Introduction Pseudo market & Event-time performance Market timing and long-run underperformance Conclusions

Pseudo Market Timing and the Long-RunUnderperformance of IPOs

by: Paul SchultzJournal of Finance, 2003

Presented by: Hector D. Torres-Aponte

Ph.D. StudentGraduate School of Business

University of Puerto Rico – Rio Piedras Campus

March 16, 2014

Introduction Pseudo market & Event-time performance Market timing and long-run underperformance Conclusions

Purpose & Contribution

The paper shows that under-performance is very likely to beobserved ex-post in an efficient market.

Based on 1973 through 1997 data reveal that when ex-anteexpected abnormal return are zero, median ex-postunder-performance for equity issuers will be significantlynegative in event time.

Pseudo market timing hypothesis is that the more firms canreceive for their equity, the more likely they are to issue stockseven if the market is efficient and managers have no timingability.

Introduction Pseudo market & Event-time performance Market timing and long-run underperformance Conclusions

Reasons for issue equity

There exists many possible reasons for issue equity:

1 Higher prices imply more investment opportunities and firm gopublic to take projects.

2 Firm could issue equity when prices are higher because theybelieve it results in less earnings dilution, because theyincorrectly believe stock prices are too high

These reasons are important?: NO!, What is important is thatmanagers in effect use trigger prices to determine when to issueequity.

Introduction Pseudo market & Event-time performance Market timing and long-run underperformance Conclusions

An example of Pseudo Market Timing

Consider a one-period returns and assume that the marketearns a return of zero and the aftermarket return of the IPO isequal to the market return plus and excess return of either+10% or -10%.

Assume that the price for recents IPO are $100 at time 0 andno companies go public is stock price for potencial IPOs are$95 or less, there is one IPO if prices are between $95 and$105 and three IPOs if prices exceed $105.

There exists two possible IPO excess returns for each period,there are 4 equally likely possible paths offerings and excessreturns.

Introduction Pseudo market & Event-time performance Market timing and long-run underperformance Conclusions

Results of Table I

The table I reveals that when average aftermarket excessreturns are calculated in event-time, that is weighting eachIPO return equally, mean excess returns are positive for thefirst path and negative for other three paths.

Introduction Pseudo market & Event-time performance Market timing and long-run underperformance Conclusions

Expected long run abnormal returns

It is difficult to calculate the effect of pseudo market timingon the likelihood of observing negative abnormal returnsfollowing IPOs.

Ritter (1991), Spiess and Affleck-Graves (1995) estimateaverage long run cumulative abnormal returns around equityofferings as

CAR =E∑

e=1

∑Nj=1 (rj ,e − rm,e)

N

where N is the total number of IPOs or SEOs, E is thenumber of event months, rj ,e is the return of stock j for eventmonth e, and rm,e is the return of the market for the eventmonth e.

Introduction Pseudo market & Event-time performance Market timing and long-run underperformance Conclusions

Expected long run abnormal returns

We know that E (XY ) = E (X )E (Y ) + Cov (X ,Y ). We canexpress the equation for the expected cumulative abnormal returnas

E(CAR

)= E

(1

N

)E

(E∑

e=1

N∑n=1

(rj ,e − rm,s)

)

+ Cov

((1

N

)( E∑e=1

N∑n=1

(re,j − rm,s)

))

Introduction Pseudo market & Event-time performance Market timing and long-run underperformance Conclusions

Data - IPO & SEO

The data of IPO & SEO are was obtained from Security DataCorporation (SDC) from 1973 through 1997.

The study exclude offerings by funds, investments companies,REITs, utilities and banks.

Mean number for IPO & SEO are 26.8 and 26.0 respectively.

The number of IPO per month range from 0 to 107.

The number of SEO per month range from 1 to 104.

Introduction Pseudo market & Event-time performance Market timing and long-run underperformance Conclusions

Returns of IPOs & SEOs

The source of this information is CRSP tapes.

The time period is up to 60 months following the offering.

A simplest way to calculate the abnormal performance is foreach month, subtract CRSP value of equal-weighted indexreturn from stock returns.

Introduction Pseudo market & Event-time performance Market timing and long-run underperformance Conclusions

Aftermarket returns for IPOs, SEOs and the Market

Table IV -Panel A presents statistics on returns and abnormalreturns of firms making equity offerings.

For the calendar-time based returns we can find a mean of0.02% for the IPOs relative to CRSP value-weighted indexand -0.12% relative to CRSP equal-weighted index. Neitherexcess return is significantly different from zero.

For SEOs, calendar-time excess returns based on thevalue-weighted index are not significant, while excess returnscalculated with the equal-weighted index are -0.30% permonth with a t-statistics of -2.40.

For the las three rows, mean returns and excess returnscalculated for each event-period month the evidence forunder-performance by IPOs is strong.

Introduction Pseudo market & Event-time performance Market timing and long-run underperformance Conclusions

Aftermarket returns for IPOs, SEOs and the Market

In panel B, correlations between the number of offerings andthe succeeding market returns grow large in absolute value asthe market return is measured over longer time period.

The panel C is analogous, but in this case the correlationsbetween the number of SEOs in a month and the subsequentreturns of the CRSP value-weighted index. The results areweaker than for IPOs, but each correlations is negative.

Introduction Pseudo market & Event-time performance Market timing and long-run underperformance Conclusions

Relation between price levels and number of offerings

In order to obtain a more realistic simulations, the authorestimate the relation between the number of IPOs and SEOsand stock prices between 1973 and 1997 period. He compileand index of recent IPOs and SEOs for the value of potencialIPOs and SEOs respectively. Note that the value of IPO andSEO indices are set to 100 at the beginning of February 1973.

For each month, an average return is calculated for all firmslisted on CRSP that had IPO or SEO in the 60 prior month.The index level at the beginning is multiplied by one plus theaverage return during the month to gen an index level for thebeginning of succeeding month.

In order to conduct a regression, the author calculate amarket index that is set equal to 100 at the begging ofFebruary 1973 and changed by the return of CRSPvalue-weighted index each month.

Introduction Pseudo market & Event-time performance Market timing and long-run underperformance Conclusions

Relation between price levels and number of offerings

In Table V - Panel A, the author shows the result for theregression between the number of IPOs on the level of theIPO index and market index at the beginning of the month.

The results indicate that the number of IPOs increase withthe index of past IPO returns with a IPO index coefficient of0.1533. The coefficient on the market index is -0.0571,suggesting that, holding the level of the IPO index constant,the number of IPOs decrease with the level of the market.

The IPO and market indices are highly correlated, and theregression results can be interpreted to mean that the numberof IPOs is determined more by the portion of returns that arespecific to IPOs that the portion of returns common to themarket as a whole.

Introduction Pseudo market & Event-time performance Market timing and long-run underperformance Conclusions

Simulations of aftermarket performance

The author estimate the distribution of the monthly returnson CRSP value-weighted index using months from February1973 through December 1997.

Also, estimate the relation between returns on IPOs and SEOsby regressing an equal-weighted average return from all IPOsor SEOs from previous 60 months on the CRSPvalue-weighted index return.

The author create 5,000 simulations of sample path for IPOsand SEOs and simulate the returns of the market each monthusing the mean and the standard deviation of the monthlyreturn on the CRSP value-weighted index over 1973 through1997.

Introduction Pseudo market & Event-time performance Market timing and long-run underperformance Conclusions

Simulations of aftermarket performance

The simulations results are reported in Table VI includingcalculations of means and cumulative abnormal returns for avariety of event periods.

As a results the author reports the distribution of simulatedmean cumulative abnormal return across the 5,000simulations and the distribution of cumulative excess returnsin the 36 months prior to an IPO.

As a conclusion of the Panel A we have that excess returnsare positive in periods prior to IPOs even though the ex-anteexcess returns are zero. This implies that the number of IPOsincreasing as the level of the IPO index rises.

Other results is that cumulative abnormal returns followingIPOs decline monotonically with the length of the holdingperiod.

Introduction Pseudo market & Event-time performance Market timing and long-run underperformance Conclusions

Simulations of aftermarket performance

In order the measure of excess return using the long-runperformance the author used a commonly use metric:

BHAR =T∏t=1

(1 + r IPOt

)−

T∏t=1

(1 + rMkt

i

)Many researchers use the measure because they measurereturns earned by investors following a buy-and-hold strategyand because their use avoids the rebalancing bias incumulative abnormal returns.

Using a buy-and-hold measure the author report that morethan 80% of the simulations produce negative buy-and-holdabnormal returns for the periods following IPOs.

Introduction Pseudo market & Event-time performance Market timing and long-run underperformance Conclusions

Pseudo market timing and characteristics of long-rununderperformance

The author discuss some others characteristics of the long-rununderperformance of equity issuers and how they fit with thepseudo market timing hypothesis.

Measures of operating performance are also poor followingequity offerings.

Poor aftermarket performance is observed in other countriesand at other times.

Offerings occur at market peaks.

Excess returns after equity offerings are more significant inevent-time that calendar-time.

Performance is particularly poor following periods of heavyIPO issuance.

Performance is also poor after debt and convertible debtissuance.

Managers do nor exploit underperformance for personal gain.

Introduction Pseudo market & Event-time performance Market timing and long-run underperformance Conclusions

Explanation for the underperformance of equity issuers

teristic o� ssuer performance.

BehavioralExplanation

Inadequate RiskAdjustment

PseudoMarketTiming

Underperformance after o¡eringsx x x

Poor operating performance after o¡eringsxx

Underperformance: Other countries, other times ? ?x

O¡erings cluster at market peaksxx

Performance is worse in event-timexp

Performance is worst after heavy issuancexp

?seussitbedretfaroopsiecnamrofrePx

Managers do not appear to pro�tx

Introduction Pseudo market & Event-time performance Market timing and long-run underperformance Conclusions

Conclusions

The author propose that the poor performance of equityissuing firms in event time is real in the sense that IPOs andSEOs have underperformance relative to their ex-anteexpectation, but this does not indicate a market inefficiency.

Using simulations with parameters estimated from historicaldata the author shows that pseudo market timing can easilylead to a level of ex-post underperformance similar todocumented for IPOs and SEOs over the past 25 years.

The author shows that using a calendar time returns ratherthan event time returns reduce the possibility of bias frompseudo marketing timing.

If event-time returns must be used, the results suggest thatthe problem can be resolve by using benchmarks that arehighly correlated with the firms being studied as possible.Also, he shows that market-model returns are preferable tomarket adjusted return.