Are valuation multiples of IPOs anchored to industry
multiples?
T.F. Boomgaard Dr. A.S.R. de Ridder This draft: 31-1-2019Abstract
This paper predicts a behavioral bias in the valuation of initial public offerings (“IPOs”). It is hypothesized that underwriters overly rely or “anchor” on industry peer multiples in the initial valuation of IPOs. Hence, IPOs priced at multiples above their industry peers are expected to have been pulled downwards by the anchor in their initial valuation, whilst IPOs priced at multiples below their industry peers are expected to have been pulled upwards in their initial valuation. After the first day of trading, when more market participants enter the market, the link to the reference price weakens. In line with the hypothesis, for a sample of 573 European IPOs from 2001 to 2018, we find IPOs with seemingly high relative valuations to experience higher returns than IPOs with seemingly low relative valuations. Furthermore, as predicted by anchoring, it is observed that the effect on first day returns increases with the size of the relative valuation distance between the IPO and industry peer multiples. The anchoring effect holds for both backward and forward looking valuation multiples and remains robust after controlling for a variety of explanatory factors on IPO pricing.
JEL classification: G24, G32
1. Introduction
This paper focuses on the pre-valuation of initial public offerings (“IPOs”) using comparable firm multiples. It has been well described that IPOs earn high first day returns. Using a sample of US firms, Ibbotson, Sindelar and Ritter (1994) found early evidence that IPOs earn first day returns between 10% and 15% after going public. Similarly, we find an average first day return of 8.5% for 576 IPOs listed on major European stock exchanges. In most literature, this phenomenon is often referred to as IPO underpricing. It is assumed that in an efficient market, the closing price at the end of the first trading day reflects the “fair” value of the IPO firm. Therefore, the difference between the offer price and the closing price at the first trading day, i.e. the first day return, can be regarded as evidence for underpricing.
An implication of IPO underpricing is that issuers leave ‘money on the table’. This represents the wealth transfer from existing shareholders of the issuing firm to the new investors. Leaving ‘money on the table’ implies that the public offering is less financially beneficial for the issuing firm than expected. In other words, the issuing firm and pre-IPO shareholders are shortchanged as they received less money for their offer than they actually deserved. Considering that positive returns are ‘costly’ for issuers and often form a serious ‘hidden cost’ to both the issuing firm and the existing shareholders, the question arises: what explains IPO (under) pricing?
Over the past decades, a broad set of academic literature has been written on factors explaining IPO pricing.1
Although the underpricing phenomenon is an endless puzzle, we highlight the main findings on IPO pricing. Firstly, underpricing can be the result of an intentional strategy of the issuer or underwriter. Brau and Fawcett (2006) state that managers consider underpricing to be relevant for attracting investors. In addition, Rock (1986) shows that underpricing is necessary to avoid declining interest of uninformed investors. Furthermore, Ljungqvist (2006) hypothesizes that underpricing is due to so-called ‘principal agent conflicts’, referring to conflicts of interest between underwriter and the issuer. From an ownership and control perspective, IPOs are underpriced in order to ensure a wide distribution of shares and in order to avoid the likelihood of being externally monitored.2 Moreover, from a legislative perspective, Ibbotson (1975) states that underwriters
deliberately underprice IPOs in order to avoid possible negative consequences (e.g. lawsuits of shareholders after overpricing). However, underpricing might not always be entirely deliberate. Behavioral explanations of underpricing show that investors bid on IPOs irrationally and that underwriters are subject to behavioral biases in the pricing process of IPOs.
This paper adds a behavioral explanation to the enduring puzzle in the world of underpricing. It will be examined whether the valuation of European IPOs is biased towards industry multiples. It is predicted that, in the process of deciding up on a listing price, underwriters and pre-IPO investors overly rely on the valuation of multiples of comparable industry peers. This relates to the so called ‘anchoring effect’, a cognitive bias in which individuals rely too heavily on an initial piece of information offered (the ‘anchor’) when making decisions.
Typically, the valuation of IPOs starts with the identification and selection of comparable industry peer companies. Thereafter, the market values of industry peers are transferred into standardized values relative to their key financials. This process of standardizing results in valuation multiples. In turn, investors will apply these valuation multiples to the financials of the IPO firms to arrive at an estimate of the fair value or the ex-ante estimate of the market value for the first trading day.
The multiples approach results in a strong focus of underwriters and pre-IPO investors on the pricing of industry peers as a reference point in the pre-valuation of IPO firms.Prior to the IPO, the valuation process involves only a limited number of participants. Depending on the pricing mechanism, offer prices are either set on a fixed basis by underwriters or on a tender basis based on investors demand. In both cases, due to a limited number of participants using homogenous valuation inputs and assumptions, a strong assimilation towards the valuation of peer companies before arriving at a suggested price for the new firm is expected.
In time, after more participants have entered the market, the anchoring effect is expected to weaken. It is presumed that the valuation of an IPO is subject to a number of assumptions varying from investor to investor. For example, investors can dispute about the selection of the peer group or the appropriate valuation multiple for a specific IPO. Therefore, the valuation of a company is often referred to as both a science and an art. 3 As the
stock goes public a broader set of investors is expected to use different peer groups and valuation methods, leading to more heterogeneous information sets for the valuation of listed firms.
Similarly to Hundtofte and Torstilla (2017), we examine the anchoring effect by comparing valuation multiples of IPO firms with valuation multiples of industry peers. A distinction can be made between both over- and underpriced IPOs. It is important to notice that, in this case, the terms “overpricing” and “underpricing” refer to the IPO pricing with respect to comparable industry peers (instead of pricing with respect to the fair value of the firm). Overpriced IPOs have higher valuation multiples at the offer compared to industry peers, whilst underpriced IPOs have lower valuation multiples at the offer compared to industry peers.
Depending on the relative valuation of IPOs, the anchoring effect is expected to have either a positive or a negative effect on first day returns. On the one hand, if the IPO firm is priced at a multiple higher than its peers (i.e. overpriced), it is expected that its valuation will (on average) have been influenced downwards from its ‘fair valuation’ by the reference price. This results in positive first day returns, since it is assumed that IPO stock prices will move to the fair value at the first day of trading. The intuition behind this is that investors likely perceive the IPO as relatively ‘expensive’ compared to industry peers. In addition, it is predicted that the further the anchor, the stronger its pull towards the valuation of the industry peers, and therefore it is expected that a larger valuation difference predicts a larger positive first day returns, all else being equal. In line with prior literature, this effect is expected to diminish in line with the log of distance.
On the other hand, if the IPO firm is priced below the median multiple of its industry peers (i.e. underpriced), it is expected that its valuation will (on average) have been influenced upwards from its ‘fair valuation’ by the reference price. In this case, the anchor is expected to have initially pulled the listing price upwards, resulting in below-average first day returns. From an investor’s perspective, the IPO likely seems to be “cheap” at the offer price, compared to the average industry multiple. Furthermore, we expect a larger distance to the anchor is expected to be associated with a larger effect size, and a stronger pull towards the valuation of peer companies. In order to find whether the effect of anchoring on IPO valuation is truly robust, other factors (potentially) predicting IPO pricing need to be included. Therefore, in this paper, we will control for three groups of control variables relating to market-specific, issue-specific and IPO-specific factors explaining first day returns. These control variables will be highlighted later on in the paper.
If the anchoring effect remains robust, after controlling for other variables explaining first day returns, then our expectations can help distinguish among alternate behavioral theories of IPO (under) pricing. For example, those that predict initial underpricing with respect to industry peers to result in higher than average first day returns and those that predict initial overpricing with respect to industry peers to lead to lower than average first day returns.4
Our expectations are in line with previous findings of Purnanandam and Swamithan (2001), who show that overvalued IPOs earn higher returns than undervalued. Similarly, Hundtofte and Torstilla (2017), also find overpriced IPOs to experience higher-than-average abnormal returns, and vice versa. More importantly, they found evidence for the anchoring effect by showing that IPOs valued at a larger distance from the anchor experienced a stronger pull effect than IPOs closer to the anchor. In other words, IPOs with the seemingly highest relative valuation were found to experience highest first day returns.
This paper differs from the aforementioned papers for a number of reasons. Firstly, there will be tested for an additional accounting multiple to determine the IPO value with respect to industry peers. Secondly, industry peers will be matched based up on a broader definition of industries than has been done in previous literature. Thirdly, there will be focused on European stock exchanges covered by the MSCI index. Most of the existing literature on relative IPO pricing focuses on the US market, therefore investigating European stock markets results in new insights.
The structure of the rest of this study is as follows. Section 2 describes and defines the anchoring phenomenon and elaborates on the theory of IPO underpricing. Section 3 tests for anchoring and develops the implications of the model for the predictability of first day returns. Furthermore, the robustness of the anchoring will be tested by controlling for other influences and checking for a variety of multiple pricing methods. Section 4 provides the results and provides a summary and conclusions.
2. Literature review
The following section will discuss the relevance and the persistency of anchoring by providing several real world examples. Thereafter, we will argue why the anchoring effect is expected to occur in the initial valuation of IPOs. This will be demonstrated by presenting earlier papers on the initial valuation of IPOs using industry multiples. Finally, in order to make sure the anchoring effect is truly robust, we will discuss for which factors explaining IPO first day returns we will control. This will lead to our hypothesis and regression coefficient expectations. 2.1 Anchoring
Tversky and Kahneman (1974) refer to anchoring as a ‘belief formation process’ under which one begins at a notable but perhaps irrelevant value, and subsequently adjusts forward to arrive at a final estimate. Typically, it can be observed that the final estimate is adjusted insufficiently from the initial value. As a result, the final estimate is biased towards the initial reference point or ‘the anchor’.
Strack and Mussweiler (1997) illustrated this by asking participants to estimate the age of Mahatma Gandhi. The first group was asked whether Ghandi became older than nine and the second group was asked whether he became older than 140. Although the numbers are clearly irrelevant values for Gandhi’s age, they produced a strong assimilation effect on subsequent estimates. As a result, participants who received the high anchor on average estimated Gandhi to have lived 67 years, whereas participants who received the low anchor on average thought that he was merely 50 years old. Tversky and Kahneman (1997) provide another example of the anchoring bias. Participants were asked to estimate the result of a product within five seconds. The exercise was presented both in an ascending sequence (1 × 2 × . . . × 8) and in a descending sequence (8 × 7 × . . . × 1). It turned out that estimates for the ascending sequence were significantly lower than for the descending sequence. Apparently, participants used the result of calculating the product for the first few numbers (which is lower for the ascending than for the descending sequence) as an anchor, to which their final estimate was then assimilated. Both studies indicate that anchoring leads to a behavioral bias in which an irrelevant reference point carries a disproportionately high weight in the participants’ estimation.
Mussweiler et al. (2016) highlight the relevance and persistency of anchoring. They make three important statements about the behavioral bias. Firstly, they argue that anchoring is an exceptionally robust phenomenon. Probably the most striking demonstration of this stems from Wilson et al. (1996), who tested whether anchoring remained persistent after informing participants about the influences of potential anchors on their estimations. It turned out that, even after providing explicit instructions to correct for anchoring, final estimates remained biased. Hence, participants’ awareness about potential distortions did not mitigate the anchoring effect, proving the anchoring bias is hard to avoid. Secondly, anchoring occurs independently of participants’ expertise. Even accomplished trial judges with an average of more than fifteen years of experience were influenced by irrelevant sentencing demands made by non-experts.5 Judges who considered high charges gave final sentences that were
much higher than judges who considered low charges. Thirdly, Mussweiler et al. (2016) state that anchoring appears to be an omnipresent phenomenon that occurs in a variety of real world settings. Examples can be found in the legal field, psychology, economics and finance. This gives rise to the belief that anchoring is also present in
the IPO pricing process. Before moving to the presence of anchoring in the pre-valuation of IPOs we will first elaborate on anchoring in economics and finance.
2.2 Anchoring effect in economics
From an economic perspective, anchoring induces market participants to reject a sound financial decision, such as buying an undervalued investment or selling an overvalued investment. Using the housing industry as an example, Northcraft and Neale (1987) find that the valuation of real estate is influenced by initial asking prices, even among professional real estate agents. If real estate agents would have behaved rationally, then the irrelevant asking price should not have any impact on the valuation of real estate.
In consumer behavior, it has been suggested that price claims in advertisements influence consumer behavior because they function as anchors in product evaluations (Biswas and Burton, 1993). The same takes place during negotiations, in which buyers tend to anchor on initial offer prices and adjust subsequently from this value to arrive at a final offer. Note that this can also be used as an advantage. Kahneman (1992) points out negotiators commonly have an interest in misleading their counterparty about their reservation prices. Therefore, high offers and low claims can be made in the hope of anchoring the others’ view of one’s true position.
Finally, in the financial world, investors are likely to anchor on a recent high (or low) point for the stock’s value. They do this in the belief that a drop (increase) in price suggests an opportunity to buy (sell) the stock at a discount (premium). Grinblatt and Keloharju (2001) found evidence for this by showing that stocks being at a monthly low induces individual investors to buy, whereas being at a monthly high induces individual investors to sell. Similarly, anchoring on recent highs also occurs in the field of mergers and acquisitions. Baker and Wurgler (2010) find that prior stock price peaks of targets serve as anchor for the estimation of offer prices, even though they are irrelevant in the valuation of a stock.
2.3 Anchoring effect in IPO pricing
All of the previous findings on the anchoring effect have in common that individuals seem to anchor on readily available information when they are asked to estimate a difficult to value or uncertain quantity. In addition, compressions towards initial valuations seems to occur whenever individuals can easily compare against reference prices.
On the one hand, the use of the multiple valuation method results in economically efficient outcomes. Multiples are relevant because they revolve around key financial statistics related to the value of the firm. Besides, the simplicity of multiples makes it easy for underwriters to communicate initial valuations at the offer with pre-IPO shareholders. In addition, Hundtofte and Torstilla (2017) state that multiple valuations are often used for defensive purposes to explain initial valuations at the offer in case of potential litigations regarding the fair value. On the other hand, the multiple valuation method can also lead to inefficient outcomes. Roosenboom (2012) documents that multiples valuations to determine fair value all suffer from a positive bias with respect to the equilibrium market value. Similarly, Hundtofte and Torstilla (2017) find a bias in the valuation of IPOs by showing that initial offer prices are valued too close to the valuation of their industry peer companies. They show this by examining 4645 IPOs listed on the major US stock exchange for the time period of 1980-2016.
Based on the findings of Hundtofte and Torstilla (2017) it is expected that valuations of industry peers also serve as a reference point in the in the valuation of European IPOs. We test this by checking for different multiples, different matching procedures and different markets than researched by Hundtofte and Torstilla (2017). In order to measure the relative valuation of IPOs compared to industry peers we follow the methodology of Kim and Ritter (1997). They were one of the first to examine the use of multiples of comparable firms as benchmarks for valuing IPOs. Two sets of comparable peers were used: recent IPOs in the same industry and comparable firms chosen by a research boutique that specializes in valuing IPOs. After testing for a variety of backward looking multiples, they found these multiples to have only a modest predictive ability on first day returns. Purnanandam and Swaminathan (2002) built on this model to test whether IPOs are overpriced relative to value metrics based on industry peer price multiples. They found that the median IPO is overvalued at the offer by 50% relative to its industry peers.
2.4 Hypothesis and assumptions regarding signs 2.4.1 Expectations regarding anchoring
Similar to Hundtofte and Torstilla (2017), the distance between the IPO multiple at the offer price and the comparable industry peer multiple is expected to explain first day returns. If the IPO firm is priced at a multiple above its peers (positive distance), it is expected that its valuation will have been influenced downwards from its ‘fair valuation’ by the reference price, resulting in a price increase at the first day. On the contrary, if the IPO firm is priced at a multiple lower than its peers (negative distance), it is expected that its valuation will have been influenced downwards from its ‘fair valuation’ by the reference price. This results in a decrease of the price during the first day. In addition, the size of the distance is expected to increase the effect on first day returns.
2.4.2 Factors influencing IPO returns
“Transaction values”, “Pricing mechanisms” and “Price revision”. Finally, IPO-specific factors are captured by “Sales Growth”, “EBITDA margin”, “Age” and the “Retention ratio”.
Market returns
Loughran and Ritter (2000) show that first day returns on IPOs are predictable based upon market movements prior to the issue. They argue that, following a market increase, IPOs that were in the pre-selling period will have above average expected first day returns. Similarly, when there is a market decline, IPOs that come to the market in the next few weeks will have below average expected first day returns. It is assumed that issuers bargain hard over the offer price in a bad state of the economy, whereas they are easy pushovers in bargaining over the offer price in a good state of the economy. Therefore, higher market returns during the filling period of the IPO are expected to result in higher underpricing. We control for market returns by taking the average return of industry peers during the filing period of the IPO.
Sentiment
In addition, Loughran (2002) argues that more favorable investor sentiment plays a role in the increasing valuation of IPOs. Similarly, Ljungqvist Nanda, and Singh (2006) claim that during periods of high market sentiment investors are overly optimistic and willing to pay higher prices. In contrast, during periods of low market sentiment, investors tend to undervalue IPOs. The European Economic Sentiment Indicator (“ESI”) will be used to measure business sentiment. This is a composite indicator made up of five sectoral confidence indicators with different weights for the industrial, services, consumer, construction and trade retail indicators. It is expected that a high sentiment ratio leads to higher first day returns.
Hot issue markets
According to Ritter (2004) IPO activity appears to move in cycles. They state that hot issue periods result in significantly higher first day IPO returns than cold issue periods. In a hot issue market, excessive optimism on the part of investors leads IPO prices to raise. Hot issue periods can be defined as periods in which the average first month performance (or after market performance) of new issues is abnormally high. We measure the average first day return of firms going public for each month between 2000 and 2018. When the average monthly IPO return exceeds the total average IPO return of our sample, then the month is defined as a hot issue period. Hot issue markets will be included as a dummy variable in the regression model. It is assumed that in a hot issue period there is high optimism around the future performance of the issuer so that there is a picture of underpricing.
Transaction value
Pricing mechanisms
IPO issuing mechanisms are expected to affect the efficiency of IPO pricing. A distinction can be made between fixed price issues and book building issues. In fixed price issues the price at which the IPO is offered is known in advance to the investors (as this is set by the listing firm). As a result, demand for the IPO is known only after the closure of the issue. On the contrary, in book building issues prices are set on a tender basis prior to the issue date. Hence, IPO demand can be known during the book building process. This allows issuers to gain private information from potential institutional investors in their estimate of the value of the IPO, which translates to lower price variations at the first day of trading. Cassia et al. (2004) support this by showing that fixed price offerings are effectively more underpriced than book built offers. We control for pricing mechanisms by including a dummy variable for book building procedures.
Price revision
Price revisions relate to book building issues. Before each IPO, underwriters set a price range for the stock of the listing firm. A price revision occurs during the filing period, the period between the announcement date and the offer date of an IPO, and represents the percentage change between the midpoint of the filed price range and the price at which the IPO is eventually traded. Hanley (1993) was the first to find that IPOs of which the offer price is revised upwards during this period experience, on average, much higher first day returns than those of which the offer price is revised downwards. Similar, we expect price revisions to have a positive effect on first day returns as it can be taken as an indicator of market demand for the IPO.
Sales growth and EBITDA margin
Firm-specific characteristics relate to sales growth and the average EBITDA margin in the year prior to the IPO . These financials measure the profitability of the firm and can explain investors’ future expectations about the listing firm. Assuming that historical rates are used to predict future performance, it is expected that IPOs with higher future forecasts experience higher first day returns.
Age and Maturity
Moreover, the maturity of the listing firm is expected to influence IPO valuations. The stage of the life cycle will be measured by the age of the firm and an alternative measure provided by De Angelo and De Angelo (2005). They measure maturity by the earned/contributed capital mix of a firm. This is a logical proxy for the lifecycle stage of a firm, because it measures the extent to which the firm is self-financing or reliant on external capital. Firms with low retained earnings (“RE”) to total equity (“TE”) ratios tend to be in the capital infusion stage, whereas firms with high RE/TE ratios tend to be more mature with ample cumulative profits that allow them to be largely self-financing and makes them good candidates to pay out dividends.
Year and exchange
Table 1 presents an overview of the expected signs of the aforementioned factors on first day returns. ---
Please put table 1 about here --- 3. Data & Methodology
The level of IPO underpricing can be measured with respect to both the fair value of the IPO firm and industry peer companies. This paper explains underpricing with respect to the fair value by investigating the relative valuation of IPOs compared to their industry peers. Both definitions will be explained in the following section. 3.1 Measuring underpricing with respect to fair value
To measure underpricing with respect to the fair value we use the standard method common for IPOs. Equation 1 shows the definition:
𝑈𝑃 =𝑃 − 𝐸
𝐸 (1)
Where: UPi is the level of underpricing, Pi is the offer date close price and Ei the offer price of the relevant IPO.
According to the efficient market hypothesis, stocks always trade at their fair value, making it impossible for investors to beat the market because there are no under- or overvalued securities. Therefore, the difference between the offer price and the closing price at the end of the first trading day is taken as evidence for IPO underpricing.
In addition, as a robustness test, one can correct for market returns by subtracting the average return of industry peers at the offer date of the IPO. Equation 2 shows the definition for the market corrected underpricing (MUPi)
for IPOi:
𝑀𝑈𝑃 =𝑃 − 𝐸
𝐸 −
𝑀 − 𝑀
𝑀 (2)
Where: Pi and Ei are defined as in equation 1, Mi0 is the average share price of the industry peers at the start of
the offer date of IPOi and Mi1 is the average share price of the industry peers at the end of the offer date.
3.2 Measuring underpricing with respect to industry peers
To measure relative underpricing compared to industry peer companies we use the model of to Purnanandam and Swaminathan (2002). For each IPO firm, a price-to-value (“P/V”) ratio is calculated where P represents the value of the IPO firm at the offer and V is the value computed from industry multiples.
Furthermore, EBITDA – i.e. earnings before interest, taxes, depreciation and amortization – is widely regarded as a ‘clean’ proxy of a company’s operational performance, as it does not factor in financing decisions, accounting decisions (to an extent) and tax environments.
Firstly, the EV/EBITDA of the listing firm at the offer date is calculated, as defined in Equation 3:
𝑃𝑖 = 𝐸𝑉
𝐸𝐵𝐼𝑇𝐷𝐴 (3)
Where: Pi is the multiple at which the IPO is priced at the offer, EVi is the enterprise value of the IPO and EBITDAi
represents the latest reported value for EBITDA prior to the IPO.
The enterprise value is calculated as the market capitalization plus debt, minority interests and preferred shares, minus total cash and cash equivalents. In addition, we derive the market capitalization at the end of the first day by multiplying the offer price with the number of shares outstanding at the end of the first day. All values are based on latest available financial reports.
Secondly, the enterprise value based on industry multiples will be computed. For each IPO in our sample we identified industry peers listed on the same stock exchange. This paper applies a different matching approach than previous papers written on the pre-valuation of IPOs using industry multiples. Different from Purnanandam and Swaminathan (2002) and Hundtofte and Torstilla (2017), peer companies will be matched based up on a broader definition of industry groupings. This is expected to result in benchmarks that are larger and, hence, more accurate. Hundtofte and Torstilla (2017) point out that matching based on Fama-French’s 48 industries results in a larger than typical set of peers used by investors during the offering process. In other words, noise could be introduced due to the use of a too narrow industry classification. It is more likely that investors match firms based on broader industry classifications. Therefore, we used a broader definition of industries provided by Standard & Poor’s Capital IQ (“Capital IQ”). As a result, firms are matched based on 28 industry groupings. In addition, in order to rule out country specific risk, peers will be matched based on the exchange on which they are listed. This prevents cases in which, for example, a German IPO is being compared to Italian industry peers (with higher associated country risk levels).
In order to rule out possible effects of outliers we use the median instead of the mean value to construct the EV/EBITDA multiple. Bhojraj and Lee (2001) show that taking the industry median multiple improves valuation accuracies. In addition, Koller et al. (2005)state that the median is less liable to be distorted by outliers than the mean.
After having matched industry peers to IPOs we can derive the median EV/EBITDA multiple of industry peers, as presented in Equation 4:
𝑉𝑖 = 𝐸𝑉
Where: Vi is the value of the IPO based on the median multiple of industry peers, EVij is the enterprise value and
EBITDAij represents the EBITDA value of peer companies matched to IPO i and active in industry j.
The P/V ratio reflects the difference between the multiple of the IPO at the offer date and the multiple computed from industry peer companies. If the IPO has a higher multiple (P) at the offer than the median multiple (V) of industry peers, then the IPO is considered overpriced. On the contrary, if the IPO has a lower price (P) at the offer than the median value (V) of industry peers, then the IPO is considered underpriced. Equation 5 shows the P/V ratio of IPOi: 𝑃 𝑉 = 𝐸𝑉 𝐸𝐵𝐼𝑇𝐷𝐴 𝐸𝑉 𝐸𝐵𝐼𝑇𝐷𝐴 (5)
Where: Pi/Vi represents the level of under- or overpricing of IPO i compared to the median of the valuation of peers matched to IPO i and active in industry j.
In addition, we will verify whether the anchoring effect remains persistent after controlling for forward-looking multiples. Kim and Ritter (1997) point out that investors are mostly interested in future earnings. Therefore, it is expected that forward-looking multiples play an important role in the assessment of IPOs. We will add two forward-looking multiples to the model. The first forward-looking multiple is based on analyst forecasts. Ideally, analyst forecasts are captured at the offer date of the IPO. However, most analysts only start covering IPO firms after they went public. Therefore, in order to increase the number of observations, we included analyst forecasts that were reported up to one month after the issue date of the IPO. 6 The second forward-looking multiple
assumes perfect foresight and takes into account the realized EBITDA in the year following the IPO. Equation 6 and 7 show multiples based on forward looking earnings.
𝑃 = 𝐸𝑉
𝐸𝐵𝐼𝑇𝐷𝐴 (6)
𝑃 = 𝐸𝑉
𝐸𝐵𝐼𝑇𝐷𝐴 (7)
Where Pforward represents the multiple at the offer date of IPO i based on forward-looking analyst forecasts and
Pperfectforesight is the multiple based on forward-looking multiples using perfect foresight. The price-to-value ratio
will be composed in the same way as it has been done for backward-looking multiples. 3.3 Data
Capital IQ provides information about the IPO transaction and gives financial information about the listing firm. We obtained data on IPO firms from 2000 to 2018 on European stock exchanges covered by the MSCI Europe.
Similar to Purnanandam and Swaminathan (2002) we applied the following screening criteria for inclusion of the IPO:
- The IPO should be covered by the Capital IQ database. - The IPO should have positive EBITDA in the prior fiscal year. - The IPO should be a non-financial firm.
- The IPO should issue ordinary shares and should not be a unit-offering, closed-end fund, real estate investment trust (REIT).
- The IPO should have at least three industry peers.
IPOs in the financial sector have been excluded from the sample, as EBITDA is not considered as an appropriate metric to use for the valuation of companies in the financial sector. Banks to a large extent rely on interest income, which is not part of EBITDA. Therefore, comparing EBITDA multiples of financial with non-financial firms can generate misleading results. In addition, we have made sure that IPOs have at least three industry peers. This mitigates the effect of outliers on market corrected returns.
We observed that Capital IQ lacks a notable amount of financial information on the issuing companies. Overall, European nation-wide markets are smaller and less developed compared to U.S. stock markets. As a result, IPO firms that were not covered by Capital IQ have been excluded from the sample.
The initial data set consisted of 1,403 IPOs. After excluding IPOs in the financial sector 1,102 IPOs remained. In line with Purnanandam and Swaminathan (2002) we excluded listing firms with a negative EBITDA. As a result, 471 IPOs dropped from the initial sample. Finally, IPOs that had less than three matching industry peers were excluded from the sample. This resulted in a final sample of 576 IPOs.
The same criteria have been used for to the selection of industry peers. On average, each IPO was matched to ten industry peers. Hence, average returns and median multiples have been constructed out of a sample of more than 10,000 industry peers. This is expected to result in accurate benchmarks.
The appendix shows the number of IPOs per exchange and per year. The majority of IPOs stem from five exchanges.7 Furthermore, it can be observed that most IPOs took place between 2014 and 2018. Two growth
exchanges (the London Stock Exchange AIM Market and the Nordic Growth Market) are included. Firms on these exchanges are expected to be associated with more risk and higher first day returns. However, as this effect is captured by a dummy variable, this will not result in biased outcomes.
3.4 Descriptive statistics
Table 3 presents the summary statistics for the variables used in the regression analysis. We find a mean transaction value of EUR 227 million, which implies most of the IPOs are small cap firms. In addition, one can observe a large deviation in the transaction value of IPO firm. We will control for this factor in the regression analysis. Mean underpricing proves to be 8%. Moreover, the level of underpricing ranges from minus 87% to plus 163% at the first day of trading. After correcting for industry returns, underpricing ranged between minus 88% to plus 161%. The correction for industry returns is not trivial, because industry returns on the day of the IPO spread from minus 7% to plus 26%. On average, however, the correction is small as the mean for the industry return corrected underpricing is still large. Similar to the findings of Purnanandam and Swaminathan (2002), the average IPO appears to be overpriced at the offer. It can be observed that the mean multiple at the offer equals 23.6x, whilst the average valuation multiple of peer multiples is found to be 12.2x. As a result, the P/V ratio has a mean value of 2.0 indicating that the average IPO is priced at a higher multiple than the median multiple of peer companies.
--- Please put table 2 about here --- 3.5 Test for normality
Before conducting regression analysis, the assumptions of linear regression need to be tested. We test for a normal distribution on the level of underpricing with the Shapiro-Wilk and Skewness/Kurtosis tests. These tests reject the null hypothesis of a normal distribution. Therefore, the non-parametric Wilcoxon signed-rank test is added to the test. The observed distribution of first day returns is skewed to the right. This indicates positive average IPO returns, which is in line with classical theories presuming that IPOs are underpriced. In addition, IPO returns seem to be centered around 6 and 7 percent. We also test whether the average level of valuation is higher for underpriced IPOs than for the point estimate of overpriced IPOs and whether it is different from the point estimate of the highest annual underpricing. For the non-parametric Wilcoxon signed-rank test we compare the median first day return of underpriced IPOs to the median percentage return for overpriced European IPOs.
3.6 Determinants of IPO underpricing
To find possible relationships between IPO underpricing and the independent variables, we conduct a regression analysis. Equation 8 shows the formula for the regression analysis (while omitting year dummies and exchange dummies):
Where Ri is the IPO return and Rm is the median industry return at the end of the offer date. The βs are equal to
an unknown parameter and measure the effects of the independent variable belonging to project i on that project’s underpricing. The error term is εi. Furthermore, ln ( ) represents the distance, if offer prices are anchored to industry averages then the β is positively defined. The regression will be conducted for both backward- and forward-looking P/V ratios on EV/EBITDA multiples. In addition, the price-to-value ratio is transformed into a log formula as it measures relative underpricing. In line with Hundtofte and Torstilla (2017), we do not expect this effect to be linear, but to diminish in line with the log of distance. Constants have been added to the log values in order to make sure that log transformations did not generate missing values for first-day losses.
We have performed a number of tests to control for potential biases in the model. Firstly, we controlled for the dummy variable trap. Two dummies appeared to be highly correlated. Therefore, these dummies were excluded from the model. Secondly, we checked for the correlations between independent variables. It was found that even the highest correlation between independent variables did not show a concern for multicollinearity. Thirdly, we performed the White heteroscedasticity test. This test did not reject the null hypothesis for homoscedasticity. Therefore, heteroscedasticity appears to be no concern. We added robust standard errors to the model, as this is a standard practice in empirical finance research. Finally, outliers have been identified by using the Cooks D approach. Data points have been deleted if they had a Cooks D value higher than 4/n, where n is the number of observations.
4. Results
4.1 Regression results
Table 3 shows regression results for both backward- and forward-looking multiples. It can be observed that distance has a positive effect on first day returns. As a robustness check, we tested the anchoring effect on both the level of underpricing at the first day and the level of underpricing corrected for industry returns. In addition, backward looking multiples appear to be more significant than forward-looking multiples. This could be due to the fact that investors have better access to backward-looking multiples. There is a smaller number of observations for analyst forecasts. Only 325 observations for analyst forecasts have been included in the model, as most of the IPOs only did not have analyst coverage yet. Furthermore, we observe a low predictive power of the model. Therefore, dummy variables for both exchanges and years have been added to the test. It is shown that impact of backward-looking multiples on IPO pricing is most significant in comparison to the forward-looking multiples. These multiples are readily available for investors and therefore might have more explanatory power.
In table 4 we have controlled for market specific factors explaining first day returns. It can be observed that the anchoring bias remains significant after controlling for all three variables. Different than expected, we find industry returns during the filling period of IPOs to have a negative impact on first day IPO returns. This effect increases after adding more control variables. In addition, the sentiment index did not have a significant impact on first day returns. A reason could be that the sentiment index is measured as a weighted index of European sentiment, which might not accurately reflect market sentiment per country or exchange. Hot markets go along with higher first day returns. After adding this variable, the predictive power of the model increases significantly.
--- Please put table 4 about here ---
In addition, we have controlled for issue specific factors. As expected, the issue size of the IPO firm has a negative effect on the excess return of the IPO. However, this effect seems to be rather small and not significant. Besides, there has been tested for the influence of pricing mechanisms on first day returns. The book building procedure is found to have a positive influence of excess returns. This differs from our expectations, as book building procedures were expected to lead to lower price variations at the first day of trading. Finally, it is observed that price revisions have a negative effect on first day returns. However, this effect proves to have little explanatory power. Price revisions only relate to IPOs using the book building procedure. Therefore, we find a lower number of observations. The anchoring effect remains robust after controlling for influences of issue specific factors. ---
Please put table 5 about here ---
Finally, we tested for different IPO specific factors. It can be observed that sales growth levels prior to the IPO have no significant influence on IPO returns. On the contrary, the level of profitability measured by the EBITDA margin does have a significant effect on IPO pricing. The higher the level of profitability prior to the IPO the higher first day returns. This could indicate that investors have high expectations about future cash flows. Younger firms are most likely being perceived as more risky, whereas older firms are associated with lower risk levels. Therefore, we find age to have a negative relationship with first day returns.
The final table shows results on IPO pricing for a variety of factors. For backward looking multiples the anchoring effect remains significant after testing for market specific, issue specific and firm specific factors explaining IPO returns. Furthermore, this is holds for checking first day returns, excess returns and first week returns. Moreover, there has been controlled for forward looking multiples. Both perfect foresight looking multiples and average analyst forecasts do not have a significant influence on IPO pricing. Besides, it can also be observed that hot issue markets, industry returns during the filling period of IPOs and the age of the IPO explain first day returns. Investors can benefit from the knowledge generated by this paper as the control variables partly explain first day IPO returns.
--- Please put table 7 about here --- 5. Conclusion
This paper adds a new explanation to the enduring puzzle of IPO (under) pricing. Using a sample of 576 European IPOs, it is found that initial valuations of IPOs are biased towards the valuation of industry peers. It is demonstrated that IPOs are priced too close to industry average multiples and that the first trading day corrects for this. As a result, IPO firms priced at multiples above industry multiples experience higher-than-average returns and IPOs priced at multiples below industry multiples experience lower than average returns.
References
Ibbotson, Roger G., Jody L. Sindelar, and Jay R. Ritter. "The market's problems with the pricing of initial public offerings." Journal of applied corporate finance 7.1 (1994): 66-74.
Ibbotson, Roger G., and Jay R. Ritter. "Initial public offerings." Handbooks in operations research and management science 9 (1995): 993-1016.
Kim, Moonchul, and Jay R. Ritter. "Valuing ipos." Journal of financial economics 53.3 (1999): 409-437.
Brau, James C., and Stanley E. Fawcett. "Initial public offerings: An analysis of theory and practice." The Journal of Finance 61.1 (2006): 399-436.
Rock, Kevin. "Why new issues are underpriced." Journal of financial economics 15.1-2 (1986): 187-212.
Ljungqvist, Alexander, Vikram Nanda, and Rajdeep Singh. "Hot markets, investor sentiment, and IPO pricing." The Journal of Business 79.4 (2006): 1667-1702.
Link, Albert N., and Michael B. Boger. The art and science of business valuation. Greenwood Publishing Group, 1999.
Hundtofte, C., and Sami Torstila. "Relative Price≠ Relative Value: Anchoring to Industry Multiples in IPO Pricing." (2017).
Swaminathan, Bhaskaran, and Amiyatosh K. Purnanandam. "Are Ipos Really Underpriced?." Review of Financial Studies 17 (2004): 811-848.
Tversky, Amos, and Daniel Kahneman. "Judgment under uncertainty: Heuristics and biases." science 185.4157 (1974): 1124-1131.
Strack, Fritz, and Thomas Mussweiler. "Explaining the enigmatic anchoring effect: mechanisms of selective accessibility." Journal of personality and social psychology 73.3 (1997): 437.
Thaler, Richard H., et al. "The effect of myopia and loss aversion on risk taking: An experimental test." The Quarterly Journal of Economics 112.2 (1997): 647-661.
Mussweiler, Thomas, and Birte Englich. "Subliminal anchoring: Judgmental consequences and underlying mechanisms." Organizational Behavior and Human Decision Processes 98.2 (2005): 133-143.
Englich, Birte, Thomas Mussweiler, and Fritz Strack. "Playing dice with criminal sentences: The influence of irrelevant anchors on experts’ judicial decision making." Personality and Social Psychology Bulletin 32.2 (2006): 188-200.
Wilson, Timothy D., et al. "A new look at anchoring effects: basic anchoring and its antecedents." Journal of Experimental Psychology: General 125.4 (1996): 387.
Englich, Mussweiler, and T. Mussweiler. "Anchoring effect." Cognitive Illusions: Intriguing Phenomena in Judgement, Thinking and Memory (2016): 223.
Northcraft, Gregory B., and Margaret A. Neale. "Experts, amateurs, and real estate: An anchoring-and-adjustment perspective on property pricing decisions." Organizational behavior and human decision processes 39.1 (1987): 84-97.
Kahneman, Daniel. "Reference points, anchors, norms, and mixed feelings." Organizational behavior and human decision processes 51.2 (1992): 296-312.
Grinblatt, Mark, and Matti Keloharju. "What makes investors trade?." The Journal of Finance 56.2 (2001): 589-616.
Baker, Malcolm, Xin Pan, and Jeffrey Wurgler. "The effect of reference point prices on mergers and acquisitions." Journal of Financial Economics 106.1 (2012): 49-71.
Roosenboom, Peter. "Valuing and pricing IPOs." Journal of Banking & Finance 36.6 (2012): 1653-1664.
Loughran, Tim, and Jay R. Ritter. "Why don’t issuers get upset about leaving money on the table in IPOs?." The Review of Financial Studies 15.2 (2002): 413-444.
Ritter, Jay R., and Ivo Welch. "A review of IPO activity, pricing, and allocations." The journal of Finance 57.4 (2002): 1795-1828.
Ljungqvist, Alexander, Vikram Nanda, and Rajdeep Singh. "Hot markets, investor sentiment, and IPO pricing." The Journal of Business 79.4 (2006): 1667-1702.
Loughran, Tim, and Jay Ritter. "Why has IPO underpricing changed over time?." Financial management (2004): 5-37.
Megginson, William L., and Kathleen A. Weiss. "Venture capitalist certification in initial public offerings." The Journal of Finance 46.3 (1991): 879-903.
Cassia, Lucio, et al. "IPO underpricing in Italy." Applied Financial Economics 14.3 (2004): 179-194.
Hanley, Kathleen Weiss. "The underpricing of initial public offerings and the partial adjustment phenomenon." Journal of financial economics 34.2 (1993): 231-250.
Loughran, Tim, and Jay Ritter. "Why has IPO underpricing changed over time?." Financial management (2004): 5-37
Table 1: Description of explanatory factors on first day returns
Variable Description Expected sign
Distance The distance between the IPO multiple at the offer and the median
multiple of industry peers +
Market specific controls
Market return filling period Average return of industry (peer companies) during IPO filling period + Sentiment Composite indicator made up of five sectoral confidence indicators
measuring European business sentiment +
Hot issue market Dummy value of 1 if the average monthly IPO return exceeds the total
average IPO return +
Issue specific controls
Transaction value Total number of shares offered x offer price -
Pricing mechanism Dummy value of 1 if the pricing mechanism concerns the book building
procedure -
Price revision Percentage of change between midpoint of initial price range and the
final offer price +
Firm specific controls
Sales growth Prior fiscal year sales growth as a measure of profitability +
EBITDA margin Prior fiscal year EBITDA margin as a measure of profitability +
Age Age of the firm at the IPO offer date -
Retention ratio Retained earnings over total equity -
Category dummies
Exchange Exchange on which the IPO is listed +/-
Table 2: Summary statistics
Variable Observations Mean Standard deviation Minimum Maximum
First day return in % 576 7.95 15.84 -87.23 163.47
First day IPO industry return in % 573 0.06 1.81 -7.41 26.43
Excess return in % 573 7.90 15.86 -87.91 161.03
Week IPO return in % 576 9.67 19.82 -87.40 159.76
EV/EBITDA (a) 576 23.62 30.60 0.37 247.01
EV/EBITDA industry (b) 576 12.21 6.87 4.84 111.91
EV/EBITDA ratio (a/b) 576 2.04 2.62 0.02 21.87
Perfect foresight EV/EBITDA ratio 523 2.46 13.34 0.02 269.04
Analyst forecast EV/EBITDA ratio 328 1.18 0.87 0.20 7.66
“Distance” ln (a/b) 576 0.33 0.81 -4.01 3.09
Perfect foresight Distance 523 0.15 0.85 -3.98 5.60
Analyst forecast Distance 328 -0.01 0.57 -1.62 2.04
Market specific controls
Market return filling period 565 1.17 11.82 -78.05 231.01
Sentiment index 576 106.64 5.57 78.20 115.30
Hot markets 576 0.69 0.46 0.00 1.00
Issue specific controls
Transaction value in € mln 570 227.09 526.05 0.50 7023.96
Pricing mechanism 574 0.86 0.35 0.00 1.00
Price revisions in % 339 7.02 32.99 -98.96 135.29
Firm specific controls
Sales growth in % 185 43.32 198.62 -100.00 2507.37
EBITDA margin in % 571 62.28 16.00 -29.59 92.16
Age 507 34.64 40.53 0.00 259.00
Table 3: Valuation distance and first day returns
The dependent variable is the level of underpricing (1) or the industry corrected level of underpricing (columns 2-4). Underpricing has been measured for both backward and forward-looking multiples. Robust standard errors are shown between parentheses. Significance of 1%, 5% and 10% is indicated by ***, ** and * respectively.
(1) (2) (3) (4)
Distance 2.56*** 2.45**
(0.97) (0.98)
Perfect foresight Distance 2.44**
(1.16)
Analyst forecast Distance 3.24*
(1.40)
Constant 7.10*** 7.09*** 7.23*** 7.01***
(0.63) (0.64) (0.62) (0.80)
Observations 574 571 518 323
Table 4: Valuation distance and first day returns controlling for market specific factors
The dependent variable is the level of underpricing corrected for industry returns (columns 1-5). There has been controlled for the effect of both years and exchanges on first day returns. Robust standard errors are shown between parentheses. Significance of 1%, 5% and 10% is indicated by ***, ** and * respectively.
(1) (2) (3) (4) (5)
Distance 2.34** 2.39** 2.34** 1.98** 2.06*
(1.01) (1.01) (1.02) (0.85) (0.84)
Market return filling period -0.46 -0.73**
(-0.38) (0.28)
Sentiment index 0.00
(0.24)
Hot markets 17.47*** 17.66***
(1.28) (1.28)
Controls Yes Yes Yes Yes Yes
Constant -5.54 -6.96 -6.10 -0.86 8.47***
(8.05) (7.79) (28.46) (5.35) (23.49)
Observations 571 571 571 571 571
Table 5: Valuation distance and first day returns controlling for issue specific factors
The dependent variable is the level of underpricing corrected for industry returns (columns 1-5). Robust standard errors are shown between parentheses. Significance of 1%, 5% and 10% is indicated by ***, ** and * respectively.
(1) (2) (3) (4) (5) Distance 2.34** 2.28** 2.46** 1.90* 1.95* (1.02) (1.02) (1.02) (1.08) (1.08) Transaction Value -0.00 -0.00 (0.00) (0.00) Pricing mechanism 3.28* 3.58 (1.67) (2.48) Price revision -0.01 -0.00 (0.02) (0.02)
Controls Yes Yes Yes Yes Yes
Constant -5.54 -2.92 -9.02 14.18 11.92
(8.05) (8.43) (8.24) (12.81) (12.24)
Observations 571 565 571 335 332
Table 6: Valuation distance and first day returns controlling for firm specific factors
The dependent variable is the level of underpricing corrected for industry returns (columns 1-5). There has been controlled for the effect of both years and exchanges on first day returns. Robust standard errors are shown between parentheses. Significance of 1%, 5% and 10% is indicated by ***, ** and * respectively.
(1) (2) (3) (4) (5) (6) Distance 2.34** 1.44 2.25** 2.32** 2.61** 2.35** (1.02) (1.66) (1.03) (1.08) (1.09) (1.14) Sales growth 0.00 (0.00) EBITDA margin 0.00*** -0.01 (0.00) (0.04) Age -0.03** -0.04** (0.01) (0.02) Retention ratio 0.02 -0.02 (0.26) (0.24)
Controls Yes Yes Yes Yes Yes Yes
Constant -5.54 -12.49 -10.17 -8.60 -10.96 -18.68***
(8.05) (12.45) (8.11) (8.57) (9.61) (6.39)
Observations 571 182 566 502 503 445
Table 7: Anchoring effect over time for a variety of control variables
The dependent variable is the level of underpricing at the first day (1) at the first week (2) or the corrected level of underpricing for industry returns at the first day of trading (column 3-5). Robust standard errors are shown between parentheses. Significance of 1%, 5% and 10% is indicated by ***, ** and * respectively.
(1) (2) (3) (4) (5) Distance 2.15** 2.15** 2.31* (0.97) (0.97) (1.35) Industry return 0.21 -0.79* -0.11 -0.39 -1.28*** (0.48) (0.48) (0.51) (0.50) (0.40) Sentiment index 0.29 0.29 0.43 0.30 0.19 (0.25) (0.25) (0.51) (0.26) (0.29) Hot markets 17.01*** 17.01*** 18.14*** 15.65*** 15.21*** (1.36) (1.36) (1.63) (1.41) (1.96) Transaction value -0.00 -0.00 -0.00 -0.00 -0.00 (0.00) (0.00) (0.00) (0.00) (0.00) Pricing mechanism 1.64 1.64 2.06 1.63 -0.23 (1.74) (1.74) (2.43) (1.76) (2.19) EBITDA margin -0.03 -0.03 -0.03 -0.02 0.01 (0.03) (0.03) (0.05) (0.03) (0.04) Age -0.03** -0.03** -0.05*** -0.04** -0.02* (0.01) (0.01) (0.01) (0.01) (0.01) Retention ratio 0.17 0.17 0.06 0.21 0.21 (0.19) (0.19) (0.35) (0.17) (0.19)
Perfect foresight Distance 1.47
(1.06)
Analyst forecast Distance 1.98
(1.31)
Controls Yes Yes Yes Yes Yes
Constant -43.59 -43.59 -77.89 -33.00 -5.867
(28.43) (28.43) (57.97) (26.39) (32.08)
Observations 442 442 442 397 250
Appendix
Graph 1: Stock Exchanges
Euronext Paris 88 15%
London Stock Exchange AIM Market 81 14%
London Stock Exchange 77 13%
Borsa Italiana 75 13%
Deutche Boerse AG 46 8%
XETRA Trading Platform 28 5%
Oslo Borse 18 3%
OMX Nordic Exchange Helsinki 16 3%
SIBE - Spanish Stock Exchange Interconnection System 12 2%
SIX Swiss Exchange 11 2%
OMX Nordic Exchange Copenhagen 8 1%
Euronext Amsterdam 7 1%
Nordic Growth Market 6 1%
Wiener Börse AG 5 1%
Euronext Brussels 4 1%
