Analyst coverage and the relation between underpricing and long run performance of IPO firms

Analyst coverage and the relation between underpricing and long run

performance of IPO firms

Naomi Peters

10013768

Universiteit van Amsterdam

Supervisor: Dr. D. Damsma

02-02-2015

Abstract

This paper hypothesizes that, if there is a relation between IPO underpricing and analyst coverage, and a relation between analyst coverage and long-run performance of issuing firms, this must imply a direct causality between underpricing and firm performance. The first relation is proven to hold for this sample, but with a reversed causality than is usually assumed; IPO firms who show great first day returns attract more analyst following of their stock, and receive more recommendations. Furthermore, results show that firms with greater analyst following outperform in terms of net income per share in the third year of going public, which suggests that the second part of the hypothesis holds too. Since both relations are established in this paper’s research, a direct causality between the level of initial returns and long-run performance is indicated, which is also confirmed by this paper’s research.

Statement of Originality

This document is written by Student Naomi Peters who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

Table of content

1. INTRODUCTION ... 4  

2. LITERATURE ... 5  

2.1INITIAL PUBLIC OFFERING ... 5

2.2UNDERPRICING ... 5

2.2.1 Value of underpricing ... 6

2.2.2 Value of analyst recommendations ... 6

2.2.3 Quality of analyst recommendations ... 6

2.2.4 Long-run value of analyst recommendations: ... 7

3. HYPOTHESIS ... 7  

3.1HYPOTHESIS DEVELOPMENT ... 7

4. DATA AND SAMPLE DESCRIPTION ... 8  

4.1SAMPLE FORMATION ... 8

4.1.1 Thompson One database ... 8

4.1.2 CRSP database ... 9

5.2RELATION BETWEEN UNDERPRICING AND ANALYST COVERAGE ... 14

5.2.2 OLS regression with underpricing as de dependent variable ... 16

5.3RELATION BETWEEN ANALYST COVERAGE AND FIRM PERFORMANCE ... 16

5.3.1 Returns by quintiles ... 17

5.3.2 OLS regression with firm performance as the dependent variable ... 18

5.4RELATION BETWEEN UNDERPRICING AND FIRM PERFORMANCE. ... 20

6. CONCLUSION AND DISCUSSION ... 20  

6.1UNDERPRICING ... 20

6.2FIRM PERFORMANCE ... 21

6.3RELATION BETWEEN UNDERPRICING AND FIRM PERFORMANCE ... 21

1. Introduction

The year 2014 was the year of the most IPO’s since the techbubble, which has once again increased the relevance of the question around IPO anomalies. Three of the most well known anomalies are the underpricing of IPOs, long-run underperformance and hot issue markets (Rajan & Servaes, 1996).

These anomalies, and their underlying factors, have been a subject of debate for many years. Consequently, many theories have been formed around the anomalies relating to initial public offerings. And, while for all three of these anomalies many different explanations have been offered, ‘analyst recommendations’ is a common variable in both studies regarding underpricing as well as in papers that study the long-run performance. Therefore, the hot issue market anomaly will be disregarded in this paper, and the focus will remain on IPO underpricing and how this relates to the long-run performance of IPO firms.

Recent studies imply that analyst coverage has increased in importance for issuing firms (Cliff & Denis, 2004, p. 2871). This can be seen in the study conducted by Rajan and Servaes (1997), who find results that suggest that more underpriced IPOs attract larger analyst following. Similarly, Dunbar (2000) and Clarke et al. (2003) report that underwriters gain market share if they have a highly ranked analyst on the team, which suggests that issuing companies place value on research coverage from (highly ranked) analysts. Moreover, Krigman et al. (2001), find that if issuing firms don’t receive the expected coverage from their underwriter they are likely to switch underwriters for the seasoned equity offer (SEO).

While one may expect that issuing firms can simply compensate underwriters for greater analyst coverage by increasing the underwriter fee, several papers show that the underwriter fee spread is only 7% for all IPOs but the smallest and largest offerings (Chen & Ritter, 2000, Hansen, 2001). Furthermore, Cliff and Denis (2004) found evidence that issuing firms may pay for research coverage via the underpricing of the initial public offering.

I hypothesize that there is a correlation between IPO underpricing and (long-run) firm value. Issuing firms expect analyst recommendations to be of value to the firm, and thus, if these firms pay for analyst coverage through IPO underpricing, this implies a correlation between underpricing and firm value.

Firstly, I will research if there exists a relation between underpricing and analyst recommendations, to see if it may be possible that issuing firms pay for research coverage by leaving money on the table. Secondly, I will investigate whether analyst recommendations truly lead to increased firm value, measured in several ways. And finally, if these two relations hold, I will test whether the relation between underpricing and firm value exists, and if analyst coverage is an explanatory variable.

This paper contributes to existing literature to the extent that, while the effect of analyst recommendations on underpricing and vice versa is studied, and the value of analyst coverage on long-run performance is examined, these hypothesizes have not yet been combined to my knowledge.

It would be interesting to examine how the two anomalies relate, and if analyst recommendations may be an explanatory factor.

This paper is structured as follows: next, the literature surrounding the anomalies will be discussed, beginning with IPO underpricing and its possible explanations, which will be followed by literature concerning the determinants of long-run performance. In the subsequent section, I will describe the sample data, characteristics and research method. Lastly, the empirical results and conclusion are discussed.

2. Literature

In the following section I will first briefly discuss the literature on initial public offerings. Since the IPO anomalies are a much-researched matter, I will then go over the existing literature surrounding IPO underpricing, and the different explanations that have been found. In the last part this paper will take a closer look at the perceived value of analyst recommendations and its quality.

2.1 Initial public offering

The initial public offering is the process of transforming a private company into a public company. One of the reasons to do so is to raise capital from investors, and to create a public market in which founders and shareholders can convert wealth into cash at some future date (Ritter & Welch, 2002, p. 1798). Firms go public by issuing shares with the help of an underwriter: often an investment bank, which provides a wide range of pre- and post-IPO services to the issuing firm (Cliff & Denis, 2004, p. 2871). Together with the underwriter, IPO firms set an offering price and the number of shares to be sold.

2.2 Underpricing

IPO underpricing is measured by the first-day initial return after going public. The level of underpricing is determined by the degree of abnormality in initial return. Since evidence has been found on IPO underpricing (Ibbotson, 1975), many economists have attempted to offer an explanation why investors receive large initial returns, and issuing firms leave money on the table. Some explanations offered are the presence of information asymmetry between underwriters and issuing companies (Baron, 1982), and the use of underpricing as a signaling device (Welch, 1989).

Chemmanur (1993) mentions that investors can directly produce information themselves. Underpricing is used to compensate externals, e.g. analysts for their costs associated with information production. Information production by many skilled externals leads to an intrinsic valuation for the issuing company (Hakenes & Nevries, 2000, p. 4).

While Chemmanur (1993) predicts that underpricing leads to analyst coverage, the more recent study of Cliff and Denis (2004) reverses the causality. They hypothesize that issuing firms pay for analyst coverage via the underpricing of the offering. Their study is based on the earlier

mentioned assumption that analyst coverage has been gaining importance in the security issuance process (Cliff and Denis, 2004).

2.2.1 Value of underpricing

Firstly, we must have an understanding of how underwriters benefit from underpricing, in order to understand why underpricing can be used as an incentive. Analyst coverage in Cliff and Denis’ (2004) study is measured by the number of recommendations made by lead underwriters because, compared to non-lead underwriters or other (unaffiliated) analysts, lead underwriters have most to gain from the underpricing (Cliff & Denis, 2004, p. 2875).

They do so in at least two ways: firstly, they can allocate IPOs that are more underpriced to favored clients hopefully in return for future investment banking business. And secondly, they can distribute shares to hedge funds and other large investors who are then more willing to do more of trading with the investment bank (Cliff & Denis, 2004, p. 2875).

2.2.2 Value of analyst recommendations

Secondly, to understand why issuing firms would deliberately underprice their initial offering shares to purchase analyst coverage, it is important to understand the perceived importance of analyst coverage for the issuing firm. As mentioned above, Cliff and Denis (2004) measured analyst coverage by the number of analyst recommendations made by lead underwriters. To start, they comment that greater analyst coverage might lead to greater investor recognition of the IPO firm, which leads to higher company value according to the Merton model (1987).

Furthermore, Cliff and Denis (2004) state that analyst coverage can generate publicity for issuing firms, which may result in increasing firm value by attracting more customers (p. 2874). They refer to Hakenes and Nevries (2000), who conclude that issuing companies are more likely to underprice their shares if they long for publicity which leads to higher operating figures (measured by turnover figures) and intrinsic value (p. 18). Thus, the issuing firm benefits from analyst coverage, and the underwriter gains from underpricing.

The research executed by Cliff and Denis (2004) confirms the correlation between IPO underpricing and the frequency of subsequent recommendations made by the lead underwriter with a 0.01 significance level. Firms in the lowest quintile of IPO underpricing receive a recommendation from the underwriter in 75% of the cases, while this rate increases to 86% for the companies in the highest quintile of underpricing (Cliff & Denis, 2004, p. 2873).

2.2.3 Quality of analyst recommendations

Next, we can ask ourselves if the quality of the recommendations is of importance for issuing firms. Cliff and Denis (2004) measure the quality by the status of the analyst who made the recommendation. A recommendation is perceived to be valuable if made by an all-star analyst as ranked in the annual

Institutional Investor survey (Cliff & Denis, 2004, p. 2873). This shows in the researches mentioned earlier conducted by Dunbar (2000) and Clarke et al. (2003): underwriters increase their market share of IPOs if they have an all-star analyst on the team.

Cliff and Denis (2004) find a positive correlation between IPO underpricing and all-star coverage: IPO firms in the highest quintile of underpricing receive recommendations made by an all-star analyst in 35% of the cases, while only 16% of the firms in the lowest quintile only receive all-all-star analyst coverage (Cliff & Denis, 2004, p. 2873). Both these results suggest that, rather than increasing the underwriter fee, issuing firms underprice their IPO to compensate investment banks for high-quality analyst coverage (Cliff & Denis, 2004, p. 2873).

2.2.4 Long-run value of analyst recommendations:

The issue that subsequently arises is: what long-run value does an analyst recommendation add? Do analyst recommendations have long-run implications for IPO firms?

A study conducted by Bradley et al. (2008) attempted to answer this question by focusing on the effect of analyst coverage on the long-run performance. They found no conclusive evidence that incremental analyst coverage is related to improved long-run performance (p. 1121). Furthermore, whether an analyst is an ‘all-star’, i.e., is reputable and high ranked, adds no value for the issuing firm in the long run (Bradley et al. 2008, p. 1121). Finally, contrary to some beliefs that affiliated analysts’ coverage might be biased due to conflict of interest and thus leads to worse performance, Bradley et al. (2008) find no difference in long-run performance between affiliated and unaffiliated coverage (p. 1122).

Conflicting with this finding is the earlier study by Das et al. (2006) who did find a positive correlation between residual analyst coverage and post-coverage stock returns (p. 1161). The higher the residual coverage, which they define as the unexpected number of analysts that choose to follow a firm, relative to that predicted by known determinants (p. 1161), the higher is the future stock performance. They attribute this result to the predictive power of analysts regarding future performance of issuing firms.

This difference in results may be due to two variables: firstly, Das et al. use analysts’ earnings forecasts as a variable, while Bradley et al. use analyst recommendations. With regards to the IPO underpricing, analyst recommendations made by lead analysts is used as an explanatory variable. Since I want to investigate whether IPO underpricing is related to long-run performance, I will follow the view of Bradley et al, who also use analyst recommendations as a variable.

3. Hypothesis

3.1 Hypothesis development

In the previous section I covered the underlying factors of IPO underpricing, the value of underpricing for underwriting firms, and the perceived value of analyst coverage for issuing firms.

I found that, on the one hand there are IPO firms that desire analyst coverage, and are willing to let their public offerings be underpriced to obtain this (Cliff & Denis, 2004, p. 2874). On the other hand there is the fact that analyst coverage has no implications for long-run performance (Bradley et al., 2008, p. 1121). Why then is it that receiving analyst coverage is so important for IPO firms? So important even that, if they don’t receive the expected coverage after the IPO, they switch underwriters for the SEO’s (Cliff & Denis, 2004, p. 2873).

This paper will address the following questions: firstly, what is the perceived value of analyst recommendations, both short- and long-term? According to Bradley et al., the level of analyst recommendations does not affect the long-run performance. However, Merton (1987) states that greater analyst coverage might lead to greater investment recognition while Cliff and Denis (2004) mention that analyst coverage can generate publicity, which in return attracts more customers and increases revenue. This study will examine the effect of analyst recommendations on the long-run performance, as well as on short- and long-run profits.

Secondly: does higher underpricing lead to increased analyst recommendations? While Cliff and Denis (2004) conclude that underpricing is a way to purchase analyst coverage, I will examine if greater underpricing truly leads to increased analyst coverage. If the results show this is the case, and increased analyst coverage is valuable in the long run, this would imply that there is a relation between IPO underpricing and long run value.

4. Data and sample description

4.1 Sample formation

The data used in this study is constructed from multiple sources in order to obtain an as detailed and complete dataset as possible. The different variables will now be discussed per database from which they are gathered.

4.1.1 Thompson One database

Firstly, the sample of IPO firms is obtained by selecting all issuing firms that have completed an initial public offering between the period January 2001 and December 2002. Data on these IPOs such as the company description, filing dates, filing price ranges, offer prices and the amount of shares offered and sold are gathered from the Thompson One database, which is consistent to the method used in prior IPO studies (Das et al., 2006), and Kim and Ritter (1999). Similar to these studies, I exclude unit offerings; offerings of foreign corporations, equity carve outs, filings of financial institutions (with SIC codes between 6000 and 6999) and finally filings of services companies (SIC code greater than 8100) from the sample (Das et al., 2006). Applying this sample criterion results in a sample that focuses on industrial firms only. Since it may be difficult to compare operating performance between industrial and financial firms, a sample with merely industrial firms is best to compare absolute differences in operating performance.

Furthermore, I only consider IPO firms that have offered shares to the U.S. public, since these offerings have most data available for research. IPOs that have been withdrawn within one year are also removed from the sample. After these exclusions, the sample consists of 103 firms that have gone public over a period of two years. With this data I can measure the degree of underpricing by computing the first day initial return (FDR).

4.1.2 CRSP database

Next, I gather data on IPO long-run performance from the CRSP database. This variable is measured by the annualized 3-year buy-and-hold return (3HPR), which is calculated by annualizing the 36 monthly returns subsequent of the IPO (Bradley et al., 2008). If an issuing firm was delisted within 3 years after going public, the delisting return has been taken into account. Furthermore, I compute the 1-year buy and hold return (1HPR), which measures the shorter-term performance of the issuing firm.

4.1.3 I/B/E/S database

Finally, data on analyst coverage is obtained from the I/B/E/S database (Das et al., 2006). I investigate all analyst recommendations within one year of going public. The data obtained on analyst recommendations contains the number of recommendations made within one year of going public (NRC), how many analysts have made the recommendations (NAN), how many recommendations the book runner has made on average (NRCL) and the average I/B/E/S recommendation code (ARC).

Also gathered from the CRSP database is information on turnover figures to measure how the company value reacts to publicity (Hakenes & Nevries, 2000). I used the following figures: return on equity (ROE), return on assets (ROA), net income per share (NET) and earnings per share (EPS), all from the third year after going public.

4.2 Variable construction

Here I will discuss some of the most important variables. Underpricing (FDR) is measured as the first day abnormal return from the offer price to the first closing price. Long-term issuing firm performance is measured by the 1- and 3-year holding period return. The 1-year holding period return (1HPR) is calculated with the returns of the first twelve months, the 3-year holding period return (3HPR) is measured by annualized return of the first thirty-six months. Additionally, I have added four turnover figures from the third year after going public. These are the net income divided by outstanding shares (NET), the return on assets (ROA), return on equity (ROE), and the earnings per share (EPS).

Following, the variables concerning analyst coverage are measured as followed: the number of recommendations (NRC) is measured by the amount of recommendations received in the first year after going public. Furthermore, the average recommendation code (ARC) is added to the model. The content of the recommendation is measured by the I/B/E/S recommendation code, which ranges from 1 to 5, with 1 being a ‘strong buy’ recommendation, followed by ‘buy’, ‘hold’, ‘sell’ and ‘strong sell’,

which is indicated by a 5. Thus the lower the recommendation code, the better the recommendation received.

Finally, I measure how many different analysts have made a recommendation (NAN), since some analyst make only one recommendation, while others cover a firm multiple times. This variable shows how many analysts an IPO attracts. Because literature shows that firms expect high coverage from underwriting firms, I also measured how many recommendations the lead manager has made in the first year after going public (NRCL). If a firm has two or more lead managers, the average number of recommendations received by its lead manager is recorded. A dummy variable has also been added to the model (Book), which is 1 if firms have received coverage from the lead underwriter and 0 if they did not. Even though this variable is quite similar to the NRCL variable, this doesn’t measure the exact amount of recommendations received of the book runner, but only if a lead recommendation has been received at all.

4.3 Descriptive statistics

In this section the overall descriptive statistics, and the descriptive statistics of the returns and recommendations is presented.

Panel A of table 1 provides us with an overview of the initial public offering characteristics. The average first day initial return is presented in the first row of panel A. As expected, the average initial return is positive, namely 11,06% which implies underpricing of the IPO. However, the initial return does show a large amount of variation, ranging from –99,74% to 76,67%. The offer price, presented in the second column, is $14.29 on average, and varies from $5 to $28. The number of unique book runners range from 1 to 3 for the issuing firms, with an average of 1.18 book runners per IPO, and 12.12 managers.

Continuing, panel B gives us information concerning the average returns of the issuing firms after the IPO. It is important to notice that not all firms have survived the first three years of going public. While this is not problematic for calculating the holding period returns– I simply add the delisting return to the calculation–, this does deliver some issues while calculating the turnover figures for the third year after going public. There seems to be no clear consensus for which firms have delisted within three years with respect to the firm industry, size, etc. Therefore, the delisting firms have been omitted for the calculation of the turnover figures, which leaves a lower number of observations.

The first two rows of panel B report the annualized 3-year holding period return and the 1-year holding period return respectively. There is quite some difference between the two variables. To start, the 3-year HPR is a positive number: 9.49% on average, while the 1-year HPR is close to zero, namely a –.6% average. This shows us that on average the holding period return during the first year after going public is only mediocre, but somewhat more rewarding if the stock is held during the first three years after the initial offering. The variation in holding period returns however is much larger for

the 3-year HPR than for the 1-year HPR. The returns range from –163.73% to 243.39% and from – 96.8% and 180.36% respectively. On the contrary, the 3-year HPR shows a lower standard deviation than the 1-year HPR.

The turnover figures are presented in the third to sixth row of panel B. As mentioned above, 19 firms have delisted within a period of three years after going public. Only 67 firms have complete data on their net income, earnings per share, return on assets and equity. There seems to be no clear consent in why some firms miss certain figures. All firms who passed the 3-year mark have reported earnings per share, and therefore EPS is the most complete variable. The average EPS is close to 1 dollar.

Table 1 Overall descriptive statistics (n=103)

Average Minimum Maximum Standard Deviation

n

A. IPO characteristics 103

Initial return (%) 11.06 –99.74 76.67 21.22

Offer price ($) 14.29 5 28 5.28

Number of book runners 1.18 1 3 0.41

Total shares offered (1.000) 8.651 800 72.500 9064

Number of managers 12.12 1 33 12.12 Issue size ($mln) 141.62 7.5 900 165.7 B. Returns 3-year HPR annualized (%) 9.49 -163,73 243.39 39.61 103 1-year HPR (%) -.6 -96.8 180.36 57.76 103 NET ($) 60.85 -97.42 1001.43 145.46 79 EPS ($) .9581 -4.12 16.5 2.26 84 ROA (%) .79 -169.71 63.93 28.32 70 ROE (%) 17.53 -187.84 335.11 63.3 70 C. Recommendations 103 Number of recommendations 9.6 0 44 7.68 Number of analysts 5.72 0 24 4.19 Number of recommendations per book runner

1.55 0 4 1.06

The last panel contains information on the recommendations issuing firms have received. Seven firms have not received any recommendations; this sets the number of recommendations, analysts and recommendations received per book runner to zero. However, for the average recommendation received the number cannot be set to 0 for these firms, since the numerical value of the content of recommendation runs from 1 to 5, 1 being the highest recommendation a firm can receive. Therefore, these seven firms are omitted for the calculation of this variable.

The results are reported in panel C. The first two rows provide us with information on the number of recommendations an issuing firm has received and how many analysts have made these recommendations respectively. Firms have received 9.6 recommendations on average, made by an average number of 5.72 analysts. The variation of the two variables is quite high with a standard deviation of 7.68 and 4.19 respectively.

Firms have received 1.55 recommendations per book runner on average, with a minimum of zero, and a maximum of four recommendations received. This variation is fairly low, especially compared to the other recommendation variables. The last row reports the content of the recommendations; the average recommendation code is a 1.96. The lowest recommendation received is only a 3, which, considering the range from 1 to 5 is quite high. This means that no firm has received a ‘sell’ or ‘strong sell’ recommendation.

4.4 Methodology

Above, the variables of interest are described. This section will enlighten how these variables will be used in the regressions performed in the remainder of this paper. Independent variables to control for underpricing, analyst coverage as well as for firm performance are, consistent with prior studies, the log of the issue size (log size), the number of managers of the IPO (NMN) and two dummy variables: Tech, which indicates whether the issuing firm is in the high-technology industry, and Book, which is equal to 1 if the lead underwriter made a recommendation and zero otherwise. The methodologies for the regressions shown in table 5 and 6 are presented below.

FDR = 𝛼 + ß1*logsize + ß2*Tech + ß3*NMN + ß4*Book + ß5*NRC + 𝜀

Log (1+NAN) = 𝛼 + ß1*logsize + ß2*Tech + ß3*NMN + ß4*Book + ß5*FDR + 𝜀

The last table, table 9, shows the 5 regression models with the firm performance variables as dependent variables. Formulas for these regressions are as follows:

3HPR = 𝛼 + ß1*logsize + ß2*Tech + ß3*Book + ß4*NMN + ß5*NRC + 𝜀

NET = 𝛼 + ß1*logsize + ß2*Tech + ß3*Book + ß4*NMN + ß5*NRC + 𝜀

NET = 𝛼 + ß1*logsize + ß2*Tech + ß3*Book + ß4*NMN + ß5*NRC + ß6*FDR + 𝜀

NET = 𝛼 + ß1*logsize + ß2*Tech + ß3*Book + ß4*NMN + ß5*FDR + ß6*NAN + 𝜀

5. Empirical results

In this section the empirical results are discussed. First I will look at the correlations of the recommendation variables to the underpricing and performance variables respectively. Then I will discuss the relation between analyst coverage and underpricing by following steps similar to Cliff and Denis’ (2004) research. Thereafter, I will look into the relation between analyst coverage and firm performance in a comparable way as Bradley et al. (2008) performed their study. When both relations are established, I will test whether a relation between underpricing and firm value is detectible, and if analyst coverage could be an explanatory variable.

5.1 Correlations

The next two sections contain correlations between underpricing and analyst coverage variables, and analyst coverage variables to performance variables respectively.

Table 2 Variable correlations

This table shows correlations of the recommendation variables to both the underpricing variable and firm performance variables. Recommendation variables shown in the left column are the number of recommendations received (NRC), the number of analysts who made a recommendation (NAN), the average number of recommendations made by the lead underwriter (NRCL) and the average recommendation code (ARC). The columns show underpricing measured by the first-day initial return (FDR), the 3- and 1-year holding period returns (3HPR and 1HPR respectively), earnings (EPS) and net income per share (NET) and finally the return on assets (ROA) and equity (ROE). For the aim of this research, correlations between recommendation variables itself and performance variables itself are not of any relevance.

FDR (%) 3HPR (%) 1HPR (%) EPS ($) NET ($) ROA (%) ROE (%) NRC 0.216** (0.03) –0.0493 (0.62) 0.1628* (0.10) 0.2231** (0.04) 0.4069*** (0.00) 0.057 (0.64) 0.1402 (0.25) NAN 0.187* (0.06) –0.0302 (0.76) 0.191* (0.05) 0.1646 (0.14) 0.5005*** (0.00) 0.0072 (0.95) 0.1217 (0.32) NRCL 0.116 (0.24) –0.1022 (0.31) –0.0382 (0.70) 0.1186 (0.28) 0.2769** (0.01) 0.1186 (0.28) 0.0715 (0.56) ARC 0.079 (0.34)

Note: the p-values are shown in parentheses, *, **, and *** show the level of significance at the 10%, 5% and 1% level respectively.

5.1.1 To underpricing

As can be seen in the first column of table 2, the FDR shows a significant correlation to the number of recommendations, and number of analysts at the 5% and 10% significance level respectively. This means that both variables could be of interest. There is however no mentionable correlation between underpricing and the recommendation code, which means that the degree of underpricing has no effect on how analysts recommend the stock in question. This last results may be explained by the fact that in this sample analysts only have issued recommendations with codes of 3 or higher. The fact that a recommendation is issued at all might be more valuable in this model than the code the recommendation contains.

5.1.2 To firm performance

The last six columns of table 2 show the holding period returns and turnover figures. The 3-year holding period only shows a small correlation to the recommendation variables. Interestingly, all correlations are negative, meaning that if a firm receives more recommendations, the return of holding the stock for three years decreases. However, the p values indicate that the coefficients for the 3-year HPR are far from significant, thus I cannot attach great meaning to these negative values. This result is different from the results obtained by Bradley et al. (2008). While they do not report the correlation between the 3-year HPR and recommendation variables, their model yield a fairly significant adjusted R2_{, which implies significant correlation unlike our model. This difference in results may be caused by} the difference in sample size and time period since the rest of our models are quite equal. Since the 3-year HPR is the main variable used by Bradley et al. (2008), we will not yet disregard the 3HPR variable and will test it in later sections.

The correlations with the 1-year holding period return seem to be more significant, but only at the 10% significance level. The fourth column reports the correlations for the earnings per share, of which only the correlation to the NRC seems to be significant at the 5% level. The following column, showing the correlations between the net income per share and recommendation variables, clearly appears the most promising. The correlations of NET to NRC and NAN are both positive, and significant at the 1% level, and the correlation to NRCL is also positive and significant at the 5% level. It seems that the net income per share variable is a suitable variable to measure how issuing firms benefit from analyst coverage. The two to last columns containing regressions of ROA and ROE show no significant correlations and are therefore disregarded for the remaining regressions of this study.

5.2 Relation between underpricing and analyst coverage

The first question addressed in this paper is if there is any relation between underpricing and analyst coverage. According to Cliff and Denis (2004), analyst coverage is bought via underpricing thus the

higher the analyst coverage, the higher the underpricing must be. Table 2 shows a correlation between the number of recommendations received and the level of underpricing at the 5% significance level.

Consistent with Cliff and Denis’ (2004) analysis, I will begin with reporting the frequency of analyst coverage. As can be seen in the table 3, 83 of the 103 firms have received a recommendation from their book runner as well as from another analyst. Furthermore, 3 firms have only received a recommendation from their book runner, while 10 firms have not received a lead-recommendation but did receive a recommendation from another analyst. The remaining 7 firms have not received any recommendations at all.

Next, I present the IPO characteristics based on the underpricing quintiles in table 4. The number of recommendations received shows an increase the higher the firms initial return is (Qhigh). The same applies to the number of analysts who made the recommendation. Similar to the results of Cliff and Denis (2004), firms in the lowest underpricing quintile receive a recommendation of the lead underwriter only 73% of the time, while this is 88% for firms in the highest underpricing quintile. The last column displays a correlation between the number of managers of the IPO and IPO underpricing.

Table 3 Frequency of analyst coverage

Shown below are characteristics of analyst coverage within 1 year after going public. The sample includes 103 issuing firms and is divided by the type of recommendations received. Per category the number and percentage of firms of the total sample are reported.

Recommendation received from… Number of firms Percentage firms

Lead underwriter and other analyst 83 80.85

Only lead underwriter 3 2.91

Only other analysts 10 9.71

Neither the lead underwriter or other 7 6.80

Table 4 Mean recommendations by quintiles

This table shows IPO characteristics as well as recommendation means by quintiles of IPO underpricing. The sample has been divided into four quintiles, running from Qlow to Qhigh. The book variable shows the percentage of firms per quintile that have received a recommendation from their lead underwriter.

FDR quintile FDR (%) NRC NAn Book (%) NMn Low –9.18 8.42 5.38 73.07 11.69 Q2 3.18 7.19 4.31 84.62 10.42 Q3 13.94 10.62 6.35 88.46 11.85 High 37.28 12.28 6.88 88 14.6

Table 5

The table below shows the coefficients of the OLS regression with underpricing as the dependent variable

FDR Intercept 20.22 (0.06) Log size –6.602 (0.03) Tech 5.387 (0.25) NMN 0.406 (0.24) Book 5.831 (0.32) NRC 0.997 (0.01) Adjusted R2 _0.062 Table 6

The table below shows the coefficients of the OLS regression with the number of analysts as the dependent variable

Log (1+NAN) Intercept 0.692 (0.00) Log size 0.412 (0.00) Tech –0.096 (0.30) NMN –0.001 (0.88) Book 0.608 (0.00) FDR 0.005 (0.02) Adjusted R2 _0.630

Note: the p-values are shown in parentheses, *, **, and *** show the level of significance at the 10%, 5% and 1% level respectively.

5.2.2 OLS regression with underpricing as de dependent variable

For the OLS regression, the recommendation variables are the explanatory variables, and the FDR is the dependent variable. Consistent with prior studies (Cliff & Denis, 2004) independent variables are the log of issue size, technology dummy, lead-recommendation dummy and the number of managers as described in the methodology section.

The regression with FDR as the dependent variable is presented in table 5. As showed, the regression model explains merely a small portion of the cross-sectional variation in underpricing with an adjusted R2 _{of just 0.062. This result differs from the results found by Cliff and Denis (2004).} Explanations for the difference in results might be the difference in sample size, and the fact that Cliff and Denis have added more control variables to their model which were omitted in this model given the focus of this paper.

5.3 Relation between analyst coverage and firm performance

The second part of my empirical research will focus on the question whether analyst recommendations are correlated to firm performance, which I measure in several ways. This structure of this section is comparable to that of the research of Bradley et al. (2008).

I begin with a cross-sectional regression of analyst coverage to examine the factors that are related to analyst coverage presented in table 6. The dependent variable of the regression is the log of one plus the number of analysts who provided coverage. Similar to Bradley et al. (2008), and the regression presented in table 5, independent variables are the log of firm size, number of managers, a dummy variable indicating high-tech firms, a dummy variable indicating whether the firm has received a recommendation of the lead-underwriter and the initial return.

The results are somewhat different from those of Bradley et al. While the issue size seems to be of great significance in explaining the amount of analyst following, the number of managers is not. The initial return is of importance too at the 1% significance level. The regressions yields an adjusted R2_{of 0.630, which means the model is rather good in explaining analyst following. Note that in this} model increased underpricing implies increased analyst following which is a reversed causality compared to the model of Cliff and Denis.

Thus, underpricing in our sample cannot be explained merely by the control variables and the number of recommendations received, but the number of analysts following a stock however can be explained by the control variables and underpricing. This result is different from our hypothesis. It seems the hypothesis of Cliff and Denis stating issuing firms purchase analyst coverage through underpricing does not hold for our sample, but that increased underpricing does imply larger analyst following. This result is more consistent with previous studies, such as the research executed by Chemmanur (1993).

5.3.1 Returns by quintiles

In panels A and B of table 7, I have sorted the sample into four quintiles based on the number of recommendations received and the number of analysts respectively. For the highest NRC quintile showed in panel A, all return variables but the 3-year HPR show the highest results. This shows us that if firms are in the high-end of receiving recommendations, they display high performance compared to the rest of the sample. Interestingly enough, this result is reversed for the 3-year holding period return. Firms who are in the lowest quintile of receiving recommendations tend to outperform in the long run in terms of the holding period return. EPS and the 1-year HPR are the only variables that show a steady increase that corresponds to the increasing number of recommendations received.

The results presented in panel B are somewhat different from those in panel A. Firms in the highest quintile of receiving recommendations of unique analysts do not tend to perform best compared to the rest of the sample. This is only different for the NET variable, which shows a steady correlation to the NAN quintiles, which we also saw table 2. The 3-year HPR shows similar results in panel B as in panel A. Firms who are in the low end of receiving unique analysts’ recommendations outperform in the long-run in terms of the holding period return.

Table 7 Mean returns by quintiles

Shown here are the mean returns per NRC quintiles in panel A, and NAN quintiles in panel B. The sample is sorted into four quintiles based on the number of recommendations received, and based on the number of analysts who made recommendations, running from low to high. The return variables are the earnings and net income per share, and the 1- and 3-year holding period return in percentages.

A. NRC quintile NRC EPS ($) NET ($) 1HPR (%) 3HPR(%)

Low 2.28 0.03 4.98 –8.23 14.51

Q2 6.6 1.18 78.14 –7.96 7.66

Q3 10.64 1.03 45.15 3.91 5.76

High 20.5 1.72 137.93 11.59 9.44

B. NAN quintile NAN

Low 1.94 0.06 6.9 –13.1 15.49

Q2 4.44 1.04 49.8 –6.5 –0.46

Q3 6.96 1.94 79.29 12.56 12.85

High 12.56 0.71 138.24 10.78 6.65

Table 8 Returns by lead coverage status

In the table below IPO characteristics and returns are shown by lead coverage status. The first row shows the results for the firms who have received a recommendation of their lead-underwriter. The bottom row does the same for firms who have not received a lead recommendation.

Lead recommendation: FDR (%) Tech (%) 3HPR (%) 1HPR (%) NET ($) EPS ($) Yes (n=86) 12.15 29.07 6.45 1.14 69 1.03 No (n=17) 5.03 23.53 24.72 –9.38 10.51 0.60

Finally, I divided the IPO sample into firms who have received a lead recommendation, and firms who have not. Consistent with Bradley et al. (2008), I find that firms outperform in terms of the 3-year holding period return if they have not received a recommendation of the lead underwriter. However, the opposite is the case for the other three performance variables. Firms do better in terms of the net income per share and earnings per share in the third year after going public when they have received recommendations from the lead underwriter and the same goes for the 1-year holding period return. Thus, while our results match those of Bradley et al. (2008), when adding extra variables as performance variables, different results are uncovered.

5.3.2 OLS regression with firm performance as the dependent variable

Following, this section will present the results of the cross-sectional regressions with performance variables as the dependent variable. Bradley et al. use the 3-year HPR as the long-run performance value of the IPO firms. As seen in table 2, the 3-year HPR does not display significant correlations to the recommendation variables in our sample. To show consistency I will first execute the regression following the research method used by Bradley et al with the 3-year HPR as the dependent variable.

Thereafter I will use the net income variable as the dependent variable while performing the regression since I expect a better fit looking at the correlations presented in table 2.

The regression is presented in table 9 and uses similar independent variables as the cross-sectional regression performed by Bradley et al. (2008). As anticipated, given the results in table 2, the regression in the first column does not do a good job in explaining the firm performance when the dependent variable is the 3-year holding period return. The adjusted R2 _{is only –0.003, which} approaches zero.

In the third column I have performed the same regression but with the net income per share of the third year after going public as the dependent variable. This model offers a much higher adjusted R2 _{of 0.229, which was predictable given the fact that NET is correlated to the NRC at the 5%} significant level.

Table 9

Shown here are the coefficients of the OLS regressions with performance variables as the dependent variables. The first two models have the 3-year holding period return as the dependent variable and models 3 to 5 use net income per share as the dependent variable. Independent variables are the log of proceeds, a dummy variable indicating whether a firm is in the high-technology sector, a dummy variable indicating whether a firm has received a lead recommendation, the number of managers of the IPO, the number of recommendations received, the number of analysts and the underpricing variable.

3HPR (1) (2) NET (3) (4) (5) Intercept 0.293 (0.16) 0.348 (0.1) –238.29 (0.00) –207.85 (0.01) –167.66 (0.02) Log size 0.011 (0.85) –0.007 (0.90) 72.196 (0.00) 62.48 (0.01) 47.93 (0.02) Tech –0.042 (0.64) –0.027 (0.76) –1.551 (0.96) 11.359 (0.74) 17.83 (0.58) Book –0.179 (0.12) –0.163 (0.16) –23.127 (0.61) –15.594 (0.73) –35.856 (0.41) NMN –0.007 (0.27) –0.006 (0.35) –2.808 (0.24) –2.493 (0.29) –2.60 (0.25) NRC 0.000 (0.98) 0.003 (0.65) 2.392 (0.41) 3.667 (0.213) FDR –0.003 (0.161) –1.329 (0.01) –1.391 (0.05) NAN 13.764 (0.01) Adjusted R2 _{–0.003 0.008 0.229 0.252 0.309}

Note: the p-values are shown in parentheses, *, **, and *** show the level of significance at the 10%, 5% and 1% level respectively.

5.4 Relation between underpricing and firm performance.

I hypothesized that causality exists between analyst coverage and underpricing, in which higher analyst coverage would imply higher underpricing. As presented in section 5.3, the causality in this research sample is reversed. Furthermore, in the second part of the empirical analysis the relation between analyst coverage and firm performance has been investigated, and a link has been found between firm performance, measured by the net income per share, and analyst coverage in terms of the number of recommendations received, and the amount of analysts covering the stock. This means that the more recommendations an issuing firm receives, the higher the net income per share is.

After establishing the relationships, it could be tested whether there is a relation between underpricing and firm performance in which analyst coverage might be an explanatory variable. Even though the first part of this paper’s hypothesis was partially incorrect, the expected causality is reversed; this last part of the research can still be conducted. Higher underpricing implies higher analyst coverage, and higher analyst coverage implies higher firm performance, which implies that higher underpricing leads to higher firm performance. Therefore, the dependent variable in this last regression is the net income as the firm performance variable. This variable is chosen given the high correlation to analyst coverage variables presented in table 2, and table 9. The independent variables in this model are similar to the independent variables used in the earlier regression with net income as the dependent variable, plus the underpricing variable (FDR).

The results of the model are presented in the one to last column of table 9. With an adjusted R2 _{of 0.252 the model is fairly sufficient in explaining the net income per share. The coefficient for} NRC is positive, indicating a positive relation between the number of recommendations and net income. The p-value however shows fairly small significance.

Since table 2 showed us an even higher correlation between net income and number of analysts than it did for net income and the number of recommendations, in the last column I replaced NRC with NAN. This gives us an even greater adjusted R2 _{of 0.309, and a positive NAN coefficient at} the significance level of 1%. Net income therefore seems to be more related to how many analysts recommend a stock, than to how many recommendations those analysts truly make. For consistency’s sake, I also performed the same regression, including the underpricing variable, with the 3-year holding period as the dependent variable. The results are presented in model 2 of table 9 and again can be seen that the adjusted R2 _{found is very low, a 0.007.}

6. Conclusion and discussion

6.1 Underpricing

This paper aimed to find a relation between underpricing and analyst coverage, in which increased analyst coverage would imply greater underpricing. Even though the models presented in table 5 combined with the correlations presented in table 2 show us such a causality, the adjusted R2 _{is very} low, namely a 0.064. Interestingly enough, the model offered in table 6, which reverses the causality

gives us a much higher adjusted R2 _{of 0.533. This is consistent with previous papers, such as that of} Chemmnanur(1993) who hypothesized that firms compensate externals via underpricing and not underwriters.

6.2 Firm performance

Secondly, this paper tried to answer in which ways analyst coverage is valuable to issuing firms in terms of firm value. Firm performance variables were the 1- and 3-year holding period return, net income and earnings per share, and the return on assets and equity. The last two variables were disregarded for further regression after the correlations in section 5.1 showed no significant correlation between these variables and the recommendation variables. Even though the 3-year HPR also showed no significant correlation, the variable was kept into consideration since it is a main variable in the similar research conducted by Bradley et al. (2008).

The subsequent research was performed in line with that of Bradley et al. (2008), and while some sections yielded similar results, such as table 8 which showed the 3-year HPR by lead underwriter status, other sections presented different outcomes. Section 5.3.2 for example showed us that although the 3-year HPR was a fitting variable in the model of Bradley et al. (2008), our model showed no significant correlations and adjusted R2_{. The net income variable however did show} significant correlations to our recommendation variables, and yielded an adjusted R2 _{of 0.229.}

6.3 Relation between underpricing and firm performance

In the last empirical research section I tried to relate underpricing directly to firm performance as measured by net income. This relation is expected to exist since the connection of analyst coverage to underpricing and firm performance respectively has been proven in sections 5.2 and 5.3. The model indeed showed that given the adjusted R2_{, 25.2% of the net income in the third year after going public} is explained by underpricing, the number of recommendations and various control variables. This percentage is even 30.9 if the number of recommendations is replaced by the number of analysts who made recommendations within the first year after going public.

Therefore, the last part of the hypothesis is correct. There is a direct positive relation between underpricing and firm performance in terms of net income per share in the third year after going public, in which the level of analyst coverage as measured by the number of recommendations received and the number of analysts who made recommendations are explanatory factors. However, the number of analysts is better in explaining this relation given a significance level of 1%. As mentioned in section 6.1, this might be explained by the reversed causality detected in table 6: higher underpricing leads to higher analyst interest and therefore a higher number of recommending analysts.

Bibliography

Bradley, D. J., Jordan, B. D., & Ritter, J. R. (2003). The quiet period goes out with a bang. Journal of Finance, 58(1), 1–36.

Bradley, D. J., Jordan, B. D., & Ritter, J. R. (2008). Analyst behavior following IPOs: The ‘bubble period’ evidence. Review of Financial Studies, 21(1), 101–133.

Bradley, D, J., Chan, K., Kim, J., & Singh, A. (2008). Are there long-run implications of analyst coverage for IPOs? Journal of Banking & Finance, 32, 1120–1132.

Chemmanur, T. (1993). The pricing of initial public offerings: A dynamic model with information production. Journal of Finance, 48, 285–304.

Cliff, M., & Denis, D. (2005). Do initial public offering firms purchase analyst coverage with underpricing? Journal of Finance, 59(6), 2871–2901.

Das, S., Guo, R., & Zhang, H. (2006). Analysts’ selective coverage and subsequent performance of newly public firms. Journal of Finance, LXI(3), 1159–1185.

Hakenes, H., & Nevries, P. (2000). Underpricing initial public offerings due to the value increasing publicity effect. Working paper, Westfalische Wilhems-Universität Munster.

Irvine, P. J. (2003). The incremental impact of analyst initiation of coverage. Journal of Corporate Finance, 9, 431–451.

James, C., & Karceski, J. (2006). Strenght of analyst coverage following IPOs. Journal of Financial Economics, 82(2), 1–34.

Liu, X., & Ritter, J. R. (2011). Local underwriter oligopolies and IPO underpricing. Journal of Financial Economics, 102(3), 579–601.

Michaely, R., & Womack, K. L. (1999). Conflict of interest and the credibility of underwriter analyst recommendations. Review of Financial Studies, 12(4), 653–686.

Rajan, R., & Servaes, H. (1997). Analyst following of initial public offerings. The Journal of Finance, 52(2), 507–529.

Ritter, J. R., & Welch, I. (2002). A review of IPO activity, pricing and allocations. The Journal of Finance, 57(4), 1795–1828.