Table of Content Chapter 1:- Conceptual Overview Chapter 2:- Research Methodology Objective of Study Scope and Rationale of Study Methodology i!itation of Study Chapter ":- #heoretical $ac%ground • • • •
Chapter &:- Case Study ' (ntroduction of Co!pany profile and )roduct *bout the wor% in co!pany done by students Chapter +:- ,ata *nalysis Chapter :- .indings $ibliography *nne/ure
1
CHAPTER - 1
2
Chapter 1:- Conceptual Overview Recent research has suggested that aggregate !ar%et li0uidity varies over ti!e and that the covariance of returns with innovations in !ar%et li0uidity is priced owever3 li0uidity has !ultiple di!ensions which incorporate %ey ele!ents of volu!e3 ti!e and transaction costs *n ideal !easure of !ar%et-wide li0uidity should therefore incorporate ele!ents of depth3 breadth and resiliency res iliency #his paper esti!ates !easures !easure s of !ar%et-wide li0uidity along each of these di!ensions di!ensions and finds that each !easure4s innovations are correlated3 that covariance of stoc% returns and innovations in each !easure is priced3 and co!bining the infor!ation in each !easure i!proves the precision of esti!ates of li0uidity ris% pre!ia ( esti!ate the li0uidity ris% pre!iu to be appro/i!ately 2-+5 per year and show that this pre!iu! is distinct fro! fir! si6e3 a security7s security7s individual li0uidity3 li0uidity3 and the covariance between changes in a security4s individual li0uidity and !ar%et-wide li0uidity *s a byproduct3 ( also docu!ent that the li0uidity ris% pre!iu! has a strong 8anuary seasonal3 which is unrelated to fir! si6e (!pact of li0uidity on various aspects of econo!y * very basic definition of li0uidity is 4the cash or !oney in a syste!4 i0uidity is !easured in ter!s of the !onetary base and the Reserve $an% of (ndia 9R$( is the sole supplier of li0uidity in the country (n general3 the supply of !onetary base by the central ban% depends on the public4s de!and for currency and the ban%ing syste!4s need for reserves to settle or discharge pay!ent obligations #hee R$( #h R$( !o !onit nitors ors the li0ui li0uidit dity y situat situatio ion n on a daily daily basis basis and and atte! atte!pt ptss to contro controll and !ode !o dera rate te li0u li0uid idit ity y cond condit itio ions ns by vary varyin ing g the the supp supply ly of ban% ban% rese reserv rves es to !eet eet its its !acroecono!ic objectives of financial stability #he periodic li0uidity assess!ent is done by the R$( based on the ban% reserves position3 and the the e/pect e/pected ed inflow inflowss and and outflo outflows ws fro! fro! both both do! do!es esti ticc opera operati tions ons and forei foreign gn flow flows s ,epending on the li0uidity forecast3 the R$( decides on a course of action to be ta%en to either supple!ent or withdraw li0uidity #hese are so!e of the factors that influence li0uidity conditions in the econo!y: ,o!estic factors *n increase in li0uidity is re0uired to cover inflation and ;,) growth Several instantaneous do!estic factors also influence the li0uidity in the syste! 3
Most co!!only3 0uarterly or annual advance ta/ pay!ents draw li0uidity out of the syste! as a lot of li0uid !oney gets loc%ed with the govern!ent On the other hand3 any large payouts by the govern!ent or higher corporate sector spending can increase the li0uidity in the syste! .unds inflows * strong econo!ic perfor!ance and the relative under-perfor!ance in the developed countries attracted the attentions of !any large global investors who were drawn towards investing here 9 .,( as well as portfolio invest!ents #his resulted in healthy capital inflows in the last few years #hese capital inflows put a lot of pressure on the li0uidity !anage!ent here as uncontrolled capital flows can result in rising inflation3 currency appreciation3 loss of co!petitiveness and reduction in !onetary control #ools to control li0uidity #he R$( !onitors the li0uidity situation periodically and ta%es necessary steps to control the situation fro! ti!e to ti!e #he R$( uses various direct and indirect policies to control the shortter! and long-ter! li0uidity position #hese are various instru!ents used by the R$( to control li0uidity: Cash reserve ratio: #he R$( uses the cash reserve ratio 9CRR as a tool to control the !ediu! to long-ter! li0uidity issues *n increase in the CRR results in an increase in the a!ount of !oney that ban%s have to !aintain with the R$( as a percentage of their deposits #his reduces the overall li0uid funds with the ban% and hence reduces the overall li0uidity i0uidity adjust!ent factor: #he li0uidity adjust!ent factor 9*. was introduced a decade ago as a part of financial refor!s *. helps in !anaging a shortter! li0uidity situation resulting fro! the large and volatile capital flows 9inflows as well as outflows Reverse repo rate: #he R$( uses the reverse repo rate for short-ter! li0uidity !anage!ent and to s!oothen interest rates in the call
4
#he repos also help in %eeping the interest rates in a predictable range3 as provided by the prevailing repo rate and reverse repo rate (n ti!es of e/cess visible li0uidity3 the call rates hover around the reverse repo rate3 whereas in ti!es of tight li0uidity3 the call rate will hover around the repo rate i0uidity i!pacts inflation *n uncontrolled and un!anaged li0uidity situation can have a severe i!pact on inflation3 rates of interest3 S#OC= M*R=>#S3 and foreign e/change rates Since the conditions in the global !ar%ets and foreign fund flows are 0uite volatile3 the job of the R$( in controlling the li0uidity condition has beco!e !ore challenging #he R$( has ta%en s!all steps in changing the !onetary policy since the beginning of this year #hese steps have shown good results in ter!s of !aintaining interest rates3 li0uidity and ;,) growth #he inflation rate is still ruling high due to various factors and analysts believe that further !onetary actions fro! the R$( along with a good rainfall and base effect will !oderate it in the ne/t couple of 0uarters DEFIITIO !ar"et Capitali#ation
#he total dollar !ar%et value of all of a co!pany4s outstanding shares Mar%et capitali6ation is calculated by !ultiplying a co!pany4s shares outstanding by the current !ar%et price of one share #he invest!ent co!!unity uses this figure to deter!ine a co!pany4s si6e3 as opposed to sales or total asset figures >?)*(@S 4Mar%et Capitali6ation4 (f a co!pany has "+ !illion shares outstanding3 each with a !ar%et value of A1BB3 the co!pany4s !ar%et capitali6ation is A"+ billion 9"+3BBB3BBB / A1BB per share Co!pany si6e is a basic deter!inant of asset allocation and ris%-return para!eters for stoc%s and stoc% !utual funds #he ter! should not be confused with a co!pany4s capitali6ation3
5
which is a financial state!ent ter! that refers to the su! of a co!pany4s shareholders4 e0uity plus long-ter! debt #he stoc%s of large3 !ediu! and s!all co!panies are referred to as large-cap3 !id-cap3 and s!all-cap3 respectively (nvest!ent professionals differ on their e/act definitions3 but the current appro/i!ate categories of !ar%et capitali6ation are: arge Cap: A1B billion plus and include the co!panies with the largest !ar%et capitali6ation Mid Cap: A2 billion to A1B billion S!all Cap: ess than A2 billion
!ar"et capitali#ation #he @ew Dor% Stoc% >/change onEall Street3 the world4s largest stoc% e/change per total $ar"et capitali#ation of its listed co!paniesF1G Market capitalization or market cap is the total dollar !ar%et value of the shares outstanding
of a publicly traded co!panyH it is e0ual to the share price ti!es the nu!ber of shares outstandingF2GF"G *s outstanding stoc% is bought and sold in public !ar%ets3 capitali6ation could be used as a pro/y for the public opinion of a co!pany4s net worth and is a deter!ining factor in so!e for!s of stoc% valuation #he invest!ent co!!unity uses this figure to deter!ine a co!pany4s si6e3 as opposed to sales or total asset figures #he total capitali6ation of stoc% !ar%ets or econo!ic regions !ay be co!pared to other econo!ic indicators #he total !ar%et capitali6ation of all publicly#R*,>, COM)*@(>S in the world was ISA+12 trillion in 8anuary 2BBJF&G and rose as high as ISA+J+ trillion in May 2BBKF+G before dropping below ISA+B trillion in *ugust 2BBK and slightly above ISA&B trillion in Septe!ber 2BBKF+G Since 2BBL3 when $itcoin beca!e the first decentrali6ed cryptocurrency and nu!erous cryptocurrencies 9altcoins have been created3 the 4!ar%et cap4 ter! has also co!e into co!!on use to describe the total dollar !ar%et value of the total a!ount of cryptocurrency in circulation 9available supplyFGFJG #he ter! so!eti!es can refer to the esti!ated !ar%et value of the total a!ount of cryptocurrency that will ever be in circulation 9total supplyFKG
Calculation
6
Mar%et cap is given by the for!ula
3 where MC is the !ar%et
capitali6ation3 @ is the nu!ber of shares outstanding3 and ) is the price per share .or e/a!ple3 if so!e co!pany has & !illion shares outstanding and the price per share is A2B3 its !ar%et cap is then AKB !illion (f the price per share rises to A213 the !ar%et cap beco!es AK& !illion (f it drops to A1L per share3 the !ar%et cap falls to AJ !illion
!ar"et cap ter$% #raditionally3 co!panies were divided into lar&e-cap3 $i'-cap3 and %$all-capF2G #he ter!s $e&a-cap and $icro-cap have also since co!e into co!!on use3FLGF1BG and nano-cap is so!eti!es heard ,ifferent nu!bers are used by different inde/esHF11G there is no official definition of3 or full consensus agree!ent about3 the e/act cutoff values #he cutoffs !ay be defined as percentiles rather than in no!inal dollars #he definitions e/pressed in no!inal dollars need to be adjusted over the decades due to inflation3 population change3 and overall !ar%et valuation 9for e/a!ple3 A1 billion was a large !ar%et cap in 1L+B3 but it is not very large now3 and they !ay be different for different countries
Relate' $ea%ure% Mar%et cap reflects only the equity value of a co!pany (t is i!portant to note that a fir!4s choice of capital structure has a significant i!pact on how the total value of a co!pany is allocated between e0uity and debt * !ore co!prehensive !easure is enterprise value 9>3 which gives effect to outstanding debt3 preferred stoc%3 and other factors .or insurance fir!s3 a value called the e!bedded value 9> has been used
7
CHAPTER - (
8
Chapter (:- Re%earch !etho'olo&) • • • •
Ob*ective of +tu') +cope an' Rationale of +tu') !etho'olo&) ,i$itation of +tu')
9
CHAPTER -
10
Chapter :- Theoretical .ac"&roun' *sset li0uidity occupies an i!portant3 but elusive3 position in the study of asset pricing Mar%et !icrostructure research has !ade it clear that li0uidity providers offer a real service $uyers and sellers !ay not arrive in the !ar%et si!ultaneously3 creating a role for li0uidity providers to transact and hold securities on a te!porary basis1 i0uidity providers are co!pensated for their e/pense and ris% e/posure via the bid
specialists and @*S,*N !ar%et !a%ers perfor! this function3 however individual investors !ay also provide li0uidity via li!it orders 2 *!ihud and Mendelson use the bid-as% spread as a pro/y for li0uidity3 $rennan and Subrah!any!an use fi/ed and variable co!ponents of transactions costs esti!ated fro! !icrostructure data3 $rennan3 Chordia and Subrah!anya! use trading volu!e3 and ,atar3 @ai%3 and Radcliffe use share turnover
such a state variable and securities differ in their return covariances with !ar%et li0uidity3 then li0uidity betas should be priced One natural approach to investigating this 0uestion is to follow the !ajority of the characteristic stoc% li0uidity literature and esti!ate !easures of syste!ic li0uidity by aggregating !icrostructure data3 but this approach suffers fro! at least two practical proble!s .irst3 the large volu!e of data per unit of ti!e !a%es it difficult to 11
co!pute even the !ost basic aggregate li0uidity !easure Second3 even the longest ti!e series of transaction data is short co!pared to the availability of lower fre0uency data )astor and Sta!baugh 92BB2 devise a !easure of the price reversal 9resiliency di!ension of !ar%et-wide li0uidity utili6ing daily returns over a long ti!e period 91L2-1LLL Controlling for the usual ris% factors3 they find a positive relationship between stoc% returns and the covariance of return with their !easure of !ar%et-wide li0uidity Ising other di!ensions of li0uidity such as depth and breadth to construct !ar%et-wide li0uidity !easures appears to re!ain an une/plored area of research #his study as%s three 0uestions .irst3 are !easures of aggregate li0uidity using depth and breadth priced3 as )astor and Sta!baugh 92BB2 find for their resiliency !easure (n addition3 is it possible to aggregate e/posure to !easures derived using the three di!ensions of li0uidity to derive an esti!ate of the price of li0uidity ris% Second3 is it possible for investors who do not care about return sensitivity to li0uidity shoc%s to invest in a portfolio that is sensitive to li0uidity shoc%s but hedged against other co!!on sources of syste!atic ris%3 using only prior infor!ation3 and earn a li0uidity ris% pre!iu! .inally3 does the pre!iu! associated with high li0uidity beta stoc%s survive after controlling for !ar%et capitali6ation3 the covariance of a stoc%4s characteristic li0uidity with changes in aggregate li0uidity3 and the level of the stoc%4s characteristic li0uidity #o preview3 ( find that esti!ates of the li0uidity ris% pre!iu! of appro/i!ately 2-+5 per year are not sensitive to the approach used for !easuring !ar%et-wide li0uidity3 that a feasible invest!ent strategy earns appro/i!ately this return before transaction costs3 and that the result survives after controlling for the three alternatives listed above ( investigate the pricing of alternative li0uidity !easures by first calculating two variations of each of three types of aggregate li0uidity !easures based on !ar%et resiliency3 depth3 and breadth #he resiliency !easure relies on the principle that order flow induces greater return & reversals when !ar%et-wide li0uidity is low3 as in )astor and Sta!baugh 92BB2 #he second type of li0uidity !easure atte!pts to capture the depth of the !ar%et and reflects the average price i!pact per unit of trading volu!e #his !easure is closely related to that used by *!ihud 92BB2 #he third type reflects the breadth of the !ar%et and is derived fro! !icrostructure data on individual stoc% bid
history ( find that return covariance with shoc%s to aggregate li0uidity is priced for all three types of li0uidity !easures #he esti!ated ris% pre!iu! is positive3 however there is a significant negative 8anuary seasonal in the li0uidity pre!iu! that is not related to fir! si6e * feasible invest!ent strategy constructed to have positive e/posure to syste!ic li0uidity shoc%s but hedged against other co!!on ris% factors earns positive returns on average and negative returns when there is a shoc% to li0uidity #he relationship between li0uidity beta and return re!ains after controlling for !ar%et capitali6ation3 the covariance between stoc% li0uidity and !ar%et-wide li0uidity3 and the li0uidity level of the individual stoc%s (n other words3 the higher return earned by stoc%s with large li0uidity betas is not due to these stoc%s being s!all3 being the!selves illi0uid3 or beco!ing particularly illi0uid when there is a !ar%et-wide li0uidity shoc% #he rest of this paper is organi6ed as follows Section (( describes each of the !easures of aggregate li0uidity e/a!ined in this paper Section ((( tests whether li0uidity ris% as !easured by return covariance with shoc%s to the aggregate li0uidity !easures is priced Section ( investigates the relationship between li0uidity betas and the covariance of individual stoc% li0uidity with aggregate li0uidity3 stoc%4s characteristic li0uidity3 and fir! si6e Section concludes
II/ !ea%ure% of A&&re&ate ,i0ui'it) #his section defines two versions for each of three types of !ar%et li0uidity !easure ( show that each of the !easures is correlated with the others3 with !ar%et returns3 and with the si6e and boo%-to-!ar%et factor returns of .a!a and .rench
A/ The Price Rever%al !ea%ure )astor and Sta!baugh esti!ate a li0uidity !easure based on the idea that price changes acco!panying large volu!e tend to be reversed when !ar%et-wide li0uidity is low #his view of volu!e related return reversals arising fro! li0uidity effects is !otivated by Ca!pbell3 ;ross!an3 and Eang 91LL"3 where ris%-averse !ar%et !a%ers 9in the sense of ;ross!an and Miller 91LKK acco!!odate order flow fro! li0uidity !otivated traders and are co!pensated with higher e/pected return .or this type of !easure3 low !ar%et-wide li0uidity refers to those states where !ar%et !a%ers re0uire a higher e/pected return to acco!!odate a given order flow #he ordinary least s0uares esti!ate of 3 )R i t P is a pro/y for stoc% i4s li0uidity in !onth t Superscripts on li0uidity !easures are used to differentiate between the various !easures used *n upper case ? denotes a generic li0uidity !easure .ollowing )astor and Sta!baugh 92BB23 a stoc%4s li0uidity is co!puted in a given !onth only if there are !ore than 1+ observations fro! 13
which to esti!ate the regression 9213 it is not the first or last !onth that the stoc% appears on CRS)3 and the share price at the end of the previous !onth is between A+ and A1BBB #he !ar%et-wide li0uidity !easure is then constructed fro! the individual stoc% !easures by averaging all of the individual !easures during the !onth and inflating by the ratio of total !ar%et capitali6ation at the end of !onth t-1 to total !ar%et capitali6ation at !onth B See )astor and Sta!baugh 92BB2 for a detailed discussion of their !easure #he rational for inflating the average li0uidity !easure by the ratio of !ar%et capitali6ations !ay not be clear )astor and Sta!baugh 92BB2 argue that 3 )R i t P can be viewed as the li0uidity cost3 in ter!s of return reversal3 of trading A1 !illion of stoc% i3 averaged across all stoc%s Since A1 !illion was a relatively larger trade in the 1LBs than in the 1LLBs3 the si!ple average coefficient will fall through ti!e (nflating the coefficient adjusts for this condition One drawbac% of the )R li0uidity !easure is the use of dollar volu!e since e0ual si6e trades !ay have different i!pact due to differences in3 for e/a!ple3 the nu!ber of shares outstanding3 differences in float3 and differences in the nu!ber and types of shareholders One possible alternative is to substitute turnover 9dollar volu!e divided by end of previous !onth !ar%et capitali6ation for dollar volu!e although3 as )astor and Sta!baugh point out3 this is si!ilar to si!ply value weighting their !easure owever even this !easure would !iss variation in return i!pact of order flow due to3 for e/a!ple3 differences in float * second alternative3 une/plored in previous studies3 is to substitute turnover scaled by average daily turnover during the previous !onth for dollar volu!e in e0uation 921 #his !easure of order flow will capture any unusual volu!e at the e/pense of ease of interpretation but without the need to inflate the average coefficient by total !ar%et capitali6ation .igure 1a plots the ti!e series of scaled 3 )R i t P and .igure 1b plots 3 !)R i t P #he series are very si!ilar with large negative li0uidity levels in !onths where li0uidity is generally considered to be low including October of 1LKJ 9the crash3 which is the largest negative value in both series3 @ove!ber of 1LJ" 9the *rab oil e!bargo3 2nd and 12th largest negative levels respectively3 Septe!ber of 1LLK 9the Russian debt and #CM crisis3 &th and 2nd3 and October of 1LLJ 9the height of the *sian financial crisis3 1"th and Lth #he overall correlation between the two series is BJ1" #able 1 reports that both series display significant autocorrelation 9B21 and B1 for )R and !)R respectively
./ The Price I$pact !ea%ure 14
*!ihud 92BB2 esti!ates a li0uidity !easure based on price i!pact =yle 91LK+ argues that spreads are an increasing function of the probability of facing an infor!ed trader3 and since the !ar%et-!a%er cannot distinguish between order flow fro! infor!ed traders and order flow fro! noise traders3 she sets prices that are an increasing function of the order i!balance that !ay indicate infor!ed trading #his i!plies an inverse relationship between price i!pact and li0uidity *lternatively3 price i!pact !easures for a particular stoc% !ay be large for reasons unrelated to asy!!etric infor!ation issues or li0uidity .or e/a!ple3 when there is a news " O!itting October of 1LKJ fro! both series reduces the esti!ated correlation to BJ *lthough influential3 o!itting this observation does not have a large i!pact on the reported correlations of the levels or innovations of the li0uidity !easures K release which i!pacts fir! value but about which there is little disagree!ent3 price change can be large and volu!e s!all resulting in a large esti!ated price i!pact My use of the price i!pact !easure follows the spirit of *!ihud but is different fro! the individual stoc% or characteristic li0uidity approach Ehen !ar%etwide li0uidity is low3 price concessions re0uired fro! ;ross!an-Miller !ar%et !a%ers are larger per unit of volu!e than when !ar%et-wide li0uidity is high $y averaging price i!pact !easures across all stoc%s the idiosyncratic effects should diversify leaving only syste!atic li0uidity Ehether this is !easurable in practice is an e!pirical issue #he !easure is defined as the negative of the daily average so that large negative values signify 4low li0uidity4 consistent with the interpretation of the )R and !)R !easures #he !ar%etwide !easure is the si!ple average of the individual stoc% !easures #he resulting ti!e series is then inflated by the ratio of total !ar%et capitali6ation at the end of !onth t-1 to total !ar%et capitali6ation at the end of !onth B #he sa!e criticis!s that apply to the use of dollar volu!e in the )R !easure also apply hereH therefore ( also e/a!ine a !odified version of the price i!pact !easure: 9 3 3 3 1 3 3 1 1 ,n !)( idt i t d n idt r , S#O P Q Q T 92& #he aggregate !easure is the si!ple average of the individual stoc% !easures and is not rescaled by !ar%et capitali6ation L .igure 1c plots the ti!e series of scaled 3 )( i t P and .igure 1d plots 3 !)( i t P Scaled 3 )( i t P is highly serially correlated 9first order serial correlation3 U1QBKL with periods of especially low li0uidity in the early 1LJBs and again in 1LLL-2BBB 3 !)( i t P also shows si!ilar periods of illi0uidity although not as severe as 3 )( i t P #he correlation between the two !easures is B+& #he correlation !atri/ for all of the aggregate li0uidity !easures is shown in #able 1 *ll of the !easures are positively correlated with each other and we 15
can reject the null that each correlation is 6ero although the correlation between scaled 3 )R i t P and scaled 3 )( i t P is only BBL C Measures $ased on $id<*s% Spread *!ihud and Mendelson 91LK3 Chordia3 Roll3 and Subrah!anya! 92BBB3 asbrouc% and Seppi 92BB13 uber!an and al%a 92BB13 8ones 92BB13 $a%er and Stein 92BB2 and !any others e/a!ine the bid-as% spread as a !easure of the characteristic li0uidity of individual stoc%s *n investor wishing to trade i!!ediately !ay always sell 9buy at the 0uoted bid 9as% price that includes a concession 9pre!iu! for i!!ediate e/ecution #herefore the spread between the bid and the as% prices3 which is the su! of the concession and pre!iu!3 divided by the !idpoint of the spread3 is a natural !easure of li0uidity Ising (SSM data fro! 1LK"-1LL2 and #*C data fro! 1LL"-2BB13 ( calculate aggregate li0uidity !easures using all @DS><*M>? stoc%s as follows& .irst3 define RNSi3d3t as the daily average relative 0uoted spread for stoc% i on day d in !onth t RNSi3d3t is the average of every best bid and offer 9$$O eligible 0uote fro! the open until just prior to the !ar%et close divided by the 0uote !idpoint R>Si3d3t is defined as the daily average relative effective spread and is the average of the absolute value of the difference between each transaction price and the !idpoint of the !ost recent 0uote3 which is at least five seconds prior to the trade3 divided by the 0uote !idpoint #he aggregate li0uidity level during !onth t is: ,t i3d3t 1 dQ1 t 1 RNS , P Q Q RNS @t t i @t 92+ & ( a! grateful to M @i!alendran for providing the 0uote and effective spread data 1B ,t i3d3t 1 dQ1 t 1 R>S , P Q Q R>S @t t i @t 92 where @t is the nu!ber of fir!s in !onth t and ,t is the nu!ber of days in !onth t (ncreasing spreads are associated with decreasing li0uidity3 therefore the leading negative sign is added so that s!aller values of P are associated with lower li0uidity3 consistent with the other !easures .igure 1e plots the ti!e series of RNS t P and .igure 1f plots R>S t P for the period 1LK"-2BB1 $oth plots show an upward trend reflecting the falling 0uoted and effective spreads during the period #here are also large negative changes in the li0uidity !easure in October of 1LKJ and Septe!ber of 1LLK Consistent with the positive ti!e trend3 both series are strongly positively serially correlated , (nnovations in *ggregate i0uidity .or asset pricing purposes it is the covariance of asset returns with innovations in the aggregate li0uidity !easure that is i!portant #his is in contrast to characteristic li0uidity where the difference in li0uidity levels i!plies differences in transaction costs that !ust be co!pensated with e/pected return #o esti!ate innovations fro! levels3 ( calculate the first difference of each li0uidity !easure as: 9 3 31 1 1 P PP V VV Q WQ @t ? ? ?? t t it it i t M @ 92J where ? Mt Q 9!t-1
capitali6ation at ti!e t-1 and ti!e 6ero3 for ? Q )R and ?Q)( and Mt ? Q1 for all others ( then regress Vt WP on its lag as well as the lagged value of the scaled series: 1 1 3 1 VV V ? ? ?? ? t t j it t PP P a b cM u W Q TW T T 92K #hus3 the predicted change depends on the lag level and the lag change #he innovation in aggregate li0uidity is ? t u #o ease co!parison of results between li0uidity !easures in later sections3 ( rescale ? t u so that the standard deviation of the innovations is of the sa!e order of !agnitude for each !easure 11 1BB B1B 1B 1BB 1BB )R )R !)R !)R )( )( tt t t t t !)( !)( RNS RNS R>S R>S t tt t t t u u u u u u Q Q ⋅ Q ⋅ Q⋅ Q ⋅ Q ⋅ 92L #able 2 shows that 92K yields innovations that are serially uncorrelated for all !easures in the full sa!ple and in both subperiods (f the three di!ensions of !ar%et-wide li0uidity are related3 then we !ight e/pect the innovations to be correlated )anel * of #able 2 shows the correlation !atri/ of the innovations over the full sa!ple period *ll of the innovations are significantly positively correlated3 both in the full sa!ple and in each subperiod #he only e/ception is the correlation between the )R and )( !easures in the second subperiod3 which is a statistically insignificant BB #he significant correlations for the later subperiod in )anel $ between the breadth !easures esti!ated fro! !icrostructure data and the resiliency and depth !easures esti!ated fro! daily data are particularly encouraging (f !ar%et-wide li0uidity !easures can be esti!ated using low fre0uency data then the cost of esti!ation is greatly reduced and the !easures can be esti!ated over !uch longer ti!e periods and for !ar%ets for which transaction level data is not available .igure 2 plots the ti!e series of the innovations in aggregate li0uidity *ll show large negative values on si!ilar dates3 October of 1LKJ in particular3 although the !agnitude of these shoc%s varies )anel $ of #able 2 shows us that the !)( !easure is highly correlated with each of the !icrostructure based !easures with an esti!ated correlation of BJ& with each #he )R3 !)R3 and )( !easures are also significantly correlated with RNS and R>S #he correlation !atrices in #able 2 suggest a si!ilarity a!ong pro/ies but do not by the!selves i!ply that !ar%et li0uidity is a priced state variable > >!pirical .eatures of the i0uidity Measures )astor and Sta!baugh 92BB2 describe a flight to 0uality effect when their !easure of !ar%et li0uidity is low Months in which li0uidity is e/ceptionally low tend to be !onths in which stoc% returns and bond returns !ove in opposite directions #able " reports the correlation between the value-weighted CRS) inde/ of @DS>-*M>? stoc%s and three fi/ed inco!e variables: !inus the change in the rate on one-!onth #reasury bills3 the return on the thirty year 12 govern!ent bond3 and the return on a portfolio of long ter! 17
corporate bonds+ Over the full sa!ple period 1L2-2BB13 the correlation between !inus the change in the rate on one-!onth #reasury bills and the !ar%et return is near 6ero and between the !ar%et return and the bond returns is positive (n !onths of low li0uidity3 defined as a li0uidity shoc% !ore than two standard deviations below the !ean3 the correlation between the !ar%et return and both !inus the treasury bill return and the govern!ent bond return is negative3 regardless of the li0uidity !easure used to identify !onths of low li0uidity #he correlation between the corporate bond return and the !ar%et return is near 6ero when low li0uidity is defined using )R or !)R and negative when using )( or !)( )anel $ reports si!ilar figures for the 1LK"-2BB1 period and include the spread based 9breadth !easures of li0uidity #he results are very si!ilar to those in )anel * *lso shown in #able " is the correlation between the !ar%et return and the e0ually weighted average percentage change in !onthly dollar volu!e for @DS>-*M>? stoc%s #he unconditional correlation between volu!e changes and !ar%et returns is positiveH however3 regardless of the !easure used to identify !onths of low li0uidity3 when li0uidity is low3 !ar%et returns and changes in volu!e are negatively correlated #able & reports correlations between innovations in each li0uidity !easure and the valueweighted CRS) inde/3 the e0ual-weighted CRS) inde/3 and the .a!a .rench factors SM$ and M >ach !easure is positively correlated with both CRS) indicesH however the correlation is driven by !onths in which the !ar%et falls .or e/a!ple3 the )R !easure has a correlation of B2L with the value-weighted inde/3 but the correlation is 'BB2 in !onths in which the inde/ return is positive and B&& when negative >ach of the !easures is positively correlated with SM$ and negatively correlated with M Ehen li0uidity is low3 large stoc%s outperfor! s!all stoc%s and value outperfor!s growth #he correlations are larger in !agnitude and significance for the price i!pact and spread based !easures than for the reversal-based !easures (t is re!ar%able is that the si/ li0uidity !easures that address the three separate di!ensions of li0uidity appear so si!ilar Months of low li0uidity are !onths in which stoc% !ar%et returns fall3 large stoc%s outperfor! s!all stoc%s3 and value outperfor!s growth #he ne/t step is to + #he corporate bond data is fro! (bbotson *ssociates 1" e/a!ine the pricing i!plications of a stoc%4s return covariance with each of these !easures3 controlling for other co!!only used sources of ris% ((( #he i0uidity Ris% )re!iu! #his section investigates whether a stoc%4s e/pected return is related to the covariance of its return with innovations in each of the li0uidity !easures after controlling for other variables that have been found to be i!portant in asset pricing #o acco!plish this ( use a 18
portfolio-based approach where the portfolios are for!ed on the basis of predicted sensitivity to li0uidity shoc%s >ach !onth3 the universe of available stoc%s is sorted into ten portfolios by predicted li0uidity beta and held for one !onth #he portfolio returns are lin%ed through ti!e to for! a single return series for each decile portfolio #hese post for!ation returns are then regressed on return based factors that are co!!only used in e!pirical asset pricing studies #o the e/tent that the intercepts are different fro! 6ero3 li0uidity sensitivity e/plains a co!ponent of returns not captured by e/posure to other factors Specifically3 for each !onth t3 ( regress the e/cess stoc% return on the li0uidity innovation3 ?3 in a regression that also includes the .a!a and .rench 91LL" factors: B 3 M S ?? it i i t i t i t i t t r RMR. SM$ M QT T T T T XX X X X Y 9"1 where ? t is the innovation calculated using one of the si/ !ethods described above .or every !onth t between ,ece!ber 1L+ and ,ece!ber 2BB13 the regression is run for every stoc% whose end of !onth price at !onth t-1 is between A+ and A1BBB and which has valid return data in at least " !onths between t-1 and t-B *lthough it see!s natural to use the esti!ated regression coefficient lX ? i to sort stoc%s into portfolios3 it is well %nown that sorting on regression coefficients in this !anner is proble!atic3 especially when the standard errors of the regression coefficients are large #his is of particular concern for the regression 9"1 since the standard errors of lX ? i for individual stoc%s are very large and therefore sorting on lX ? i 3 in effect3 leads to sorting on esti!ation errors 1& #o !itigate this proble! ( use a $ayesian approach to for! the esti!ates of ? X i then sort into portfolios based on these esti!ates #he $ayesian esti!ates of ? X i are only used to sort stoc%s into portfolios3 all point esti!ates reported in the tables are the result of classical econo!etric techni0ues Specifically3 ( esti!ate 9"1 for every stoc% at !onth t3 then treat each esti!ate of the vector X Q FXB 3 XM3 XS 3 X3 X?G as a draw fro! a !ultivariate nor!al distribution with esti!ated covariance Z #he $ayesian esti!ate of X3 lbi 3 is then esti!ated as: l 9 9 9 l 9 X X QZ T Z T 1 1 1 2 2 4 4 i i b ?? ?? i s s i 9"2 where 9 1 2 4 s i ? ? is the esti!ated covariance !atri/ of lX i #he $ayesian esti!ate of X is a weighted average of the OS esti!ate of Xi and the average of X across all stoc%s at ti!e t where the weight on Xi is the inverse of the covariance !atri/ of Xi #his esti!ator shrin%s the esti!ate of the coefficient vector for each stoc% towards the population average with the a!ount of shrin%age an inverse function of the precision of the esti!ate for the individual stoc% )astor and Sta!baugh 9)S use a si!ilar approach to infer that their li0uidity !easure is priced3 although they use a different !ethod of sorting stoc%s into portfolios #hey !odel the ti!e variation in X ? i 19
e/plicitly using the full sa!ple up to ti!e t to esti!ate the para!eters ( prefer !y !ethod for sorting into portfolios for three reasons .irst3 although )S !odel ti!e variation in the li0uidity beta3 they assu!e the other factor loadings and the para!eters of the !odel for ti!e variation in li0uidity beta do not change over a sa!ple period of up to "+ years Second3 the !odel of ti!e variation proposed by )S captures very little of the variation in li0uidity betas as !easured by R2 3 and the coefficients are unstable through ti!e #hird3 the insa!ple loadings on innovations in li0uidity are not as the !odel predicts *ppendi/ * discusses the )S !ethodology3 elaborates on the above points3 and co!pares it to the !ethod used in this paper * *sset )ricing #ests ( test for the e/istence of a li0uidity ris% pre!iu! in two ways .irst ( esti!ate the abnor!al return to each predicted li0uidity beta sorted decile portfolio using the three-factor !odel of .a!a and .rench and e/a!ine the intercepts #he difference in abnor!al return between the 1+ e/tre!e deciles provides infor!ation about a co!ponent of e/pected returns not captured by the three-factor !odel #he second test uses the infor!ation in all ten decile return series to esti!ate the li0uidity ris% pre!iu! directly 1 .a!a .rench *lphas #he ti!e series returns for each li0uidity beta decile portfolio are regressed on the three .a!a-.rench factors that are co!!only used in e!pirical asset pricing studies #o the e/tent that the regression intercepts3 or alphas3 differ fro! 6ero3 X? e/plains a co!ponent of e/pected returns not captured by e/posure to the other factors #able +3 panel * shows the alphas fro! .a!a-.rench regressions of the e/cess return on each e0ual-weighted decile portfolio for each li0uidity !easure #he intercepts are generally negative for the portfolio of those stoc%s with the least sensitivity to li0uidity shoc%s and increasing as we !ove to portfolios with a greater sensitivity #he spread in intercepts between the !ost and least sensitive portfolios is positive3 ranging fro! B"2 to "11 percent per year @one of the spreads differs significantly fro! 6ero )anel $ reports results fro! value-weighting the decile portfolios #he pattern in intercepts in )anel * repeats although the pattern is less apparent .ive of the si/ spreads in intercept are positive and none of the spreads are statistically significant Since there is a si6e co!ponent to li0uidity3 s!aller fir!s are concentrated in the e/tre!e deciles3 it see!s appropriate to chec% for a 8anuary seasonal in order to verify that the li0uidity pre!iu! is not a rediscovery of the si6e<8anuary effectJ #o isolate the seasonal effect in the intercepts ( esti!ate 9"1 and 9"2 separately for 8anuaries and for all other !onths #he portfolio returns are the sa!e as those above #able reports the results Ehen the portfolios are e0ual-weighted3 the spread in annuali6ed intercepts for non20
8anuary !onths ranges fro! BL" to 2J percent .ew of the individual portfolio intercepts differ significantly fro! 6ero3 but both price reversal !easures and one of the two price i!pact .or each li0uidity !easure3 the ten e0uations are stac%ed and esti!ated using ;MM #he point esti!ates will be identical to those fro! e0uation by e0uation OS but the standard errors are corrected for autocorrelation and conditional heteros%edasticity #his !ethod !a%es tests of cross-e0uation restrictions si!ple J Inreported tests fail to identify a 8anuary seasonal in any of the si/ !ar%et-wide li0uidity !easures 1 !easures4 portfolio intercept spreads are statistically significant *lthough the spread in intercepts for the !icrostructure based !easures are of si!ilar !agnitude 9annuali6ed spreads of 2+ and 1KJ3 the shorter ti!e series results in an inability to reject that the spreads are 6ero (nterestingly3 the spreads in 8anuary are negative for five of the si/ !easures Only the RNS !easure3 which is based on 0uoted spreads and does not include transaction prices3 has a positive spread in 8anuary )anel $ of #able reports value-weighted results si!ilar to the e0ualweighted results .ive of si/ !easures have positive intercept spreads in non-8anuary !onths although none are statistically significant Only the RNS !easure is associated with a positive intercept spread in 8anuary #he difference in intercept spreads in 8anuaries vs @on-8anuaries presents an interesting pu66le >ven after including the SM$ factor3 the !odel of .a!a and .rench does not fully e/plain the 8anuary effect Since s!all fir!s are concentrated in the lowest and highest li0uidity beta sorted deciles3 we !ight be te!pted to argue that the inability of the three factor !odel to e/plain the 8anuary effect in s!all stoc% returns is confounding any li0uidity effects owever3 as ( will show later3 the negative 8anuary seasonal in the li0uidity pre!iu! is not confined to s!all stoc%s but e/ists across all si6e 0uintiles 2 ,irect >sti!ation of the i0uidity Ris% )re!iu! #he previous section infers the e/istence of a li0uidity ris% pre!iu! fro! the spread in abnor!al returns between the highest and lowest decile of predicted li0uidity sensitivity (t is also possible to esti!ate the li0uidity ris% pre!iu! directly using infor!ation fro! all ten portfolios .or each !easure of aggregate li0uidity ?3 define the ti!e series regression: B ? ? t t tt r $. e QT T T X X 9"" where rt is a 1B/1 vector of e/cess returns on the decile portfolios3 .t is a "/1 vector containing the reali6ations of the .a!a-.rench factors RMR.3 SM$3 and M3 $ is a 1B/" !atri/ of factor loadings3 X? is a 1B/1 vector of li0uidity betas3 and ? t is the innovation in aggregate li0uidity !easure ? *ssu!e the portfolios are priced by: 9 3 . ?? >r $ t Q T [ X [ 9"& 1J where > 9 i denotes the unconditional e/pectation3 and [i is the ris% pre!iu! for factor i Since . are returns on portfolios3 let 21
9 . [ Q > . # #a%ing e/pectations of both sides of 9""3 substituting 9"&3 and solving for XB gives: B 9 9 ?? ? X X[ Q > t 9"+ ( esti!ate the vector of para!eters b Q FXB $ X/ [?G using the ;eneral Method of Mo!ents of ansen 91LK2 #he ;MM esti!ator of b !ini!i6es g9b4 E-1g9b where g9b is the sa!ple average of ft9b3 9 9 9 4 4 B 1 3 t t t ? ? t t ? t tt ? ? tt t t h e f b > h . e r $. X X \ \
⊗ Q
\\\\
Q Q 9" and E is a consistent esti!ator of the spectral density of ft K #he esti!ates of the li0uidity ris% pre!iu! [? for each of the li0uidity !easures as well as the associated t-statistics for both e0ual-weighted and value-weighted portfolios are reported in #able J #he !agnitude of [? depends on the arbitrary scaling of ?3 but the scaling does not affect the t-statistics or the product X[3 therefore #able J also reports 9X1B-X1[ for each of the li0uidity !easures #he first colu!n uses the full ti!e series of predicted li0uidity beta portfolio returns and ignores the 8anuary seasonal3 and is co!parable to #able + #he second colu!n uses the sa!e ti!e series but drops all 8anuary observations fro! the sa!ple and is co!parable to #able Ehen portfolios are e0ual-weighted and the full sa!ple is used3 the esti!ated li0uidity ris% pre!iu! [ is positive for all si/ !easures and is statistically significant for four of si/ Ehen 8anuaries are dropped fro! the sa!ple3 five of si/ are significant3 and the point esti!ates are generally larger 9X1B-X1[ is always positive3 ranging fro! B&B to &2L percent per year using the full sa!ple and 11J to &"B percent per year using only non-8anuary !onths *ll of the K ( use an iterated ;MM esti!ator where the !o!ent conditions are e0ually weighted in the first step and the value of b that !ini!i6es the objective function used with the NS %ernel to esti!ate the spectral density of ft 1K values of 9X1B-X1[ are statistically significant with the e/ception of the !odified price i!pact !easure3 !)( Splitting the sa!ple shows the price of li0uidity to be appro/i!ately constant through ti!e )anel $ of #able J reports results using value-weighted portfolios #he results are si!ilar to the e0ual-weighted results >sti!ates of the return for bearing syste!atic li0uidity ris% as !easured by 9X1B-X1[ in non-8anuaries ranges fro! 1J2 to +B2 percent per year $ edged )ortfolio Returns (f a portfolio constructed to have a positive sensitivity to li0uidity ris% earns a ris% pre!iu! then we would e/pect the portfolio to do well on average and to do poorly when there is a !ar%et li0uidity shoc% #o test whether this is in fact the case3 ( for! a feasible portfolio for each li0uidity !easure that is long decile 1B 9high predicted li0uidity beta and short decile 1 9low li0uidity beta #he returns to this portfolio are then hedged for the usual .a!a-.rench ris% factor e/posure using factor loadings esti!ated using data fro! !onths t-1 through t-B #able K reports 22
the results Ehen the .a!a-.rench factor-neutral portfolios are e0ualweighted3 the li0uidity trading strategy earns fro! 1& to &2 basis points 91K5 and +B&5 annuali6ed per !onth #he profits fro! portfolios for!ed on the four non-!icrostructure data based !easures that are available for the full sa!ple period are all statistically significant .ive of the si/ portfolios have negative returns in !onths when li0uidity is low3 and all si/ have s!aller returns in low li0uidity !onths than in other !onths #he last three colu!ns drop 8anuaries fro! the sa!ple with little effect Ehen the feasible .a!a-.rench factorneutral portfolios are value weighted3 all si/ earn positive returns3 four of si/ are negative in low li0uidity !onths and five of si/ earn lower returns on average when li0uidity is low than in other !onths Statistical significance is generally lower than when portfolios are e0ual-weighted C Co!bining i0uidity Measures (f the three di!ensions of li0uidity are related3 then it should be possible to co!bine the infor!ation contained in each variable to i!prove our esti!ates of li0uidity ris% factor e/posures and li0uidity pre!ia #o this end ( for! a new set of decile portfolios based on the su! of the 1L portfolio assign!ents fro! each of the si/ individual li0uidity !easures .or each stoc% that has a portfolio assign!ent for each of the si/ !easures 9four prior to 1LKJ3 ( su! the portfolio nu!bers to which each is assigned then sort this su!!ary statistic into deciles ( then repeat the e/peri!ent outlined in Section * using the su!!ary deciles as test assets #able L presents the results #he spread in annuali6ed intercepts is a statistically significant 2&5 when portfolios are e0ual-weighted and 2115 when portfolios are value weighted Ee are unable to reject that the spread in intercepts is 6ero when portfolios are value-weighted $oth point esti!ates are within the range of those esti!ated in #able + using the individual li0uidity !easures *n inspection of the tstatistics associated with the individual intercepts shows that the esti!ation error associated with the intercepts is !uch s!aller with the aggregate !easure than with individual !easures #he annuali6ed spread in intercepts is a statistically significant "1"5 when portfolios are e0ualweighted and 8anuaries are o!itted and 2+5 when value-weighted ,irect esti!ation of the li0uidity ris% pre!ia using the !ethod of #able K is difficult since the portfolios have been for!ed based on infor!ation contained in all si/ li0uidity !easures owever3 it is possible to esti!ate the returns to an invest!ent strategy long decile 1B and short decile 1 for!ed using the aggregate !easure and hedged against any e/posure to the .a!a.rench ris% factors #he results to such a strategy are reported in )anel $ of #able L #he strategy using e0ual-weighted portfolios earns a statistically significant return of &" basis points per !onth 9+15 annuali6ed3 23
larger than the return earned by any of the si/ individual li0uidity !easure based strategies in #able K Ehen the portfolio is value-weighted3 the invest!ent strategy earns a statistically significant +B basis points per !onth3 again larger than that earned by the feasible invest!ent strategy based on any of the si/ individual li0uidity !easures (f a low li0uidity !onth is defined as a !onth in which all available li0uidity !easures are greater than two standard deviations below their !ean3 the e0ual weighted strategy earns an average '2&2 basis points in low li0uidity !onths and the value weighted strategy an average of T"" basis points per !onth *lthough the average !onthly return for the value weighted strategy is positive3 it is the average of only three observations #he !edian return of these three observations is -"2& basis points and the average return is below that of the other !onths (f 2B low li0uidity is defined as any li0uidity !easure being !ore than two standard deviations below is average3 then both e0ual and valueweighted strategies earn negative returns in low li0uidity !onths and significantly positive returns in other !onths ( (ndividual Stoc% i0uidity #he previous sections as% whether stoc%s whose return covaries with each of several !ar%etwide li0uidity !easures earn higher returns #his section as%s whether the covariance between stoc% return and !ar%et-wide li0uidity shoc%s 9li0uidity return beta is a pro/y for the covariance between changes in a stoc%4s characteristic li0uidity and !ar%et-wide li0uidity shoc%s 9li0uidity spread betas3 or a pro/y for the stoc%4s characteristic li0uidity level3 or si!ply a pro/y for !ar%et capitali6ation * i0uidity Spread $eta *!ihud and Mendelson 91LK develop a !odel in which e/pected returns are an increasing function of the bid
precipitated a fall in !ar%et li0uidity3 the value of the fund4s portfolio value dropped triggering a need to li0uidate positions to !eet !argin calls #he anticipation of #CM4s need to li0uidate further eroded the value of the fund4s positions )rior to 1LLK3 did #CM earn a li0uidity pre!iu! for 21 holding illi0uid securitiesL 3 holding securities whose returns were sensitive to li0uidity shoc%s3 or both #o investigate whether the covariance of individual stoc% li0uidity with aggregate li0uidity is priced3 each !onth ( regress changes in individual stoc% li0uidity !easures on lag changes in individual stoc% li0uidity !easures3 the lag level of the individual stoc%4s li0uidity3 and shoc%s to aggregate li0uidity: 3 31 31 3 V VV ? ? ? ?? it it it t it P PP a b c d u W Q TW T T T 9&1 where 3 V ? i t WP : #he change in characteristic li0uidity of stoc% i fro! !onth t-1 to t using li0uidity !easure ? 3 1 V ? i t P : #he characteristic li0uidity of stoc% i using li0uidity !easure ? ? t : #he shoc% to !ar%et-wide li0uidity at ti!e t using li0uidity !easure ? and ? corresponds to one of the si/ li0uidity !easures: )R3 !)R3 )(3 !)(3 RNS3 or R>S #he coefficient vector XQFa b c dG is adjusted using the $ayesian techni0ue described in section ((( ( then sort the stoc%s into deciles by the $ayesian esti!ate of the coefficient on aggregate li0uidity shoc%s3 ? ( refer to this coefficient as a li0uidity spread beta and refer to the li0uidity beta discussed in the first three sections as a li0uidity return beta #he resulting decile portfolios are then regressed on the .a!a.rench ris% factors and the difference in annuali6ed intercepts between deciles 1B and 1 is e/a!ined for evidence of variation in alpha across the decile portfolios in a !anner si!ilar to #ables + and #able 1B reports the results .or brevity the individual decile intercepts have been o!itted and only the annuali6ed difference in the e/tre!e deciles is reported #he first three colu!ns represent the difference in alphas for e0ual-weighted portfolios for the full sa!ple and for the sa!ple split by 8anuary vs @on-8anuary >/a!ining the @on-8anuary colu!n there is so!e evidence3 particularly for the !)( 9!odified price i!pact !easure and the RNS 9relative 0uoted L #his argu!ent applies to long positions Many of #CM4s trading strategies involved the si!ultaneous purchase of long and short positions in si!ilar securities whose prices were e/pected to converge 22 spread !easure that stoc%s which beco!e relatively !ore illi0uid when the !ar%et beco!es !ore illi0uid3 conditional on the previous !onths change in li0uidity and li0uidity level3 earn negative abnor!al returns3 precisely the opposite of what we !ight e/pect #hese results !ust be interpreted with caution Sorting on the sensitivity of changes in individual stoc% li0uidity to changes in aggregate li0uidity is also a sort on !ar%et capitali6ation .or e/a!ple3 the ratio of the average !ar%et capitali6ation of 25
decile 1 to decile 1B for the !)( !easure reported in the first si/ colu!ns of #able 1B is &2B and !ar%et capitali6ation decreases nearly !onotonically between decile 1 and decile 1B #he si6e relative for the RNS !easure is !uch larger 9"KJB and !ar%et capitali6ation is also !onotonically decreasing across deciles1B $ Characteristic i0uidity #o investigate the role of the level of characteristic li0uidity3 each !onth ( sort all stoc%s with available data into ten portfolios by the average of their characteristic li0uidity over !onths t-2 to t-& ( s%ip !onth t-1 to avoid issues associated with bidS 9relative effective spread !easures are used as a !easure of li0uidity *gain3 we !ust interpret the results with caution since sorting on characteristic li0uidity is si!ilar to a sort on !ar%et capitali6ation with the !ost illi0uid stoc%s in decile 1 also being the s!allest stoc%s #able 1B provides wea% evidence that !ore li0uid stoc%s have higher ris% adjusted returns than illi0uid stoc%s *lthough this result contradicts *!ihud and Mendelson 91LK3 it is consistent with >leswarapu and Reinganu! 91LL" (n particular3 >leswarapu and Reinganu! find stoc%s that are particularly illi0uid as !easured by relative 0uoted bid derived brea%points3 then within each si6e 0uintile ( sort into 0uintiles by li0uidity !easure3 either li0uidity return beta3 li0uidity spread beta3 or characteristic li0uidity3 to for! 2+ portfolios #hese portfolios are lin%ed through ti!e and regressed on the .a!a-.rench ris% factors and3 as before3 the difference in alphas is e/a!ined for evidence of abnor!al return associated with li0uidity while controlling for !ar%et capitali6ation ( also e/a!ine the relationship between li0uidity return betas and both li0uidity spread betas and characteristic li0uidity by first sorting into 0uintiles by either spread beta or characteristic li0uidity and then sorting on li0uidity return beta within each 0uintile (n the interest of brevity3 the results reported in #able 11 are only for e0ual-weighted portfolios and only report the difference in alpha by control 26
variable 0uintile >/a!ining the spread in alphas fro! li0uidity return betas while controlling for !ar%et capitali6ation3 there is no obvious relationship between si6e 0uintile and spread in abnor!al return #he 0uintile with the largest spread in intercepts varies by !easure with the 0uintile of the largest stoc%s having the biggest spread in intercepts 9in non-8anuaries for three of si/ !easures #he !ost surprising result co!es fro! the 8anuary !onths #he negative spread in abnor!al returns is not confined to the s!allest stoc%s3 indeed the biggest negative spread in intercepts is always in one of the biggest three of the five 0uintiles Ehile the classic 8anuary effect is closely related to !ar%et capitali6ation3 the relative underperfor!ance of high li0uidity return beta stoc%s in 8anuary is not li!ited to s!aller fir!s Consistent with #able 1B3 when using the reversal-based li0uidity !easures )R and !)R3 there is no evidence that li0uidity spread betas or characteristic li0uidity is priced after controlling for !ar%et capitali6ation #he price i!pact !easure !)(4s significant negative spread in alphas when sorted by li0uidity spread beta continues when controlling for si6e although the effect is larger for the s!aller 0uintiles #he si!ilar results using the spread based !easure RNS also persists across si6e deciles #he significant positive spread in alphas between a portfolio of stoc%s with low characteristic li0uidity as !easured by !)( and high characteristic li0uidity is largest a!ong the s!allest stoc%s and virtually disappears by the largest 0uintile 2& because there is little variability in the !easure for larger stoc%s #he significant positive spread using the spread-based RNS as a !easure of characteristic li0uidity al!ost disappears after controlling for si6e #o verify that li0uidity return betas are not pro/ies for characteristic li0uidity or li0uidity spread betas3 ( use the latter variables as control variables and sort stoc%s into 0uintiles by the control variable before sorting by li0uidity return beta .or each li0uidity !easure3 sorting first by li0uidity spread beta or characteristic li0uidity then by li0uidity return beta has little effect #he point esti!ates by 0uintile are si!ilar to those of the univariate sort in #able + and #able (n su!!ary3 there is wea% evidence that li0uidity spread beta and characteristic li0uidity are priced3 but the ris% pre!ia do not have the e/pected sign Stoc%s with a high covariance between changes in individual li0uidity and shoc%s to !ar%et-wide li0uidity 9after controlling for the lagged level and lagged change in characteristic li0uidity earn lower ris% adjusted returns than those with a low covariance #here is also wea% evidence that stoc%s which are relatively li0uid earn higher average ris% adjusted returns than stoc%s which are relatively illi0uid in non-8anuary !onths #here is strong evidence that illi0uid stoc%s 9as !easured by characteristic li0uidity do 27
earn larger abnor!al returns in 8anuary3 but the effect is concentrated in the s!allest two si6e 0uintiles Most i!portant3 there does not appear to be any relationship between the higher abnor!al returns earned by high li0uidity return beta stoc%s and either li0uidity spread betas or the stoc%4s characteristic li0uidity Conclusion @u!erous authors have found that illi0uid stoc%s earn higher average returns3 presu!ably to co!pensate for the higher costs of transacting #his paper addresses the related but separate issue of !ar%etwide li0uidity (f the ability to transact a given volu!e with !ini!al price concession varies through ti!e and there e/ist cross sectional differences in a stoc%7s return covariance with !easures of !ar%et-wide li0uidity3 then this covariance should carry with it a higher e/pected return Measures of !ar%et-wide li0uidity designed to capture three related3 but separate di!ensions of li0uidity yield si!ilar results3 na!ely that covariance with !ar%et-wide !easures 2+ of li0uidity carries a ris% pre!iu! #his ris% pre!iu! varies fro! appro/i!ately 2 to +5 per year depending on the !easure #his esti!ate of the li0uidity ris% pre!iu! associated with covariance with !ar%et-wide li0uidity shoc%s is !uch lower than that esti!ated by )astor and Sta!baugh 92BB2 but is still statistically significant .our of the si/ !easures used do not re0uire transaction level data and thus can be constructed over !uch long ti!e spans and in !ar%ets for which transaction level data is unavailable Covariance of return with !ar%et-wide li0uidity does not appear related to the covariance between changes in a stoc%7s characteristic li0uidity and !ar%et-wide li0uidity (n other words3 stoc%s whose price is sensitive to !ar%et-wide li0uidity shoc%s do not necessarily beco!e the!selves !ore illi0uid when !ar%et li0uidity is low #his is consistent with fund !anagers selling !ore li0uid stoc%s to !eet !argin calls or rede!ption re0uests when !ar%et-wide li0uidity is low Several surprising results provide avenues for future research .irst3 there is a strong 8anuary seasonal in the li0uidity ris% pre!ia but not in the li0uidity !easures the!selves *lthough high li0uidity beta stoc%s earn higher ris%-adjusted returns on average3 they earn lower returns in 8anuary #his 8anuary effect in li0uidity is not related to fir! si6e3 rather it e/ists across all si6e 0uintiles Second3 li0uidity spread betas3 or the covariance between changes in a stoc%4s own characteristic li0uidity and shoc%s to !ar%et-wide li0uidity3 is associated with lower average returns (n other words3 stoc%s that beco!e particularly illi0uid when !ar%ets beco!e illi0uid earn below average returns3 a counterintuitive result #he negative average return associated with li0uidity spread betas is particularly surprising because one !ight have e/pected that an individual stoc%4s li0uidity is particularly i!portant when the !ar%et is less li0uid overall (f aggregate 28
li0uidity falls when !ar%et returns are large and negative3 then investors who !ust sell will sell those invest!ents that have the best individual characteristic li0uidity so as to !ini!i6e the transaction costs associated with li0uidity #his reasoning is consistent with the financing of !argin investors by uninfor!ed outside lenders who react to losses by cutting lending 9see Shleifer and ishny 91LLJ and ?iong 91LLL #his would result in a high covariance between the return of stoc%s whose individual li0uidity is high and the aggregate li0uidity state variable Ehy then should investors de!and a 2 pre!iu! for holding these stoc%s Ehy do investors not re0uire a pre!iu! for holding stoc%s whose characteristic li0uidity worsens when aggregate li0uidity falls #hese 0uestions provide an interesting avenue for further research
29
CHAPTER -
30
Chapter :- Ca%e +tu') 2 Intro'uction of Co$pan) profile an' Pro'uct About the wor" in co$pan) 'one b) %tu'ent%
31
CHAPTER - 3
32
Chapter 3:- Data Anal)%i%
33
CHAPTER - 4
34
Chapter 4:- Fin'in&%
35
.iblio&raph)
36
Anne5ure 1 *!ihud3 Da%ov3 2BB23 (lli0uidity and Stoc% Returns: Cross-Section and #i!e-Series >ffects3 8ournal of .inancial Mar%ets3 +3 "1-+ 2 *!ihud3 Da%ov and ai! Mendelson3 1LK3 *sset )ricing and the $id-*s% Spread3 8ournal of .inancial >cono!ics3 1J 22"-2&L " *ndrews3 , E =3 1LL13 ]eteros%edasticity and *utocorrelation Consistent Covariance Matri/ >sti!ation3^ >cono!etrica3 +L3 K1J-K+K & $a%er3 Malcol! and 8ere!y C Stein3 2BB23 Mar%et i0uidity as a Senti!ent (ndicator3 arvard (nstitute of >cono!ic Research ,iscussion )aper @u!ber 1LJJ + $rennan3 Michael 8 and *vanidhar Subrah!anya!3 1LL3 Mar%et Microstructure and *sset )ricing: On the Co!pensation for (lli0uidity in Stoc% Returns3 8ournal of .inancial >cono!ics3 &13 &&1-&& Ca!pbell3 8ohn D3 Sanford 8 ;ross!an and 8iang Eang3 1LL"3 #rading olu!e and Serial Correlation in Stoc% Returns3 #he Nuarterly 8ournal of >cono!ics3 1BK3 LB+L"L J Chordia3 #arun3 Richard Roll3 and *vanidhar Subrah!anya!3 2BBB3 Co!!onality in i0uidity3 8ournal of .inancial >cono!ics3 +3 "-2K K Cochrane3 8ohn 3 2BB13 *sset )ricing3 )rinceton Iniversity )ress3 )rinceton @8 ,atar3 inay #3 @arayan D @ai%3 and Robert Radcliffe3 1LLK3 i0uidity and Stoc% Returns: *n *lternative #est 8ournal of .inancial Mar%ets3 13 2B"-21L L >leswarapu3 en%at R3 and Marc R Reinganu!3 1LL"3 #he Seasonal $ehavior of the i0uidity Ris% )re!iu! in *sset )ricing3 8ournal of .inancial >cono!ics3 "&3 "J"-"K 1B .a!a3 >ugene . and =enneth R .rench3 1LL"3 Co!!on Ris% .actors in the Returns on Stoc%s and $onds3 8ournal of .inancial >cono!ics3 ""3 "-+ 11 .a!a3 >ugene . and 8a!es , Mac$eth3 1LJ"3 Ris%3 Return3 and >0uilibriu!: >!pirical #ests3 8ournal of )olitical >cono!y3 K13 BJ-"
37
12 .ernande63 .ran% *3 1LLL3 i0uidity Ris%: @ew *pproaches to Measure!ent and Monitoring3 Securities (ndustry *ssociation Eor%ing )aper ;ross!an3 Sanford 8 and Merton Miller3 1LKK3 i0uidity and Mar%et Structure3 #he 8ournal of .inance3 &"3 1J-"" 1" ansen3 )3 1LK23 ]arge Sa!ple )roperties of ;enerali6ed Method of Mo!ents >sti!ators3^ >cono!etrica3 +B3 1B2L-1B+& 2K 1& asbrouc%3 8oel and ,uane 8 Seppi3 2BB1!3 Co!!on .actors in )rices3 Order .lows3 and i0uidity3 8ournal of .inancial >cono!ics3 +L3 "K"-&11 1+ uber!an3 ;ur and ,o!ini%a al%a3 2BB13 Syste!atic i0uidity3 #he 8ournal of .inancial Research3 2&3 11-1JK 1 8ones3 Charles M3 2BB13 * Century of Stoc% Mar%et i0uidity and #rading Costs3 Colu!bia Iniversity Eor%ing )aper =yle3 *lbert S3 1LK+3 Continuous *uctions and (nsider #rading3 >cono!etrica3 +" 1"1+-1""+ 1J owenstein3 Roger3 2BBB3 Ehen ;enius .ailed3 @ew Dor%3 Rando! ouse )astor3 ubos and Robert . Sta!baugh3 2BB23 i0uidity Ris% and >/pected Stoc% Returns3 8ournal of )olitical >cono!y3 forthco!ing Shleifer3 *ndrei and Robert ishny3 1LLJ3 i!its of *rbitrage3 #he 8ournal of .inance3 +23 "+-++ 1K Shan%en3 8ay3 1LLB3 (nterte!poral *sset )ricing: *n >!pirical (nvestigation3 8ournal of >cono!etrics3 &+3 LL-12B 1L Shan%en3 8ay3 1LL23 On the >sti!ation of $eta )ricing Models3 Review of .inancial Studies +3 1-2& ?iong3 Eei3 2BB13 Convergence #rading with Eealth >ffects: *n *!plification Mechanis! in .inancial Mar%ets3 8ournal of .inancial >cono!ics3 2 2&J-2L2
38
