ACTIVE VERSUS PASSIVE INVESTING - AN ANALYSIS OF UK EQUITY MARKETS,

Transcription

1 ACTIVE VERSUS PASSIVE INVESTING - AN ANALYSIS OF UK EQUITY MARKETS, Barnes, E. 1 and Scott, M. Unversty College Cork, Ireland. ABSTRACT Ths study examnes the pattern of actve versus passve tradng n UK equtes over the perod We descrbe a metrc to analyse tradng actvty and volumes n the UK FTSE350 and AIM markets, wth emphass on ndustral and sze-based effects. 12,1 Our fndngs ndcate that actve stock pckng has been consstently declnng n the UK market over the perod studed for all markets, sze quntles and n vrtually every ndustral sector. Moreover, tradng patterns reveal a pronounced sze effect wth sgnfcantly less stock pckng n larger captalsaton stocks vs-à-vs smaller stocks. Patterns of nvestment n the AIM suggest an ncrease n ndex tradng over tme but hgher overall levels of stock pckng relatve to the FTSE350 lst. EFM Classfcaton: 350, Address for Correspondence: Dr Edel Barnes, Natonal Unversty of Ireland, Cork, O Rahlly Buldng, UCC, College Road, Cork, Ireland. Tel: Emal: e.barnes@ucc.e. 100

2 I. INTRODUCTION Theores of effcent markets, standard paradgms of academc and emprcal fnance, have clear mplcatons for asset combnaton and dversfcaton decsons. If markets are effcent and operate well, prces should reflect all avalable nformaton regardng frms fnancal poston and future prospects and t should not be possble to beat the market other than by chance. Investors should only be able to earn abnormal returns by havng access to prvate frm nformaton, superor forecastng ablty or through chance. In consequence, rather than ncur the sgnfcant prvate costs of research to obtan propretary nformaton, nvestors should be as well off nvestng (passvely) n a market ndex whch ncludes a broad range of dfferent securtes. Wth ths approach, the volume of trade n any partcular stock should reflect the weght of that frm n the market portfolo/ndex, and market weghtng should explan fully the varaton n volume of trade. However n a clmate of low nterest rates, as nvestors seek superor returns one mght expect sgnfcant actve nvestment as dstnct from passve nvestment. Ths leads to an upsurge n the use of skll and research on the part of professonal nvestors to dentfy msprced securtes and trade on that msprcng, a process whch s costly and whch offers no guarantee that benefts wll outwegh the very substantal costs of nformaton acquston and tradng. Carhart (1997) among others documents the magntude of actve vs-à-vs passve tradng costs and notes that, n terms of net returns, actvely managed nvestment funds have tended to under-perform ther passvely managed counterparts. 2 If the benefts of actve fund management consstently fal to outwegh the costs passve nvestment s surely more constructve for nvestors. 101

3 Despte extensve emprcal evdence on patterns n, and costs of actve vs-à-vs passve tradng, evdence largely relates to US markets and comparatvely lttle research has been conducted nto patterns of tradng n UK markets. Gven the global sgnfcance of the London market, we consder that an n-depth examnaton of actve vs-à-vs passve tradng patterns for the UK s merted, and that such evdence would represent an nterestng ncrement to the body of lterature at ths pont. The purpose of ths study s to examne the pattern of actve versus passve tradng n UK equtes over the perod nclusve. Drawng on the two fund separaton theorem (Lo and Wang, 2000; Bhattacharya and Galpn, 2005) we descrbe a metrc to analyse tradng actvty and volumes n the UK FTSE350 and AIM markets, wth emphass on ndustral and sze-based effects. 12,1 Our fndngs ndcate that actve stock pckng has been consstently declnng n the UK market over the perod studed for all markets, sze quntles and n vrtually every ndustral sector, whch evdence s consstent wth patterns of tradng documented for the US and some other markets. Our fndngs n respect of UK tradng patterns reveal a pronounced sze effect wth sgnfcantly less stock pckng n larger captalsaton stocks vs-à-vs smaller stocks. Patterns of nvestment n the AIM suggest and ncrease n ndex tradng over tme but hgher overall levels of stock pckng relatve to the FTSE350 lst. Our paper s structured as follows. The next secton presents an analyss of the theoretcal motvatons for and emprcal evdence pertanng to stock and ndex tradng and s followed by secton three whch descrbes our sample and the methodology we apply. The fourth secton outlnes the results of our analyss together wth a dscusson of those results and ther consstence wth the extant 102

4 lterature. In our fnal secton we dentfy some lmtatons of our analyss together wth some avenues for further study, and conclude. II. LITERATURE REVIEW Carhart (1997) among others documents the magntude of actve vs-à-vs passve tradng costs and notes that actvely managed nvestment funds have tended to be substantally more costly for nvestors reducng net nvestment returns. 2 Ths author examnes persstence n fund performance for equty mutual funds n the US for the perod and fnds that persstence s almost completely explaned by common stock factors and nvestment expenses. Over the long term he concludes that there s no sgnfcant momentum effect (the beneft of contnung to hold last year s wnnng stocks, dentfed by Fama and French, 1996) and that expense ratos, transactons costs and turnover are negatvely related to mutual fund performance. 4 Essentally hs fndngs are not supportve of the exstence of sgnfcant stock selectvty sklls among mutual fund managers for the perod of hs study. Jensen (1968) dentfed stock selecton ablty and dversfcaton/ rsk mnmsaton as separate fund management responsbltes and based on the Sharpe/Lntner CAPM model, examned fund managers predctve ablty n an analyss of US fund managers over the perod , the regresson ntercept term or alpha representng stock selecton ablty. 9 Hs fndngs ndcate that over the sample perod the mean fund was unable to generate suffcent returns to cover tradng costs and would not have outperformed a passve buy and hold nvestment approach. In lght of the hstorcally poor returns to actve fund management, Gruber (1996) queres why nvestors choose to buy actvely managed funds on fndng that 103

5 actve management adds value but that fund charges exceed ths value added. 8 Wermers (2000) re-examnes the value-added by mutual fund managers based on hypothetcal stocks-only funds and concludes that whle such funds outperformed the CRSP on average for hs study perod wth hgher turnover funds dong relatvely better, the net effect of transactons costs and non-stock holdngs resulted n hs sample funds underperformng a passve ndexng approach by 1% per annum on average. 17 Grnblatt and Ttman (1989, 1993) report mutual fund out performance consstent wth Wermers (2000) fndngs but ther fndngs wth respect to the substantal drag on net returns of actvely managed fund transactons costs are consstent wth Wermers. 6 A useful comment on the actve versus passve debate s provded n Malkel (2003). 13 In summary the body of lterature seems to ndcate that actve management does not justfy the fees typcally charged for ths servce. If the benefts of actve fund management are consstently neglgble or even negatve, passve nvestment s surely more constructve for nvestors and one would expect to observe ndexng as the domnant nvestment phlosophy f markets truly are effcent. To date the man body of lterature relatng to the prevalence of actve versus passve nvestors s not well developed. Such lterature as exsts regardng persstence n performance, effcent markets and mutual fund performance have been carred out n the US and typcally on US data. A Bhattacharya and Galpn (2005) paper ncorporates an mportant contrbuton to the debate by developng a metrc to measure ndexng. 1 These authors collected share volume and shares outstandng data from CRSP for NYSE, AMEX and NASDAQ lsted stocks for the perod July 1962 December 2004, and for 43 other markets around the world from DataStream for the perod January 1995 July 2004, 104

6 n order to conduct cross sectonal monthly regressons. The 43 non-us markets are classfed as emergng markets (22) and developed markets (21). A key fndng s that there appears to be more stock pckng n emergng markets (maxmum 63%) vs-àvs developed markets (maxmum 45%), whch result s ntutve gven the greater coverage of stocks and sounder nsttutonal arrangements n developed markets. Important exceptons are Germany whch appears to have more stock pckng than one would expect for a developed country (maxmum 71%) and Russa whch appears to have surprsngly lttle stock pckng (maxmum 35%). Notably the maxmum proporton of stock pckng was lowest n the US wth 29% and greatest n Chna (maxmum 80%). A further key fndng s that stock pckng appears to be declnng systematcally around the world, wth ths declne beng most pronounced n emergng markets although the US data reveal a declne to a low of 24% n the 2000s compared to an average level of stock pckng n the late 1960s of 60%. When these authors examne ther US data more mnutely some further trends and patterns are apparent. Consstent wth the practcaltes of ndexng, the practce s sgnfcantly more extensve for S+P 500 vs-à-vs non-s+p 500 stocks although ndexng appears to be ganng n popularty for both categores of shares. Share turnover s also relatvely greater for the larger non-s+p 500 shares. At all ponts examned, ndexng seems to be greater for NYSE-lsted vs-à-vs AMEX-lsted stocks and ndexng n the NASDAQ resembled that n the AMEX n the 1980s but more closely resembled tradng n the NYSE post-2000 at whch tme stock pckng n NASDAQ-traded stocks started to declne notceably. There has been a consstent declne n stock-pckng over tme n all three markets however, and an apparent sze effect as there seems to be greater ndexng n larger stocks across all the US markets examned. Furthermore, parttonng by age, the authors fnd less stock pckng n 105

7 older stocks vs-à-vs young frm stocks. Agan stock-pckng s observed to be n declne across frms of all ages and across the 10 Fama and French (1997) ndustry classfcatons, although the maxmum proporton of stock pckng s hgher n telecommuncatons whch the authors descrbe as exctng relatve to borng utltes. Bhattacharya and Galpn hypothesse that analysts have expertse n dentfyng msprced stocks and pck stocks that others should pck later. 1 Usng IBES data on analyst followng they fnd, nconsstent wth ther prors, that nvestors conduct more stock pckng n stocks that analysts do not pck and hypothesse that ths seems plausble f by undertakng and actng on ther own research analysts consequently reduce the payoff to stock pckng on one s own account. Agan stock pckng appears to be n declne across both analyst-followed and non-followed stocks wth ndexng beng more pronounced n stocks followed by greater numbers of analysts. In lght of fndngs that stock pckng s declnng across all markets and subdvsons of the data studed, Bhattacharya et al. queston the long-run steady state fracton of stock-pckers and develop a model based on frm specfc rsks and payoffs, tradng costs and the market prce of rsk (the market Sharpe rato) whch s then appled to US data for the perod Ther fndngs suggest that frmspecfc rsk has been ncreasng over tme and that stock-pckng has declned n tandem. At a long-run estmate of a net beneft to stock-pckng measure, they estmate a steady state maxmum proporton of stock-pckng of approxmately 11%, at whch level the authors predct that stock-pckng wll eventually settle n the US. The Unted Kngdom s one of the developed markets examned by Bhattacharya et al. (2005). 1 In terms of world rankngs of stock pckng, the UK ranks 9 th (21 st ) over the perod ( ) respectvely wth a maxmum proporton of stock pckng 106

8 of 47% (51%) respectvely. Whle the estmated dfferental s not large, t s nevertheless nterestng that the UK s one of very few markets n whch the extent of ndexng actually declned over that perod, n consequence of whch we consder that a fuller exploraton of tradng patterns n the UK mght yeld noteworthy fndngs. We also perceve the potental to examne more closely the role of ndustry, and of frm age or establshment n lght of the exstence snce 1995 of tradng n the UK Alternatve Investment Market (AIM). It s to ths analyss that we now turn. III. DATA AND METHODOLOGY The man objectve of our analyss here s to nvestgate, llustrate and explan any varaton n the patterns of actve vs-à-vs passve equty tradng over the perod nclusve for the FTSE350 and AIM markets, and specfcally to explore any trends n stock-pckng versus ndexng for the perod. We seek to ascertan the extent to whch tradng volume s explaned by stock pckng n the UK, whether there s a sze and/or ndustry effect n such tradng and whether patterns that apply to the FTSE350 man lst are also apparent n AIM tradng. Our methodologcal approach s based on that of Bhattacharya and Galpn (2005). 1 Ther metrc draws on nsghts of Lo and Wang (2000) who n turn base ther theoretcal dscusson on Tobn s (1958) two-fund separaton theorem. 12,16 Brefly, f the two-fund separaton theorem holds and everybody n the world ndexes between a rsk-free asset and a value-weghted proxy for the market portfolo, wth no prce changes between trades, share turnover for each stock defned as share tradng volume scaled by number of shares outstandng, should be dentcal for all stocks n the portfolo. Essentally (dollar) tradng volume n any stock should be entrely explaned by the market captalzaton of that stock. Regressng share tradng volume on number of shares 107

9 outstandng for each stock would yeld a beta of 1 and an R 2 =1 f all nvestment n the market s ndexng. To the extent that R 2 dffers from 1, there has been a devaton from ndexng whch could reflect ether stock pckng or alternatve nvestment strateges such as ndexng to an alternatve market ndex, hedgng dervatve postons etc. Thus R 2 n the followng regresson Ln( VOL) = α + β Ln( NOSH ) + ε [1] represents the proporton/extent of ndexng n a gven market and (1-R 2 ) represents the maxmum proporton of nvestment tradng that can be explaned by stock-pckng. VOL s the monthly volume of shares traded scaled by market captalsaton, NOSH s the value of shares outstandng for each stock at the end of that tradng month (adjusted for closely held shares). The ntercept term α represents the log of turnover, and the regresson coeffcent β descrbes the relaton between tradng volume and shares outstandng. The error term may be nterpreted as a measure of abnormal volume at the frm-level. It s mportant to note that our stock-pckng metrc represents the maxmum volume of shares traded that can be explaned by stock pckng, as t mplctly assumes that nvestors are ndexers or not. (R 2 wll dffer from 1 f agents ether pck ndvdual stocks n whch to nvest or alternatvely ndex to talored portfolos such as hedge funds of funds or exchange traded funds, whch latter have enjoyed ncreasng popularty n recent tmes.) The metrc does not dstngush between stock pckng and the actvtes of hedge funds and funds of funds for example. However we consder that ts appeal les n ts smplcty, understandablty and ease of computaton, requrng nether a hghly quanttatve background nor apprecaton of complex statstcs for ts comprehenson. It yelds a 108

10 measure whch by default descrbes the extent of ndexng n the market and n consequence allows us to nfer trends n approaches to nvestment over the perod studed. Our analyss of the nature of stock tradng actvty n the UK centres on the FTSE350 lst whch we consder offers a happy medum between the small number of stocks that consttute the FT100 man lst and the larger FTALLSH ndex whch would present consderable data challenges. For comparatve purposes we also analyse tradng patterns for the newer AIM market whch commenced tradng n June, 1995 and whch offers smaller frms an opportunty to access captal wthout the rgorous lstng requrements of a full lstng. Companes that lst and enjoy share tradng on the AIM are typcally smaller and younger than those on the man lst. For each month over the perod January 1991 December 2005 we obtan (aggregate) tradng volume and NOSH data (at month end) for every frm n our sample and conduct monthly regressons as n equaton [1] above. To be ncluded n our sample a share must be on ordnary common share and be lsted n ts own country. There was some varaton n the consttuents of the 350 lst, some companes dsappearng over tme and others not havng obtaned a lstng untl after the sample perod commenced. We select at random 210 companes on whch to base our analyss, representng 60% of the consttuent frms at any pont n tme. These data were obtaned from DataStream. For our sze analyss we partton our sample companes nto quntles accordng to market captalsaton for every month, quntle 1 (5) contanng the largest (smallest) stocks by market value respectvely and we conduct dfference of means tests on (1-R 2 ) measures to assess any sze effect. For our ndustry analyss we base our analyss on the DataStream ndustry classfcatons (25). Some categores had fewer than 4 companes so we reclassfed these frms under the 109

11 other classfcaton, resultng n 17 dstnct groupngs for the FTSE350 sample. Codes rangng from 1-17 nclusve were accorded to each frm to facltate our dfferentaton by ndustry. We do not seek to explore the exstence of a sze or ndustry effect n our AIM sample for whch just 10 years of data were avalable January 1996 December 2005 and we omt the perod 1 June December 1995 to allow for market settlng n ths ntroductory tradng perod. Our metrcs of key nterest are R 2 and by extenson (1-R 2 ) whch represent the proporton off ndexng (maxmum proporton of stock pckng) respectvely, though the ntercept term whch represents log of turnover also provdes some useful hnts about the absolute volume of trade n the varous data sets. We conduct the Ryan-Joner test of normalty and the Durbn-Watson test for autocorrelaton and fnd no non-statonarty n our data. Skewness s predctably a feature as turnover s necessarly bounded by 0 whch nduces postve skewness. We employ the Whte test for heteroskedastcty, agan ths s not a feature of our data though t mght plausbly have been present n such tme-seres data. In consequence we utlse OLS and base our tests of sgnfcance on parametrc P-values and (Fscher) F-statstcs, and our t-statstcs are of the 2-sded test of the null β=1. As the error term n our cross-sectonal regresson represents a measure of abnormal volume at the frm level, we obtan monthly returns for each frm over the sample perod from DataStream and relate them to ths abnormal volume measure as follows: R t = α + β ( AVol) + ε [2] t where R t s the frm-level return for frm n month t, AVol t s abnormal volume from equaton [1], α, β are regresson coeffcents and e the error term, to explore 110

12 whether abnormal volume mght have explanatory or predctve power for returns. Table I below descrbes our data for both FTSE350 and AIM companes at 31 December 2005, the end pont of our sample perod. Table 1 about here Clearly and unsurprsngly the mean FTSE350 frm s larger, enjoys sgnfcantly greater aggregate monthly tradng volume and has sgnfcantly greater numbers of shares outstandng than ts AIM counterpart. There s no mnmum market captalsaton requrement for an AIM lstng and the FTSE350 market has substantally greater market lqudty. IV. RESULTS Table 2 presents the results of applyng equaton [1] above to our FTSE350 data, where R 2 (1-R 2 ) represent the proportons of ndexng (maxmum stock-pckng) respectvely. Our sample perod pre(post)-dates that of Bhattacharya and Galpn (2005) by some 4 (1) years. 1 We are unclear about the specfc stocks that consttute ther UK lst so that comparsons are somewhat problematc other than n general mport and theme. Table 2 about here Throughout our beta value s greater than 1 at the 1% level so that whle volume was approxmately lnear n NOSH an ncrease n shares outstandng resulted n a greater percentage change n the volume of tradng wth ths effect beng more pronounced through tme. Our F-statstcs suggest that the regresson s hghly sgnfcant n every perod studed. R 2, the measure of proportonate ndexng shows a clear trend 111

13 upwards and there s a correspondng declne n the extent of stock-pckng and other non-ndexng trades, whch accords both wth our prors and wth evdence for the US and other markets documented by Bhattacharya and Galpn (2005). 1 Our mean (maxmum proporton of) stock pckng at 31% appears lower than the medan reported by Bhattacharya et al. of 49% and we report a systematc declne n stock pckng over tme whle Bhattacharya reports a slght ncrease n stock-pckng for the later years n hs sample (to 51% for the perod). 1 Our dfference of means tests ndcate that the level of ndexng was sgnfcantly lower n 1991 relatve to both the average over (t-stat 21.62, p-value 0.000) and to the level recorded for 2005 at the end of our sample perod (t-stat 24.61, p-value 0.000). These fndngs are consstent wth those of Bhattacharya et al. (2005) who document a declne from 60% to 24% over the perod 1960s-2000s for US markets. 1 Fgure 1 below hghlghts ths pronounced declne n stock pckng over tme for the FTSE350: Fgure 1 about here Table 3 below presents the fndngs of our sze analyss for quntles of the FTSE350 where quntle 1 (5) represents the largest (smallest) stocks by market captalsaton respectvely and metrcs are mean values for the perod. For all quntles the model statstcs ndcate sgnfcance at the 1% level and there s a clear sze effect evdent n the data wth ndexng beng sgnfcantly greater n larger stocks vs-à-vs smaller ones. Dfference of means tests confrm ths sze effect, (t-stat 22.05; p-value 0.000), and also that wthn each quntle there has been a systematc and sgnfcant declne n stock pckng over tme, a pattern that s evdent n Fgure 2 below. Fgure 2 about here 112

14 Table 3 about here These fndngs are consstent wth Bhattacharya and Galpn (2005), who document a smlar sze effect and tme trend for US stocks. 1 We are unsurprsed wth these data, stock pckng tends to be more prevalent n markets where there s less publc dsclosure of stock-specfc nformaton and analyst followng (and consequent publcaton of prce-senstve nformaton) s greater for larger captalsaton stocks. For our ndustry analyss we partton our FTSE350 stocks nto the DataStream classfcatons as dscussed n Secton Three above. The mean number of companes per ndustry was wth a maxmum (mnmum) of 19 (4) respectvely. Table 4 presents our fndngs wth respect to these groupngs for the perod : Table 4 about here There s consderable varaton n the relatve domnance of each nvestment phlosophy across ndustry type wth stock pckng n the Electrcal and Utlty (Chemcal and Pharmaceutcal) ndustres beng sgnfcantly greater (less) than the mean. Whle not reported here, our (1-R 2 ) measures ndcate a systematc declne n stock pckng over the perod studed for every ndustral groupng. To an extent our fndngs are consstent wth those of Bhattacharya and Galpn who report greater ndexng n the borng utlty sector as do we, however we fnd no exctng telecoms effect, stock pckng n ths UK sector havng fallen over tme rather than the reverse whch appears to have been the US experence. If analysts mprove the nformaton envronment of the stocks they research and pck, thus reducng the benefts of stock pckng, t seems ntutve that the returns to nformaton gatherng and n consequence stock pckng wll be greater n stocks that 113

15 have less analyst followng. In the UK stocks n the FTSE350 have wdespread followng but ths s much less the case n AIM-lsted stocks whch tend to be smaller, younger, start-up enterprses wthout the tradng hstory or vsblty of larger stocks. Table 5 reports our ndexng (non-ndexng) metrcs for AIM-lsted stocks for the perod nclusve, the AIM havng commenced tradng only n June Table 5 about here In 1996, the frst full year of tradng n AIM-lsted stocks thn tradng would lkely be a feature of the exchange and the concept of ndexng substantally premature formal ndexng essentally became possble only from 01/1999 when approxmately 320 frms were lsted though AIM frm numbers were 500 from For the perod as a whole, mean ndexng s ncreasng albet 2000 saw somewhat of a resurgence of the stock pckng practce, whch effect s lkely due to the popularty of hghtechnology and start-up stocks at the tme, a large number of whch would have been lsted on the AIM. Only towards the end of our sample perod does volume traded approach lnearty wth shares outstandng, n the earler years of the exchange s exstence, volume traded fell substantally short of outstandng shares. When we compare ndexng (non-ndexng) n the FTSE350 wth the AIM (for the perod ) our dfference of means tests ndcate that ndexng was sgnfcantly greater (lower) n the FTSE350 (AIM) stocks overall and n each calendar year, and our ntercepts suggest greater turnover n FTSE350 shares but more tradng n larger AIM stocks vs-à-vs smaller ones. At the end of our sample perod ndexng n the FTSE350 averaged 79.5% compared wth 45.8% for the AIM. From a practcal perspectve (and ndexng s the practcal manfestaton or mplementaton of the tenets of modern portfolo theory) ndexng s of course far easer for the FTSE

16 stocks and would not have been possble before 1999 for the AIM. However both groups ndcate a systematc trend upwards n ndexng at the expense of stockpckng, whch s more pronounced n the AIM, possbly because stock pckng started at a substantally greater level, also because the declne n stock pckng for the FTSE350, for whch we have a longer tme seres of data, pre-dates ths comparatve perod. It remans to be seen whether levels of stock-pckng for these exchanges wll converge over tme or whether there s always lkely to be somewhat less ndexng n the AIM vs-à-vs the FTSE350, whch pattern has been observed for the NASDAQ relatve to the NYSE for the US market. There are substantal tax breaks avalable to nvestors that buy and hold AIM-lsted stocks, whch provde a dsncentve to more actve tradng n ndvdual shares. Fgure 3 below depcts these tradng patterns for UK markets for the perod : Fgure 3 about here In summary, we report here a sgnfcant declne n stock-pckng for both FTSE350 and AIM markets over tme, whch results are robust to sze and ndustry sector, and n general small captalsaton stocks are shown to attract greater stock pckng actvty than larger captalsaton stocks. Whle we do not undertake any systematc analyss of an abnormal tradng volume assocaton wth frm returns, we dentfy ths area as potentally yeldng nterestng research fndngs movng forward. V. SUMMARY In effcent markets asset prces fully reflect all avalable frm-specfc nformaton and t should not be possble to beat the market other than by chance. If asset prces do not reflect all relevant nformaton, t may be possble to earn superor 115

17 returns by undertakng research to dentfy value-relevant frm nformaton and takng acton thereon. The purpose of ths study was to examne the pattern of actve versus passve tradng n UK equtes over the perod nclusve. Our metrc to analyse tradng actvty and volumes on the UK FTSE350 and AIM markets draws on the two fund separaton theorem (Lo and Wang, 2000; Bhattacharya and Galpn, 2005), and we explore ndustral and sze-based effects. 12,1 Our fndngs ndcate that actve stock pckng has been consstently declnng n the UK market over the perod studed for all markets, sze quntles and n vrtually every ndustral sector, although the AIM dd see a bref resurgence of stock pckng around at the heght of the dot-com nvestment bubble. Moreover, tradng patterns n the larger captalsaton FTSE350 lst reveal a pronounced sze effect wth sgnfcantly less stock pckng n larger captalsaton stocks vs-à-vs smaller stocks. Patterns of nvestment n the AIM suggest an ncrease n ndex tradng over tme but hgher overall levels of stock pckng relatve to the FTSE350. Ths s lkely due to the shorter hstory of the AIM and the characterstcs of stocks traded thereon; however t wll be nterestng to observe whether tradng patterns converge wth those of the FTSE350 as has been observed for the NASDAQ vs-à-vs the NYSE markets, when we have a longer tme seres of data for the AIM. Our results are not especally surprsng and are largely consstent wth those of Bhattacharya and Galpn (2005) although we do report a level of stock pckng for our FTSE350 that s substantally less than that whch BG report for ther undefned UK market. 1 If our consttuent stocks are on average larger than thers, taken n conjuncton wth our pronounced sze effect there may be a resoluton of the dfferental here. We fal to fnd any well-defned exctement/boredom factor n patterns of ndustral tradng, though we report the greatest relatve extent of 116

18 ndexng n the Chemcal and Pharmaceutcal sector whch s charactersed n the UK by relatvely small numbers of large captalsaton stocks. The evoluton of the estmated stock pckng mpact over tme s an ssue whch merts some consderaton. The perod covered by our study spans several cycles n the market from recesson through recovery and expanson and back to recesson from whch a further recovery eventually materalsed n the early years of ths century. Over the perod, many economc and geo-poltcal events occurred whch may well have nfluenced nvestor behavour for whch our statstcs proxy. Throughout that perod there has been a pronounced and consstent declne n actve stock-pckng at the expense of passve nvestment, and we remnd ourselves that our stock-pckng metrc ncludes all non-ndexng behavour. Essentally and notwthstandng developments such as the ntroducton of ETFs and the ncreasng actvty of hedge funds, actve nvestment appears to be n declne. Whether ths declne contnues or whether stock pckng wll eventually settle at some long run steady state level remans to be seen. Was ths level, f t emerges, to be sgnfcantly lower than the mean level we report at the end of our sample, such a development would have serous mplcatons for fnancal actvty n The Cty and for a fund management ndustry whch has exhbted unprecedented growth n the past decade. We recognse the smplstc nature of the BG metrc we compute n respect of stock pckng n that t essentally measures all non-ndexng nvestment behavour. 1 An nterestng avenue for further study nvolves a more granular exploraton of the mpact f any of Exchange Traded Funds on nvestors decson choces and whether ths relatvely low-cost nvestment approach whch amplfes the net returns dfferental between ndexng vs-à-vs actve nvestment has substantally hastened 117

19 the observed declne n stock-pckng. These are themes whch we hope to pursue movng forward. 118

20 Table I: Descrptve Statstcs Market FTSE350 AIM No. of Companes Mean MV m 6, Mean Volume 000s Mean NOSH 000s Varables Ln(VO) Ln(NOSH) Ln(VO) Ln(NOSH) No of obs Mean Medan SE (Mean) Std. Devaton Mnmum Maxmum Skewness Kurtoss Durbn-Watson Ryan-Joner MV=Market Captalsaton; Volume=aggregate volume of shares traded per month; NOSH=number of shares outstandng at end of calendar month, statstcs presented are averages across all shares n each lst respectvely at 31 December,

21 Table 2: (Maxmum Proporton of) Stock Pckng n the FTSE350 Year R 2 (1-R 2 ) F-stat Beta P-value Model: Ln( VOL) = α + β Ln( NOSH ) + ε Vol = volume of shares traded; NOSH = number of shares outstandng; measures are mean annual results based on monthly regressons descrbed by the model. 120

22 Fgure 1: Stock Pckng n the FTSE Level of Stock Pckng n the FTSE350 Level of Stock Pckng n the FTSE FTSE350 Average Level of Stock Pckng Year 121

23 Fgure 2: (Maxmum Proporton of) Stock Pckng n the FTSE Stock Pckng wthn Sze Analyss 100.0% 90.0% Average Level of Stock Pckng 80.0% 70.0% 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0% Year Quntle 1 Quntle 2 Quntle 3 Quntle 4 Quntle 5 122

24 Table 3: (Maxmum Proporton of) Stock-Pckng n the FTSE350 by Market Captalsaton, Quntle R 2 (1-R 2 ) F-stat Beta P-value ( ) Model: Ln( VOL) = α + β Ln( NOSH ) + ε Vol = volume of shares traded; NOSH = number of shares outstandng; measures are mean annual results based on monthly regressons descrbed by the model. 123

25 Table 4: (Maxmum Proporton of) Stock Pckng n the FTSE350 by Industry Industry R 2 (1-R 2 ) Beta F-stat P-value Rank Electrcal + Utltes Real Estate Equty Investment Other* Telecoms Aero Defence Computers Food, Drugs, Retal Food Producers Household G+S Support Servces Engneerng, Transport Travel + Lesure Insurance Meda Banks + Gen Fnance Chemcal, Pharmaceutcal Model: Ln( VOL) = α + β Ln( NOSH ) + ε Vol = volume of shares traded; NOSH = number of shares outstandng; measures are mean annual results based on monthly regressons descrbed by the model. Other* classfcaton ncludes 40 companes from the followng ndustres; auto and parts, beverages, tobacco, Personnel, H/C and Servces, Mnng, Constructon, for whch there were fewer than 4 frm-ndustry observatons. 124

26 Table 5: (Maxmum Proporton of) Stock Pckng n the AIM Perod R 2 (1-R 2 ) F-stat Beta P-value Model: Ln( VOL) = α + β Ln( NOSH ) + ε Vol = volume of shares traded; NOSH = number of shares outstandng; measures are mean annual results based on monthly regressons descrbed by the model. 125

27 Fgure 3: (Maxmum Proporton of) Stock-Pckng n the FTSE350, AIM for the perod nclusve. Stock Pckng n the AIM Versus FTSE Level of Stock Pckng AIM FTSE Year 126

28 REFERENCES Bhattacharya, U and Galpn, N. (2005). Is Stock Pckng Declnng Around the World? AFA 2007 Chcago Meetngs Paper, avalable at SSRN: Carhart, M. (1997). On Persstence of Mutual Funds. The Journal of Fnance, Vol. 52, No. 1, Economst. (2006) Passve Aggresson; Amercas Shareholders Settle for Guaranteed Medocrty: January, 76, and Alpha Bettng, September 16 th, 85. Fama, E and French, K. (1996). Multfactor explanatons of asset prcng anomales. Journal of Fnance Vol. 51, Grnblatt, M. and Ttman, S. (1989). Mutual Fund Performance: An analyss of quarterly portfolo holdngs. Journal of Busness, Vol. 62, Grnblatt, M. and Ttman, S. (1993). Performance measurement wthout benchmarks: An examnaton of mutual fund returns. Journal of Busness, Vol. 66, Grossman, S. and Stgltz, J. (1980). On the Impossblty of Informatonally Effcent Markets. The Amercan Economc Revew, Vol. 70, No. 3, Gruber, M. (1996). Another Puzzle: The Growth n Actvely Managed Mutual Funds. The Journal of Fnance, Vol. 51, No.3, Jensen, M. (1968). The Performance of Mutual Funds n the Perod The Journal of Fnance, Vol. 23, No. 2, Karpoff, J.M. (1987). The Relaton between Prce Changes and Tradng Volume: A Survey. Journal of Fnancal and Quanttatve Analyss Vol. 22, No. 1, Lntner, J. (1965). Securty Prces, Rsk and maxmal Gans from Dversfcaton. Journal of Fnance, XIX, Lo, A. and Wang, J. (2000). Tradng Volume: Defntons, Data Analyss, and Implcatons of Portfolo Theory. The Revew of Fnancal Studes, Vol. 13, No. 2, Malkel, B.G. (2003). The Effcent Markets Hypothess and ts Crtcs. European Fnancal Management, Vol. 9, No. 1, Markowtz, H. (1952). Portfolo Selecton. The Journal of Fnance, Vol. 7, No. 1, Sharpe, W. (1964). Captal Asset Prces: A Theory of Market Equlbrum Under Condtons of Rsk Journal of Fnance, XIX,

29 Tobn, J. (1958). Lqudty preference as behavour towards rsk. Revew of Economc Studes 25, Wermers, R. (2000). Mutual fund Performance: An Emprcal Decomposton nto Stock-Pckng Talent, Style, Transacton Costs and Expenses. Journal of Fnance Vol. 55, No.4,