futures markets. The NFSF was establshed n February 2000 for the purpose of respondng to occurrences at home or abroad so as to preserve stablty n captal markets and other fnancal markets and guarantee natonal stablty. The NFSF Management Commttee mplements the fund operatonal strategy and ensures ts procurement, deployment, and admnstraton. The NFSF frst ntervened n the Tawan Futures Exchange (TAIFEX) n October 2000 n order to prop up the TWSE Captalzaton Weghted Index (TAIEX). Usng ntraday NFSF tradng data obtaned from the TAIFEX, ths paper nvestgates the polcy performance of the NFSF regardng Tawan s stock ndex futures markets, focusng on assessment of futures slppage to evaluate the polcy objectves of the NFSF. A concept smlar to market mpact costs n stock markets, slppage can be decomposed nto ts lqudty effects and nformaton effects. Ths study explores the slppage caused by NFSF actons to examne whether the NFSF s able to provde lqudty to and sgnal postve nformaton n the futures markets. To our knowledge, ths s the frst study to explore ths ssue usng unque ntraday NFSF tradng data. Data and Research Desgn Ths study analyzes a unque set of ntraday NFSF tradng data for the perod from October 19, 2000 to November 15, 2000. Ths data set provdes a detal Po-Ka Huang s an Assocate Professor of Department of Fnance at Shh Hsn Unversty. E-mal: pkhuang@cc.shu.edu.tw hstory of order flows and transactons wth whch to examne the slppage, lqudty, and nformaton effects of trade by the NFSF, whch manly buys and sells TAIEX futures. In addton to ndvdual trades, trade packages are used to assess the extent of block tradng by the NFSF. If nsttutonal nvestors execute large orders through a seres of smaller trades, ther consecutve purchases (sales) generate upward (downward) prce pressure on the market over a perod of tme. Ths prce pressure wll contnue untl the nsttutonal nvestor exts the market, thus preventng a reversal n prce untl the order s fully executed. When an ndvdual trade s part of a block order and the benchmark prce s ntraday, t can appear to have an nformaton effect, regardless of whether the trade actually contans nformaton. Therefore, gven that some large trades by the NFSF are splt nto a sequence of smaller trades, Frno et al. (2008) s followed n aggregatng trades f they satsfed the followng crtera: (1) they have the same trade dentfcaton,.e., the NFSF; (2) they are n the same trade drecton,.e., buy or sell; and (3) they consecutvely execute wthout a one-day tradng break. Ths study defnes buy packages and sell packages to nclude successve purchases or sales, respectvely, of futures by the NFSF. Followng Holthausen, Leftwch, and Mayers (1987) and Frno et al. (2008), slppage s decomposed nto a lqudty effect and an nformaton effect as follows: Slppage ln Prce OpenngPrce 100 Lqudty Effect ln Prce ClosngPrce 100 Informaton Effect ln ClosngPrce OpenngPrce 100 (1) (2) (3) For each ndvdual trade, openng prce s the prce for frst trade durng regular tradng hours on the day of the trade, closng prce s the prce for the last trade durng regular tradng hours on the day of the trade, and prce s the trade prce. For each trade package, openng prce s the prce for the frst trade durng regular tradng hours on the frst day of the Management Revew Vol. 33 (July 2014), 109-116
package, closng prce s the prce for the last trade durng regular tradng hours on the same day of the package, and prce s the volume-weghted average prce of the package. To ncrease the robustness of the emprcal results, prces generated 15 mnutes before and 15 mnutes after an ndvdual trade or each trade package are also used as the openng prce and closng prce, respectvely. Indvdual trades are categorzed nto one of the followng three trade sze classes: trade sze 1, between one and fve lots; trade sze 2, between sx and nne lots; and trade sze 3, between 10 and 1,000 lots. For comparson, trade packages are also categorzed nto the followng three trade sze classes: trade sze 1, equal to or less than 100 lots; trade sze 2, between 114 and 600 lots; and trade sze 3: equal to and above 650 lots. To test whether slppage, lqudty, and nformaton effects are postvely correlated wth NFSF trade sze n futures markets, regresson analyss s performed usng the followng model: s 3 0 k Dummy SzeDummy, k k 1 ( ) (4) where s represents the slppage, lqudty effect, and nformaton effect of th ndvdual trade or trade package, respectvely; Dummy denotes the dummy varable that equals 1 f the th ndvdual trade or trade package s a buy order and 0 f otherwse; and SzeDummy,k also denotes the dummy varable and equals 1 f the th ndvdual trade or trade package falls nto the kth trade sze class and equals 0 f otherwse. Emprcal Analyss and Results Table 1 presents descrptve statstcs hghlghtng the characterstcs of the NFSF ndvdual trade data. The sample conssts of 529 buy and 976 sell trades. In terms of tradng frequency, the most actve tradng occurred n the largest trade sze class for ndvdual buy trades, but n the smallest trade sze class for ndvdual sell trades. Lkewse, the most actve tradng occurs n the largest trade sze class for buy trade packages, but n the smallest trade sze class for sell trade packages (data not shown). Analyss of these fndngs ndcates that the nterventon strategy of the NFSF s manly based on seldom large buy trade and small sell trade many a tme. executon of ndvdual trades conveys nformaton. These trades can thus strengthen nvestor confdence. Executon of buy orders also provdes lqudty because the lqudty effect s negatve for all trade szes and the prce further ncreases after executon of an NFSF buy order. As Holthausen, Leftwch, and Mayers (1987) have shown, ths suggests that the seller of a large block gves the NFSF a prce concesson as compensaton for nventory and search costs, and the block prce s below the prevous equlbrum prce. However, when redefnng a group of NFSF ndvdual trades as a block trade, analyss of ts trade package produces no evdence of a statstcally sgnfcant slppage, lqudty, or nformaton effect (table not shown here). Table 3 shows the results of regresson analyss. As can be observed n Table 3, the lqudty and nformaton effects are not necessarly related to trade sze. Ths phenomenon may be owng to the fact that some NFSF large trades are splt nto a sequence of smaller trades. Table 4 shows the results regardng the relatonshp between NFSF trade package sze and slppage. Most of the coeffcents are not sgnfcant, ndcatng that the lqudty and nformaton effects are not related to the NFSF trade package nterventon. These results are consstent wth the fndng of Frno, Kruk, and Lepone (2007) that the sgnfcant nformaton effect prevously observed n ndvdual trades loses ts statstcal sgnfcance once the trade package benchmarks are appled to all ndvdual trades n the package. Applyng the method used by Bhanot and Kadapakkam (2006), a robustness check s performed to verfy that the NFSF trade packages contan no nformaton effects. To do so, TX, TF, and TE stock ndex futures are consdered substtutes, although ther underlyng assets dffer somewhat, and NFSF buys TX and not TF and TE. Accordng to the nformaton effect hypothess, the NFSF s decson to defend the value of TX provdes a valuable sgnal to the traders of TX. The consequent proppng up of TX s value should have been benefcal to TF and TE, although the mpact would be more drect for TX. Table 5 shows that none of the stock ndex futures experences sgnfcantly abnormal average returns durng the nterventon perod, wth TX experencng a return of 0.5927%, TF of 0.1763%, and TE of 0.9898%. Table 5 also shows that no sgnfcant dfferences are found regardng abnormal returns among these three stock ndex futures, consstent wth Frno and Oetomo s (2005) fndng that trade packages executed n futures markets convey lttle nformaton. Table 2 reports the fndngs regardng slppage, lqudty, and nformaton effects of NFSF executon of ndvdual trades. Executon of buy orders has a statstcally sgnfcant nformaton effect that s larger than that of the sell orders, ndcatng that NFSF 110 Management Revew, July 2014
Frequency Prce Medan Volume Total Medan Table 1 Descrptve Statstcs by the NFSF s Indvdual Trade Sze Categores All Trade Sze 1 Trade Sze 2 Trade Sze 3 529 976 153 326 60 449 316 201 6,002.85 5,869.64 6,028.25 5,882.69 6,083.77 5,878.83 5,975.19 5,827.95 6,150.00 5,928.00 6,100.00 5,930.50 6,190.00 5,930.00 6,110.00 5,920.00 18,205 8,060 454 1,170 441 3,529 17,310 3,361 34.41 8.26 2.97 3.59 7.35 7.86 54.78 16.72 13.00 7.00 3.00 4.00 7.00 8.00 26.50 12.00 The Polcy Performance of NFSF 111
Table 2 Slppage, Lqudty, and Informaton Effects for NFSF s Indvdual Trades Slppage Lqudty Effect Informaton Effect Panel A 15 mn Trade Sze 1 0.0562 ** 0.1691 *** -0.1514 *** 0.0420 0.2076 *** 0.1271 *** Medan 0.0000 0.0000 *** -0.0969 *** 0.0000 *** 0.2197 *** 0.0000 ** Trade Sze 2 0.0496 0.1485 *** -0.1411 ** 0.0310 0.1907 ** 0.1175 *** Medan -0.0483 0.0000 *** -0.1873 ** 0.0000 *** 0.2393 0.0000 *** Trade Sze 3 0.2365 *** -0.0264-0.1079 ** 0.0728 ** 0.3444 *** -0.0991 *** Medan 0.0729 *** 0.0000-0.0484 ** 0.0000 ** 0.2098 *** 0.0000 *** All 0.1631 *** 0.1194 *** -0.1242 *** 0.0433 ** 0.2874 *** 0.0761 *** Medan 0.0484 *** 0.0000 *** -0.0807 *** 0.0000 *** 0.2195 *** 0.0000 *** Panel B Trade Sze 1-0.1781 2.1988 *** -0.0058 0.6021 *** -0.1722 1.5968 *** Medan 0.2432 2.1598 *** -0.1614 0.0161 *** 0.8084 2.2840 *** Trade Sze 2 0.6871 *** 2.2312 *** -0.2622 ** 0.4388 *** 0.9494 *** 1.7924 *** Medan 0.7279 *** 2.2166 *** -0.0807 *** 0.0000 *** 0.8084 *** 2.6818 *** Trade Sze 3 0.4119 *** 3.0020 *** -0.1964 *** 0.3134 *** 0.6083 *** 2.6886 *** Medan 0.6634 *** 4.1457 *** -0.0806 *** 0.0000 *** 0.8084 *** 4.3147 *** All 0.2725 *** 2.3791 *** -0.1487 *** 0.4675 *** 0.4212 *** 1.9116 *** Medan 0.6251 *** 2.2840 *** -0.0807 *** 0.0000 *** 0.8084 *** 2.8248 *** 15 mn refers to those prces generated 15 mnutes before and 15 mnutes after an ndvdual trade or each trade package. refers to the frst trade and last trade prce durng regular tradng hours on the day of the trade. ***, **, and * ndcate sgnfcance at the 1%, 5%, and 10% level respectvely 112 Management Revew, July 2014
Table 3 Regresson Results between NFSF s Indvdual Trade Szes and Slppage Slppage Lqudty Effect Informaton Effect 15 mn 15 mn 15 mn Intercept 0.0562-0.1781-0.1514-0.0058 0.2076-0.1722 (1.9319) * (-1.5761) (-4.3686) *** (-0.0623) (4.2259) *** (-0.9675) Dummy SzeDummy2-0.0066 0.8652 0.0103-0.2564-0.0169 1.1216 (-0.1200) (4.5141) *** (0.1422) (-1.6089) (-0.1706) (4.0240) *** Dummy SzeDummy3 0.1804 0.5900 0.0436-0.1905 0.1368 0.7805 (3.4277) *** (3.8022) *** (0.8044) (-1.6609) * (1.8766) * (3.6541) *** Dummy SzeDummy1 0.1129 2.3769 0.1934 0.6079-0.0805 1.7690 (2.5897)*** (15.4698) *** (4.26135) *** (5.1776) *** (-1.2616) (7.5337) *** Dummy SzeDummy2 0.0924 2.4093 0.1824 0.4446-0.0901 1.9646 (2.4744) ** (17.6322) *** (4.13271) *** (4.1010) *** (-1.4843) (9.3044) *** Dummy SzeDummy3-0.0825 3.1801 0.2242 0.3193-0.3067 2.8608 (-1.8632) * (18.3013) *** (4.85912) *** (2.9763) *** (-4.9905) *** (12.2258) *** Adj R 2 0.0175 0.2654 0.0167 0.0663 0.0257 0.1047 F statstc 6.3577 *** 109.6621 *** 6.1078 *** 22.3528 *** 8.9433 *** 36.1645 *** 15 mn refers to those prces generated 15 mnutes before and 15 mnutes after an ndvdual trade or each trade package. refers to the frst trade and last trade prce durng regular tradng hours on the day of the trade. The numbers n the parentheses are t-statstcs computed usng Whte (1980) heteroscedastc consstent standard errors. ***, **, and * ndcate sgnfcance at the 1%, 5%, and 10% level respectvely. The Polcy Performance of NFSF 113
Table 4 Regresson Results between NFSF s Trade Packages Szes and Slppage Slppage Lqudty Effect Informaton Effect 15 mn 15 mn 15 mn Intercept 0.4219 0.8067 0.0843-0.0902 0.3376 0.8969 (2.0117) ** (1.4401) (0.4576) (-0.1285) (2.8415) *** (1.0824) Dummy SzeDummy2-0.4425-1.3644 0.4957 0.0042-0.9382-1.3686 (-1.4995) (-1.8608) * (1.8625) * (0.0053) (-2.6604) *** (-1.3025) Dummy SzeDummy3 0.8753 1.2067-0.1506 1.9466 1.0258-0.7400 (2.0814) ** (0.6690) (-0.4591) (1.4040) (1.7714) * (-0.6203) Dummy SzeDummy1 0.3039 0.0177 0.4429 0.9764-0.1390-0.9587 (1.1568) (0.0295) (1.9481) * (1.2405) (-0.5771) (-1.0640) Dummy SzeDummy2-0.6290 0.2671 0.5537 0.3812-1.1827-0.1141 (-1.5043) (0.1824) (1.5049) (0.4668) (-2.3613) ** (-0.0790) Dummy SzeDummy3 0.4754 1.1176 0.0286 0.8465 0.4468 0.2712 (0.7336) (1.0310) (0.1295) (0.8761) (0.6612) (0.1595) Adj R 2 0.2091-0.0065 0.0744 0.0123 0.2391-0.0627 F statstc 3.2205 ** 0.9460 1.6756 1.1042 3.6399 *** 0.5046 15 mn refers to those prces generated 15 mnutes before and 15 mnutes after an ndvdual trade or each trade package. refers to the frst trade and last trade prce durng regular tradng hours on the day of the trade. The numbers n the parentheses are t-statstcs computed usng Whte (1980) heteroscedastc consstent standard errors. ***, **, and * ndcate sgnfcance at the 1%, 5%, and 10% level respectvely. 114 Management Revew, July 2014
Table 5 Robustness Check for Informaton Effect Perod Delvery Months Daly Average N TX TF TE Dfference (TX-TF) Before Interventon 89/9/21~89/10/18 89/10 Raw returns 20-1.3296-0.3846-1.8742-0.9450 Abnormal returns 20-1.2823-0.2708-1.8718* -1.0115 (-1.49) (-0.35) (-1.99) (0.88) Durng Interventon 89/10/19~89/11/15 89/11 Raw returns 21 0.5455 0.0625 0.9873 0.4830 Abnormal returns 21 0.5927 0.1763 0.9898 0.4164 (0.88) (0.26) (1.07) (-0.43) After Interventon 89/11/16~89/12/20 89/12 Raw returns 27-0.3549 0.1150-0.6311-0.4699 Abnormal returns 27-0.3077 0.2287-0.6287-0.5364 (-0.53) (0.36) (-1.02) (0.63) Estmaton Perod 88/12/16~89/9/20 89/1~89/9 Raw returns 199-0.0472-0.1138-0.0025 0.0666 TX refers to TAIEX Futures. TF refers to Fnance Sector Index Futures. TE refers to Electronc Sector Index Futures. Daly average s dsplayed as a percentage. The numbers n parentheses below the abnormal returns are the correspondng t-statstcs. ***, **, and * ndcate sgnfcance at the 1%, 5%, Dfference (TX-TF) 0.5446 0.5895 (-0.46) -0.4418-0.3971 (0.35) 0.2762 0.3210 (-0.38) -0.0447 The Polcy Performance of NFSF 115
Concluson Ths study nvestgates the effectveness of mplementaton of NFSF polcy on the TAIEX futures market usng unque ntraday NFSF tradng data. The polcy objectves of NFSF are provdng lqudty and sgnalng postve nformaton. Analyss of futures slppage, a varable smlar to market-mpact cost on stock markets that can be decomposed nto ts lqudty and nformaton effects, s performed to evaluate the effectveness of NFSF polcy. The results ndcate that executon of ndvdual trades by the NFSF conveys nformaton, provdng lqudty and the earnng of a lqudty premum n the sample perod. However, executon of NFSF trade packages does not convey nformaton. Fnally, the results ndcate that the lqudty and nformaton effects of NFSF trades are not necessarly postvely related to trade sze. REFERENCE Bhanot, Karan and Palan-Rajan Kadapakkam (2006), Anatomy of a Government Interventon n Index Stocks: Prce Pressure or Informaton Effects? Journal of Busness, 79(2), 963-986. Frno, Alex and Teddy Oetomo (2005), Slppage n Futures Markets: Evdence from the Sydney Futures Exchange, Journal of Futures Markets, 25(12), 1129-1146. ----, Jennfer Kruk, and Andrew Lepone (2007), Transactons n Futures Markets: Informed or Unnformed? Journal of Futures Markets, 27 (12), 1159-1174. ----, Johan Bjursell, George H. K. Wang, and Andrew Lepone (2008), Large Trades and Intraday Futures Prce Behavor, Journal of Futures Markets, 28(12), 1147-1181. Holthausen, Robert W., Rchard W. Leftwch, and Davd Mayers (1987), The Effect of Large Block Transactons on Securty Prces: A Cross- Sectonal Analyss, Journal of Fnancal Economcs, 19(2), 237-267. Whte, Halbert (1980), A Heteroskedastcty- Consstent Covarance Matrx Estmator and a Drect Test for Heteroskedastcty, Econometrca, 48(4), 817-838. 116 Management Revew, July 2014