PhD Course in Initial Public Oerings

Ernst Maug1

Revised February 24, 2019

1Professor of Corporate Finance, University of Mannheim; Homepage: http://http://cf.bwl.uni-mannheim.de/de/people/maug/, Tel: +49 (621) 181-1952, E-Mail: [email protected]. This note is made publicly available subject to the condition that any user noties the author of its use. Please bring any errors and omissions to the attention of the author. Ernst Maug PhD Course in Corporate Finance The Puzzles

Initial public oerings (IPOs) are challenging for nancial economists and have generated a large literature: IPOs are underpriced. On average, the oering price to investors is much lower than the price that prevails in secondary trading immediately after the IPO. Apparently, issuers leave money on the table. Why are investment bankers not able to get more of the value of shares to the initial owners of the rm? IPOs happen in waves with hot issue markets and periods in which almost no rms go public. Moreover, more rms issue shares in those times when underpricing is largest. IPOs appear to underperform in the long term. If researchers compare the performance of IPO rms over a 3-5 year period after the IPO and match IPO rms with seasoned rms, the investment in IPO rms looks much worse.

Ernst Maug PhD Course in Corporate Finance Disappearing public rms - another puzzle?

Figure: IPOs in the United States. Figure 1 From Doidge, Craig, Kathleen M. Kahle, George Andrew Karolyi, and René M. Stulz, 2018, Eclipse of the Public Corporation or Eclipse of the Public Markets?, ECGI - Finance Working Paper 547/2018.

Ernst Maug PhD Course in Corporate Finance IPO Underpricing

Based on Rock (1986)

A company oers 1 share of its common stock for the rst time in an initial public oering (IPO). Value of company is uncertain:

 V¯ with probability p V = V with probability 1 − p

The share is oered at an initial oering price P0; the price cannot be revised.

Ernst Maug PhD Course in Corporate Finance There are two groups of investors: Informed investors acquire information about the company.

They apply for 0 < QI ≤ 1 shares if and only if they expect to break at least even, otherwise they do not apply for shares. Uninformed investors do not acquire any information and only know the probability distribution for rm values. They apply

for QU ≥ 1 shares if and only if they expect to break at least even, otherwise they do not apply for shares. Note: If uninformed investors and informed investors apply for shares, then demand exceeds supply. Assume also that the number of shares cannot be increased in this case (no  option).

Ernst Maug PhD Course in Corporate Finance We also assume: Investors cannot sell the shares. The information acquired by informed investors is of the form of a binary signal σ ∈ {σ; σ} where:  Pr (σ = σ |V = V ) = Pr σ = σ V = V¯ = 1 − .

Hence  is the error of the signal. Assume for now that  = 0, i. e. informed investors receive a perfect signal. If demand exceeds supply, shares are rationed pro rata, i. e. every investor receives a proportion 1 per share ordered. QU +QI

Ernst Maug PhD Course in Corporate Finance From these assumptions it is easy to derive the decision rule of informed investors:

Apply if P0 ≤ V , Do not apply if P0 > V .

It is clear that the issuer will never issue shares below V , and no   investor will ever buy shares if P0 > V¯. Hence P0 ∈ V , V¯ and the policy of the informed investor can be rewritten as:

Apply if V = V¯, Do not apply if V = V .

Ernst Maug PhD Course in Corporate Finance Then we can write the incentive compatibility constraint for the uninformed investors as follows:

QU  p V¯ − P0 + (1 − p)(V − P0) ≥ 0. QU + QI This implies that the oering price is bounded from above. Dene α ≡ QU . Then: QU +QI

αpV¯ + (1 − p) V P0 ≤ . 1 − (1 − α) p Uninformed investors will not participate if this constraint is violated.

Ernst Maug PhD Course in Corporate Finance This shows immediately that the oer is underpriced. The average price in secondary trading is:

P1 = pV¯ + (1 − p) V .

Then the average IPO discount is:

 α  ∆ = P1 − P0 = pV¯ 1 − IPO 1 − (1 − α) p  1  + (1 − p) V 1 − 1 − (1 − α) p 1 − α = p (1 − p) V¯ − V  . 1 − (1 − α) p

Ernst Maug PhD Course in Corporate Finance Note that: The variance of a Bernoulli random variable like rm value is 2 p (1 − p) V¯ − V  , hence the IPO discount is closely related to the uncertainty or of the share price. The IPO discount is zero only if α = 1 as long as volatility is not zero. Hence, 1 − α is a parameter for the degree of adverse selection. Exercise Generalize the model above for the case where  > 0 and informed investors' information is only imperfect. Derive new expressions for

P0 and ∆IPO .

Ernst Maug PhD Course in Corporate Finance Exercise Extend the model of Rock (1986) to include underwriter price support in the as follows. Assume the rm hires an underwriter who is committed to buy all shares from the rm at the price PF . The underwriter sells the shares in the IPO for P0 and agrees to buy shares back in the secondary market for some support price P. Only investors who bought shares in the IPO can sell them back at the support price.

Ernst Maug PhD Course in Corporate Finance Solve the model using the following steps: Assume that P ∈ V , V¯. Rewrite the condition for the uninformed investors to participate in the IPO. Show that P0 becomes a function of P. Derive this function and show whether it is decreasing or increasing. Write down the payo of the underwriter. This must include the payo from buying shares from the rm and selling them to investors, and the expected cost of price support, i. e. the cost of buying shares above their intrinsic value. (Hint: assume the underwriter buys these shares from investors and immediately sells them in the secondary market for the

intrinsic value P1). From this, what is the price PF the underwriter can oer the rm and still make a positive prot?

Ernst Maug PhD Course in Corporate Finance Assume many banks compete for the issue, so the underwriter makes zero prots in equilibrium and the underwriter who buys the shares from the rm at the highest

price PF conducts the oering. Which level of price support P is oered in equilibrium? Verify the initial assumption that   P ∈ V , V¯ . Hence, what is the equilibrium solution for P0 and PF ? Show that the equilibrium has the following properties: (1) the adverse selection problem is eliminated completely, and the rm receives a fair price of the stock, and (2) the IPO is overpriced. Comment on this solution. Note: Models of IPO price support were published by Chowdhry and Nanda (1996) and Benveniste, BuSaba and Wilhelm (1996).

Ernst Maug PhD Course in Corporate Finance Motivation

Based on Schultz (2008)

Question: Why do IPOs underperform in the long run? Potential explanation: behavioral eects: companies issue equity when the market is overvalued problem: violates market eciency Alternative: Pseudo-market timing: companies issue more equity at higher stock market levels Then IPOs cluster at ex post peaks

Ernst Maug PhD Course in Corporate Finance Idea

The argument in ve steps:

1 More companies are taken public when stock market values are higher than when they are lower.

2 If stock prices decline, then the number of IPOs goes down subsequently.

3 Then the ex post peak of the stock market also becomes the ex post peak of IPO activity.

4 Event studies weight returns by the number of IPOs at the beginning of the period, hence they give more weight to these ex post peaks followed by negative returns.

5 Event study returns are lower than calendar returns.

Ernst Maug PhD Course in Corporate Finance Example

Two periods In each period, the excess return on the stock market is either +10% or -10%. The current level of the IPO price is equal to 100. IPO activity depends on the price level if P<95: no IPOs if 95105: 3 IPOs With two periods and two states there are four equally likely scenarios.

Ernst Maug PhD Course in Corporate Finance Pseudo market timing in a two-period model

Ernst Maug PhD Course in Corporate Finance Two-period model: Calculations

Period 0 Period 1 Total X Return Scen. Price # X Ret Price # X Ret IPOs Cal. Event 1 100 1 0.10 110 3 0.10 4 0.10 0.10 2 100 1 0.10 110 3 -0.10 4 0.00 -0.05 3 100 1 -0.10 90 0 0.10 1 0.00 -0.10 4 100 1 -0.10 90 0 -0.10 1 -0.10 -0.10 Avg. 0.00 0.00 -0.04 Consider scenario 2:

Calendar day return = (XRet0 + XRet1) /2 = (0.10 − 0.10) /2 = 0.00 Event day return = (1 × XRet0 + 3 × XRet2) /4 = (1 × 0.10 + 3 × −0.10) /4 = −0.05.

Ernst Maug PhD Course in Corporate Finance A general restatement

CAARs in long-term event studies are calculated as:

t=T "i=N # X 1 X CAAR = (r − r ) . N i,t m,t t=1 i=1 Here: r = return t = event month (usually runs from 1 to 36 or 60) i = stock of IPO rm (runs from 1 to number of IPOs) m = market or matching stock

Ernst Maug PhD Course in Corporate Finance Why average CAR-estimates are biased

Standard assumption: N is xed. Assumption here: N is a random variable

t=T i=N !  1  X X E (CAAR) = E E (r − r ) N i,t m,t t=1 i=1 t=T i=N ! 1 X X + Cov , (r − r ) . N i,t m,t t=1 i=1 N is positively correlated with returns Covariance in second line is negative (correlation with 1/N) Estimate of CAAR is biased downward

Ernst Maug PhD Course in Corporate Finance How big is this eect?

Need to calibrate magnitude of this eect and compare it to returns from statistical studies Research design:

1 Measure empirical sensitivity of number of IPOs to value of stock market and to value of rms that recently went public 2 Simulate economy with realistic parameters for returns to the market, IPO-portfolio, and issuance activity 3 Calibrate model so that markets in the simulated economy are ecient and no real market timing is possible 4 Measure eect of pseudo market timing using usual buy-and-hold returns (BHAR)

Ernst Maug PhD Course in Corporate Finance Step 1: Measure sensitivity of IPOs to returns

Research design: Construct IPO-index from rms that went public in 60 months prior, calculate return Index value is 100 in February 1973, subsequently index value is IPO-Index(t)=IPO-Index(t-1)*(1+mean return to IPO rms) Regress IPO activity from 1973 to 1997 on IPO-index, stock market index, and time

#IPOs = −1.974 − 0.144t − 0.057Markett + 0.153IPOt . (−11.66) (19.43) The R-squared of this regression is 77.8%.

Ernst Maug PhD Course in Corporate Finance Steps 2-4: Run Monte Carlo simulations

Step 2: Estimate statistics of returns Mean monthly return is 1.12% with standard deviation of 4.52% Regress return to IPO-portfolio on market return: is 1.31 times higher with residual variance of 4.27% Step 3: Calibrate economy and run 5,000 times Generate returns from normal distribution with parameters as estimated in step 2 for market returns For IPO-portfolio, use slope coecient and variance of disturbance term from regression, but use intercept to equate expected IPO returns with expected market returns Generate returns for 300 months (25 years) Step 4: Calculate BHARs, average across 5,000 runs

Ernst Maug PhD Course in Corporate Finance Simulated excess returns and wealth relatives

BHARs

1-12 1-24 1-36 1-60 Median -5.29% -10.97% -17.06% -31.75% Mean -5.26% -10.18% -15.18% -26.50% Std. error 0.10% 0.19% 0.29% 0.55% t-statistic -54.77 -54.35 -52.74 -47.94 Percent<0 81.80% 82.10% 82.70% 82.80%

Wealth relatives

1-36 1-60 Median 85.29% 77.81% Mean 86.43% 80.82% Std. error 0.20% 0.29%

Ernst Maug PhD Course in Corporate Finance Comparison to other theories

Behav. Inad. Psd. market risk market timing adj. timing 1 Underperformance after oerings """ 2 Poor operating perf. after oerings "" 3 Underperf.: Other countries, other times ?? " 4 Oering clustering at market peaks "" 5 Perf. is worse in event-time "" 6 Perf. is worst after heavy issuance "" 7 Perf. is poor after debt issues ? " 8 Managers do not appear to prot "

Ernst Maug PhD Course in Corporate Finance Conclusions

It is better to measure returns in calendar time, not in event time, to account for the endogeneity (dependence on stock market ) of the event itself. Benchmark against index that not only has the same expected return under the null, but is also highly correlated with the event itself. More detailed discussion of econometric issues: Baker, Taliaferro, Wurgler, JF 2006; Viswanathan & Wei, RFS 2008.

Ernst Maug PhD Course in Corporate Finance Motivation

Based on Khanna, Noe and Sonti (2008)

Puzzling observations about IPOs: Hot issue markets: why do many rms suddenly decide to go public? What is the window of opportunity? Underpricing is higher in hot markets than in cold markets. Why do rms not shift to markets where they have to leave less money on the table? Why does competition not eliminate banks' rents in hot markets? Why are rms during hot markets younger, less protable, and with less insider ownership?

Ernst Maug PhD Course in Corporate Finance The Model

Economy has N entrepreneurs who may go public. Each is matched with one of a continuum of underwriters: A fraction ρ of projects is good (G) with payo X = 1, 1 − ρ is bad (B) with payo X = 0. Going public has an opportunity cost w. Each rm bargains with the underwriter it is matched with

who sets the oer price ps . Firms capture a fraction β of the issue price, underwriters 1 − β. Underwriters' benet from higher prices through higher fees; they are penalized for overpricing IPOs. Underwriters hire a quantity η ∈ [0, 1] of bankers who cost θ and screen projects. With probability 1 − η they receive an uninformed signal U. With probability η they become perfectly informed so that Pr (H |1) = Pr (L |0) = η.

Ernst Maug PhD Course in Corporate Finance Timing of Events

Ernst Maug PhD Course in Corporate Finance Solution

1 Average quality of the IPO pool is ρ , assuming π = ρ+α(1−ρ) that all G projects and α of the B projects go public (single crossing). This is the fundamental value of the shares. 2 There is no underpricing for s = L, H. There is underpricing for ∗ . s = U : pU < π 3 The dierence in issue prices between G and B rms is βη.G entrepreneurs issue with probability 1.

Ernst Maug PhD Course in Corporate Finance A sequentially rational equilibrium is a triple (π, η, θ) such that: 1 B rms play a mixed strategy: 1 ∗ . β ( − η) pU = w 2 The market for bankers clears: 1 ρ 1. Nη (ρ + α ( − ρ)) = Nη π ≤ 3 Underwriters hire screening labor (bankers) such that ∆V = θ. 4 Underpricing: ∗ . pU < π

Ernst Maug PhD Course in Corporate Finance Proposition The larger the pool of potential IPOs (N ↑) in an overheated equilibrium, the lower the average quality of rms that want to go public (π ↓).

The higher the average quality of IPOs issued, the higher the benet for B rms to go public. The higher the probability of screening, the lower the benet for B rms to go public. Hence, B's indierence condition implies that a higher π has to be compensated by a higher η. A higher N (or ρ) shifts the indierence curve in π − η−space to the right and equilibrium π down. Fixed supply of bankers reduces π: A higher quality ρ or a larger size N of the IPO pool reduces the quality of IPO applicants.

Ernst Maug PhD Course in Corporate Finance Ernst Maug PhD Course in Corporate Finance Hot and Cold Markets

Proposition If the number of good projects is below β , there exists an ρN β−w equilibrium in which only good projects try to obtain funding. If ρN is above this cut-o, some bad rms apply for funding.

Implications: There is a discontinuous shift at some point such that above the threshold, there are more rms apply for funds. Such a shock is more likely to come from market-wide shocks than from industry-specic IPO waves.

Ernst Maug PhD Course in Corporate Finance Discussion

The story in a nutshell: Hot markets are ignited for neoclassical reasons: The number of good rms that want to go public crosses a critical threshold. Once suciently many investment bankers are too busy screening projects, the quality of screening declines, which opens a window of opportunity for bad rms. Bad rms are drawn into the market, which becomes even more crowded, deteriorating the quality of screening further. The quality of IPOs declines and the uncertainty about quality increases, leading to more underpricing.

Ernst Maug PhD Course in Corporate Finance