Analyst recommendations and strategic distortion around lockup period expiration

(1)

1

MSc Business Economics Specialization: Finance

“Analyst Recommendations and Strategic Distortion

around Lockup Period Expiration”

MSc Thesis

Luuk de Wildt Student number: 10166548

Date: 15 December 2015

Supervisor: Jens Martin

(2)

2

Statement of Originality

This document is written by Student Luuk de Wildt 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.

(3)

3 Abstract

This paper examines the behavior of security analysts around the expiration of the lockup period, using a sample of US IPOs during the period 1996-2014. Both the change in the level of recommendations issued as well as the level of strategic distortion are examined. This is the first study in which the change in strategic distortion around the lockup period is examined. I find that a drop in the level of recommendations after the expiration of the lockup period for unaffiliated analysts, indicating that those analysts issue upward biased recommendations during the lockup period. For analysts affiliated with the lead manager I find an increase in the level of recommendations after the lockup period expiration. The level of strategic distortion does not change significantly after the expiration of the lockup period for unaffiliated analysts. For affiliated analysts I find a significant increase in the level of strategic distortion after the lockup period expiration.

(4)

4

1. Introduction

The expiration of the lockup period is a popular topic of interest for many research papers. As argued by Field and Hanka (2001), a lot of this interest in lockup expiration is derived from the fear that insiders will try to sell huge blocks of shares. The IPO of Alibaba and following lockup expirations provides a recent example of the interest in lockup period expirations. Investors feared for an enormous selling pressure, as a huge group of insiders were allowed to sell their shares. The impact was a decline in the stock price of about 0.9% (Reuters).

Essentially all IPOs include share lockup agreements. Those are voluntary agreements between the underwriting investment bank and the original pre-IPO shareholders, regarding the sale of shares. The lockup agreement prohibits pre-IPO shareholders and other insiders from selling shares without the permission of the underwriting investment bank for a specific period, which lasts typically for 180 days (Field & Hanka, 2001).

Theoretically, insiders can also sell their equity stake during the IPO itself by issuing their secondary shares. However, insiders generally wait until the end of the lockup period to do this. Brau and Fawcett (2006) show that this is because insiders fear that selling large secondary equity stakes during the IPO will lead to bad market signals. Brav and Gompers (2003) find a high selling pressure at the expiration of the lockup period.

Martin (2006) explains that this pressure is caused by both insiders, who wish to exit, as well as by investment banks, which support the share price until the end of the lockup period in order to maintain a good reputation.

Prior evidence on analyst behaviour shows that analysts in general issue optimistic recommendation. Michaely and Womack (1999) show that analysts deviate from their natural role as providers of neutral investment recommendation during IPOs, by issuing overoptimistic recommendations. Ljunqvist et al. (2009) demonstrate that even unaffiliated analysts issue upward recommendations. They argue that analysts do this in order to increase their chances of being part of a future IPO deal. Aggarwal et al. (2002) develop a model in which manager strategically underprice the IPO. This underpricing creates information momentum and in turn attracts new investors. Therefore, at the end of the lockup period managers now can sell their shares at a higher price than during the IPO itself (Aggarwal, 2002).

Malmendier and Shanthikumar (2014) make a distinction in the two ways analysts can provide information: recommendations and earnings forecasts. They construct a two-tongues metric for strategic distortion, which measures the difference between the recommendation surprise and the earnings forecast surprise. Strategic distortion concerns the misalignment of

(6)

6

incentives. It actually reflects incentives to please the management of companies and to trigger small investors to purchase the specific stock (Malmendier & Shanthikumar, 2014). Analysts that distort strategically issue very optimistic recommendations and at the same time lower, beatable earnings forecasts.

Prior research mainly focuses on IPO underpricing and price support in the first weeks after the IPO. For investment banks this is important, since underwriting successful IPOs increases their chances of being part of a future deal (Michaely and Womack, 1999). Jens Martin (2006) in his paper argues that this price support actually lasts as long as the entire lockup period. He therefore states that price support can also be interpreted as a way to allow insiders who wish to exit, can do this at a high price.

This paper, in contrast to most of the prior research, focuses on price support during the entire lockup period. It presents the first study in which strategic distortion is directly related to the expiration of the lockup period. Based on the papers of Jens Martin (2006) and Malmendier and Shanthikumar (2014) I develop two main hypotheses. First, I expect the level of recommendations to be significantly higher during the entire lockup period compared to after the lockup period. This can be seen as a way to support the stock price. Second, I expect the level of strategic distortion to be significantly higher during the lockup period compared to after the lockup period.

This paper finds evidence that is consistent with the first hypothesis. However, my findings are not consistent with the second hypothesis. Using US IPO data, I find a significant drop in the level of recommendations after the lockup period for unaffiliated analysts. For affiliated analysts, however, I find an increase in the level of recommendations after lockup period expiration. For the level of strategic distortion, I find a significant drop in the level of strategic distortion for affiliated analysts after the expiration of the lockup period.

The remainder of this paper is structured as follows. Section two provides an overview of the existing literature regarding IPO lockups and lockup expirations. Section three describes the data and provides some summary statistics. Section 4 provides an overview of the methodology used. In section 5 the results are presented and discussed. Section 6 provides a robustness check. Finally, section 7 presents a conclusion.

(7)

7

2. Theoretical review

2.1 The lockup period; agreements and insider selling

Essentially all IPOs include share lockup agreements. Those are voluntary agreements between the underwriting investment bank and the original pre-IPO shareholders, regarding the sale of shares. The lockup agreement prohibits pre-IPO shareholders and other insiders from selling shares without the permission of the underwriting investment bank for a specific period, which lasts typically for 180 days (Field & Hanka, 2001). Typically the lockup period covers the shares not sold during the IPO. However, not all pre-IPO shares are necessarily subject to a lockup agreement. Field and Hanka (2001) find that on average 95 percent of the shares hold by pre-IPO shareholders are subject to lockup agreements.

Next to selling, the lockup period also prohibits insiders from offering, short selling, contracts to sell or any other possible way to reduce their ownership in the company doing an IPO without the permission of the underwriting investment bank (Martin, 2008). Insiders commonly own quite a large portion of the company of the firm going public. Lockup agreements make sure that those insiders will maintain a significant ownership position in the firm and thereby aligns their interests with the interests of new shareholders (Bradley et al., 2001). This gives a credible signal to market that insiders will not cash out in advance of bad news. During the lockup period, trading in IPO shares is very different from trading in well established companies (Field and Hanka, 2001). Only a small fraction of the shares are traded and there is little to no selling of insiders and other pre-IPO shareholders.

Field and Hanka (2001) as well as Keasler (2001) both argue that underwriters prefer a longer lockup period. This is because of the stabilization effect it has for the IPO, since insiders cannot sell their shares. On the other hand, insiders prefer a shorter lockup period, because lockup periods come at the cost of illiquidity (Field and Hanka, 2001).

The expiration of the lockup period is not the first possibility for insiders to reduce their equity stake in a company. Theoretically, insiders can also sell their equity stake during the IPO itself by issuing their secondary shares. However, insiders generally wait until the end of the lockup period to do this. Brau and Fawcett (2006) show that this is because insiders fear that selling large secondary equity stakes during the IPO will lead to bad market signals. This in turn leads to a lower share price for which insiders are able to sell their secondary shares. Aggarwal et al. (2002) also look at lockup expiration selling. They find that historical IPO data for developed countries shows an average first-day underpricing of about 15%. So, by selling

(8)

8

shares during the IPO, insiders would leave some serious money left on the table (Martin, 2008). In their paper Aggarwal et al. (2002) develop a model in which manager strategically underprice the IPO. This underpricing creates information momentum and in turn attracts new investors. Therefore, at the end of the lockup period managers now can sell their shares at a higher price than during the IPO itself (Aggarwal, 2002).

As explained above, the expiration of the lockup period essentially forms the first real opportunity for insiders to sell their secondary shares. After the expiration of the lockup period, insiders are free to sell their shares up to certain volume limits (Field & Hanka, 2001). Brav and Gompers (2003) find that insiders sell their shares as soon as the lockup period expires. Field and Hanka (2001) find a permanent increase of 40 percent in trading volume around the expiration of the lockup period, accompanied by a significant negative three-day abnormal return of -1.5 percent. Martin (2008) looks at open market transactions by insiders to find out whether they bought or sold shares. He finds a sell-to-buy ratio of 7 to 1 for his sample three years after their respective IPOs, indicating that insider sales outnumber insider purchases over the long run.

2.2 The value of coverage, analyst incentives and the lockup period

In this chapter I describe the value of analyst coverage for firms. Moreover, I discuss the kind of incentives analysts face during the lockup period, as well as in after the termination of the lockup period.

Analyst coverage is valuable for companies, as is demonstrated by various prior studies (Liu and Ritter, 2010). Empirical evidence on the reaction of the market to analyst coverage indicates that it is rational for companies to estimate the value of analyst coverage. Bradley et. al (2008) find a 3% abnormal announcement return for companies initiating unanticipated analyst coverage in the year following the IPO. Kelly and Ljungqvist (2007) report that firm value drops when analysts terminate their coverage of a specific firm.

At the end of the lockup period, there is a high selling pressure (Brav and Gompers, 2003). For pre-IPO corporate insiders this forms the first moment on which they can sell their shares. Therefore, for the insiders that wish to exit, the share price at the end of the lockup period is of enormous importance. Investment banks at the same time want to maintain their good reputation by supporting the share price during the entire lockup period (Martin, 2008).

This leads to pressure on the analysts. Analysts want to maintain their reputation as provider of good and reliable information. However, at the same time they also want to increase their chance of being part of a future underwriting syndicate as well. The latter can be achieved

(9)

9

by issuing upward biased recommendations, which is in favour of company management. Of course, the performance of the stock price of the firm going IPO determines the importance of issuing upward biased recommendation. The urge for good recommendations will be higher in case of bad stock performance following in the period directly after the IPO.

Michaely and Womack (1999) in their paper describe the behavior of analysts. In effect, the role of analysts is to provide information on investment decisions on a neutral basis. However, they show that during IPOs analysts indeed deviate from this natural role and issue biased, overoptimistic recommendations. Michaely and Womack (1999) show that affiliated analyst issue more favorable recommendations. However, those stocks perform worse in the short- and long-term. Degeorge et. al (2007) and Ljungqvist et. al (2009) demonstrate that even analysts not affiliated with the underwriting of the IPO issue upward biased recommendations. The reason for this is that doing so, increases the chance for their investment bank to be part of a future underwriting syndicate.

2.3 Specific insider groups

As described before, analysts in general have incentives to give overly optimistic recommendations to support insiders after an IPO. Moreover, analysts show distortion in their behaviour by issuing overly positive recommendations in combination with moderate to low earnings forecasts. In this section I will look at two specific groups of insiders, which both have a strong interest in the share price at the end of the lockup period. Both of these groups have some kind of leverage over the investment banks and therefore might be more likely to receive this favourable treatment by analysts.

The first group of special insiders is venture capitalists. Looking at the IPO market, venture capitalists are important players. Venture capitalists have various ways to exit their investments, of which IPOs are a very popular one (Gompers and Lerner, 1998). The performance of a venture capitalist investment, which is usually measured by the internal rate of return, is calculated over the investment period. The investment period starts on the day of the investment and runs until the moment that the shares are transferred to the limited partners (Gompers and Lerner, 1998). Based on the performance over this investment period, fees are calculated.

Liu and Ritter (2010) generate a new theory related to the underpricing of IPOs backed by venture capitalists which they call “the analyst lust theory”. They demonstrate that venture capitalists indeed are focused on the share price at the day that shares in the company are

(10)

10

distributed to the limited partners. This exit moment is usually between six months to one year after the IPO (Liu and Ritter, 2010).

Venture capitalists are repeated players in the market for IPOs. Therefore, they have some kind of leverage over the investment banks, because they can bring future lucrative business to the underwriting investment banks. Chen and Ritter (2000) show that typically 7% of the IPO proceeds are divided between the underwriting investment banks, which leads to huge amounts of money for the investment banks.

The second group of insiders consists of the management of companies. Unlike the venture capitalists, management is not a repeated manager in the IPO market. However, the management team of a company leads the company and makes a lot of investment related decisions, including m&a activities and SEOs, which might need future assistance from investment banks (Martin, 2008). Therefore, the company management can actually bring new lucrative future business to the investment banks. This gives them leverage over the investment banks, since they are very willing to retain the specific company as customer.

2.4 Strategic distortion

As is shown in a large body of previous research, analyst recommendations in general are positively biased. Michaely and Womack (2005) provide a detailed overview of previous literature regarding analyst recommendations. The explanations for this positive bias vary widely, however they fall into two main categories: strategic and non-strategic (Malmendier & Shanthikumar, 2014). In their research, Malmendier and Shantikumar (2014) look at two different ways in which analysts can provide information: recommendations and earnings forecasts.

Analysts in general aim to please the management of companies by issuing optimistic recommendations in order to generate new corporate finance business, as well as to persuade investors to invest in those stocks (Michaely & Womack, 1999). Moreover, buy-side clients force the analysts to provide positive recommendations on the stocks they hold (Boni and Womack, 2002).

Nonstrategic distortion reflects the fact that analyst have too-positive expectations (Malmendier & Shanthikumar, 2014). These might be due to self-selection into covering the stocks they think are doing well, or due to trust (Teoh & Wong, 2002). Non-strategic distortion is characterized by the fact that both the recommendations and forecasts issued are very optimistic.

(11)

11

Strategic distortion concerns the misalignment of incentives. It actually reflects incentives to please the management of companies and to trigger small investors to purchase the specific stock (Malmendier & Shanthikumar, 2014). Analysts that distort strategically issue very optimistic recommendations and at the same time lower, beatable earnings forecasts.

A large part of previous research provides evidence regarding distortions in the recommendations of analysts. However, it is not so clear why those distortions occur. Malmendier and Shantikumar (2014) try to make this cause more transparent, and provide empirical evidence in which they make a distinction between the moral hazard and selection hypotheses.

2.5 Trade reactions to analyst recommendations and forecasts

This section describes how investors trade, based on the recommendations and forecasts of analysts. As described above, recommendations are positively biased. Therefore, I expect investors to account for this distortion in their trading.

Malmendier and Shantikumar (2007) analyze how investors account for those distortions. In their research they make a distinction between large and small traders. They find that both small and large traders show significant trade reactions following the issue of a recommendation. However, only large traders account for the distortion in recommendations and adjust their trading for this (Malmendier and Shantikumar). Small traders follow the recommendations, without any discouting. Moreover, small traders attach more importance to whether earnings forecasts are met (Malmendier and Shantikumar).

Given these findings, this gives analysts incentives to distort their recommendations in an upward direction. Martin (2008) states that issuing biased recommendations is costly. However, based on the research of Malmendier and Shantikumar (2007) it seems to be not so costly. This gives even larger incentives for analysts to issue those upward biased recommendations. The same does not hold for forecasts, since management likes beatable forecasts.

(12)

12

3. Data sources and descriptive statistics

This research focuses on companies going public between 1996 and 2014. My starting point is 1996, since there is very little recommendation data available for earlier years. I include companies conducting an IPO and issuing common shares, based on data provided by the Thomson One database. Firms must be included in the Centre for Research in Security Prices (CRSP) database sometime after the public offering, but before the expiration of the lockup period. All firms in my sample are listed on the New York Stock Exchange (NYSE), NASDAQ or American Stock Exchange (AMEX) subsequent to their IPO.

Consistent with prior studies, I exclude IPOs by closed-end funds, American depository receipts (ADRs), real estate investment trusts (REITs), unit offerings, financial companies and utilities (Brav & Gompers, 2003). Moreover, I also drop all offerings for which the offer price is below $5 as well as firms for which information regarding the lockup period is not available, since this is the central topic of this study (Ritter & Zhang, 2007). I use Thomson One to obtain offer prices, length of the lockup period, primary and secondary shares offered, insider ownership at time of the offering, proceeds and stock exchange on which the company went public.

I obtain data regarding analyst recommendations, quarterly earnings forecasts and information about brokerage houses and analyst identities from the IBES database. I obtain stock price data from the CRSP database.

I use quarterly earnings forecasts occurring between the previous earnings announcement and the announcement to which the quarterly forecast relates, obtained from the IBES database. Split-adjusted data is used. Although the split-adjusted data might be imprecise due to IBES’s rounding’s of forecasts, this imprecision is very small and will most likely not affect my results.

IBES converts the recommendations of brokers into a numerical, based on codes ranging from 1 to 5. Just like Malmendier & Shanthikumar (2014), I reverse this coding to 5 = strong buy, 4 = buy, 3 = hold, 2 = sell and 1 = strong sell. By doing this, a higher recommendation reflects a better recommendation and thus makes the interpretation of the results easier.

IBES provides the recommendation and earnings forecast data in two separate files. In order to match a specific analyst’s recommendation and earnings forecast, I use the analyst identity code variables “amaskcd” and “analyst” provided by IBES.

Thomson One provides a list of all lead- and co-managers involved in the IPO underwriter syndicate. I make use of this list to group the recommendations based on the

(13)

13

affiliation of the analyst who issued the recommendation. I manually merge this data to the recommendations data provided by IBES, thereby using the abbreviated broker codes provided by IBES. Due to time constraints, I am only able to do this for the lead-managers. Therefore, I consider an analyst to be affiliated if the analyst is affiliated with the lead-manager of the specific IPO. Besides this, I use the total number of underwriters (co- and lead managers) data provided by Thomson One in my research as well.

To determine the reputation of the underwriter, I use the Ritter (2003) underwriter ranking. I obtain this data from Ritter’s own website. Since IBES only provides abbreviated broker codes, I manually adjust the full broker names used in the ranking to the abbreviated ones in IBES. By doing this, it is possible to merge the underwriter ranking data with IBES recommendations and forecast data. Underwriter ranking data is available for different time periods, since underwriter reputation changes over time. If underwriter ranking data is missing for a specific time period, I use the same rank as the preceding time period for which data was available. Brokers not included in the Ritter underwriter ranking receive the lowest score possible, which is equal to one. The reason for doing this is that the list only contains the brokers with the highest reputations, so the remaining brokers per definition get the lowest score possible (Ritter, 2003).

3.1 Summary statistics

Table 1 presents summary statistics for the sample. After removing companies for which no lockup information was available, as well as companies for which no financial data was available in CRSP, my dataset comprised 961 IPOs. Approximately half of these IPOs (49.4%) were backed by a venture capitalist. I screen my sample in order to control for data inconsistencies, such as analyst recommendations outside the 1-5 range, as well as for outliers. In about 70% of the IPOs in my sample, only primary shares were offered. This indicates the theory stated before, namely that insiders in general refrain from selling their secondary shares during the IPO itself and wait for this until the expiration of the lockup period. The relatively high mean value for pre-IPO insider ownership percentage (63%) shows that possible insider selling around the lockup expiration indeed is an important topic, given the possible high selling pressure and possibly huge effects on the share price. The length of the lockup period is very concentrated around 180 days. Although the mean value of 173 days is a little lower, approximately 80% of the IPOs in my sample included a 180 days lockup period.

(14)

14

INSERT TABLE 1 HERE

Below I present a correlation matrix from the main variables of interest I use in my regressions. As can be seen from the table, the correlation between % of pre-IPO insider ownership and crisis is missing. After examining the data, I find that there is no % of pre-IPO insider ownership available for the IPOs during the crisis years in the Thomson One equities database. Therefore, no correlation is presented there. Unfortunately I have no means to gather these data in other ways.

(15)

15

4. Hypotheses

In this section, I describe the hypotheses I formulate based on the findings described in the previous section. For each hypothesis, I first summarize the main findings I based that specific hypothesis on.

Based on previous studies, I expect that analysts will try to support the stock prices of firms during the lockup period, for the following reasons. First of all, analysts want to attract future business for their investment bank. A way to maintain their reputation is to support the stock price of a specific company until the end of the lockup period, so the insiders can sell at a high price (Field & Hanka, 2001). Second, these insiders and other large shareholders put pressure on the analysts to issue those high recommendations, so they can indeed exit at a high price (Martin, 2008). However, as Boni and Womack (2006) show, issuing overly optimistic, biased recommendations is costly for the analyst. Therefore, after the end of the lockup period I expect the analyst to issue lower, unbiased recommendations. I expect this, because at that point the pressure of the investment banks on the analysts to stimulate the stock price reduces, as insiders had the possibility to sell their shares. Based on these arguments, I formulate the following hypothesis:

Hypothesis 1: Analysts recommendations during the lockup period are significantly better than recommendations after the lockup expiration

Table 1 present summary statistics, which seem to be in line with the first hypothesis. During the lockup period, I find a mean analyst recommendation of 4.09. The mean analyst recommendation after the lockup period is 3.79. So indeed, the mean analyst recommendation during the lockup period is higher, which is in line with the first hypothesis. Moreover, during the lockup period I find a mean recommendation surprise of 0.0021954 which is higher than the mean recommendation surprise after the lockup period of -0.0079698. The methodology section describes how I test this hypothesis formally, by making use of regression analysis.

Previous findings show that both affiliated and unaffiliated analysts issue overoptimistic recommendations after and IPO. Degeorge, Derrien and Womack (2007) find that non-affiliated analysts issue overoptimistic recommendations for firms after an IPO. As Michaely and Womack (1999) show, affiliated analysts issue biased recommendations after an IPO with the purpose of increasing the financial returns for their investment bank. They show that this bias

(16)

16

is even larger for affiliated analysts. Moreover Huang and Zhang (2009) find that the bias in analyst recommendations regarding an SEO is larger for affiliated analysts. Based on these findings, I formulate the following two hypotheses:

Hypothesis 2: Analysts recommendations from affiliated analysts are significantly better than recommendations from unaffiliated analysts

Hypothesis 2.1: The drop in the level of analyst recommendations at the end of the lockup period is significantly larger for affiliated analysts than for unaffiliated analysts

As described before, analysts in general issue distorted recommendations. Nonstrategic distortion reflects the fact that analyst have too-positive expectations (Malmendier & Shanthikumar, 2014). Strategic distortion concerns the misalignment of incentives. They show that strategic distortion is a way to please company management and to trigger small investors to purchase the company stock covered by the specific security analyst. Analysts do this by issuing high recommendations in combination with lower, beatable earnings per share forecasts (Malmendier & Shanthikumar, 2014). Based on these findings, and the literature provided about analyst incentives in combination with the lockup expiration, I formulate the following two hypotheses:

Hypothesis 3: The level of strategic distortion will be significantly higher during the lockup period than after the lockup expiration

Hypothesis 3.1: The recommendation surprise will be significantly higher during the lockup period than after the lockup period

As described before, affiliated analysts in general face more pronounced incentives than unaffiliated analysts. Therefore, besides issuing higher recommendations, I also expect affiliated analysts to show a higher level of strategic distortion compared to unaffiliated analysts. I formulate the following hypothesis which I use to test whether this is the case:

Hypothesis 4: The level of strategic distortion is significantly higher for affiliated analysts

Specific groups of insiders might be more likely to receive favorable treatment from the investment banking analysts. Venture capitalist performance highly depends on the stock price at the expiration of the lockup period (Gompers & Lerner, 1998). Venture capitalists are

(17)

17

repeated players in the market for IPOs and therefore have some kind of leverage over the investment banks. Management leads the company and makes a lot of investment related decisions, including m&a activities and SEOs, which might need future assistance from investment banks (Martin, 2006). They therefore have some leverage over the investment bank as well. Based on those arguments, I formulate the following hypothesis:

Hypothesis 5: Analyst recommendations for VC-backed companies and companies with high pre-IPO management ownership are significantly higher

(18)

18

5. Methodology

This chapter describes the methodology I use to test my hypotheses. My methodology is based on the papers of Martin (2008) en Malmendier and Shanthikumar (2014), which are both closely related to my research question.

The following standard firm control variable are included in all of the regressions, both OLS and order probit regressions, I run:

- Proceeds in USD million (including overallotment options);

- Exchange were the firm was listed on (NYSE, AMEX or NASDAQ); - Underwriter rank score;

- Lead manager dummy;

- Number of underwriters involved in the IPO underwriting syndicate; - Venture capital backed dummy;

Unfortunately, for the firms in my sample, Thomson One does not provide data regarding insider ownership pre-IPO for the years during the financial crisis. Therefore, I am not able to include both a crisis dummy and the pre-IPO ownership variable together in one regression model. This would lead to perfect multicollinearity problems. In the regressions I include a crisis dummy, I therefore exclude the pre-IPO insiders ownership variable.

I make a distinction between recommendations and earnings forecasts issued during the lockup period and after the lockup period. This leads to two subsamples: during the lockup period and after the lockup expiration. The period “after lockup expiration” includes all recommendations and earnings forecasts made during the 60 calendar days following the lockup expiration date. I use a 60-day period in order to measure the shift in analyst behavior directly after the lockup period expiration. The 60 calendar days provide the analysts with enough time to change their expectations and corresponding recommendations and forecasts. In the robustness checks section, I will recalculate my results whereby I relax my 60-days assumption and use a 30 and 50 day time period as well.

Malmendier and Shanthikumar (2014) construct a measure, the “two-tongues metric”, which captures the strategic distortion part in analyst behavior. In my paper, I use this same measure for strategic distortion. Since my focus is on the change in analyst behavior directly following the lockup period, I use quarterly earnings forecast data instead of yearly forecast data. The reason for doing this is that quarterly earnings forecasts are issued and updated more frequently and thus provide a better understanding of the change in analyst behavior around the lockup period expiration.

(19)

19

In order to test the first hypothesis, I run an ordered probit regression in which analyst recommendation is the dependent variable. A lockup expiration dummy variable is included, which is equal to one for recommendations issued after the expiration of the lockup period and zero for recommendations issued during the lockup period. Furthermore, standard firm control variables are included in this regression.

Pr(recj = i) = Pr(Ki-l < β1lockupexpirationj + ∑𝑛_𝑙=2 β1 control_variables + uj < Ki (1)

Reci (1,2,3,4,5) presents the recommendations issued by the analysts, where a higher number represents a better recommendation, and ui represents the normally distributed error term. The lead manager dummy variable, included as standard control variable, gives me insight in the impact of affiliation on the recommendation issued.

Using probit regression makes the interpretation of the regression coefficients different compared to standard OLS-regression. You cannot simply look at the coefficients in the regression output to see the effect of that variable on probability of your dependent variable. However, in general the following holds: a positive coefficient indicates a positive effect on the probability of your dependent variable, while a negative coefficient indicates a negative effect on the probability of your dependent variable. Therefore, based on my first hypothesis, I expect the lockup expiration dummy variable to be negative and significantly different from zero.

In order to test the third hypothesis, I make use of a normal OLS regression with the two tongues metric as dependent variable. In this case I can make use of a normal standard OLS-regression because of the rescaling in the two tongues metric (Malmendier & Shanthikumar, 2014). In order to construct the metric both the ordinal recommendations data and continuous numerical earnings forecast data are combined and subsequently transformed into a continuous numerical variable. A lockup expiration dummy variable is included. This variable is equal to one if both the recommendation and the forecast are issued after the expiration of the lockup period and zero otherwise. Furthermore, standard firm control variables are included in this regression.

(20)

20

The lead manager dummy provides me insight in the impact of analyst affiliation on the level of strategic distortion displayed by the analyst. Adding an interaction term “leadmanager x lockupexpiration” provides me insight on the impact of affiliated analyst issuing recommendations and forecasts after the end of the lockup period.

In order to test whether companies with high management ownership and VC-backed firms are more likely to receive high recommendations, I run the following ordered probit regression, in which analyst recommendation is the dependent variable:

Pr(recj = i) = Pr(Ki-l < β1lockupexpirationj + β2 vcbackedxleadmanagerxlockupexpriation + β3 highinsiderownershipxleadmanagerxlockupexpiration + β4 vcbackedxlockupexpiration + ∑𝑛

𝑙=2 β5 control_variables + uj < Ki (3)

In this regression, i (1,2,3,4,5) presents the recommendation issued by an analyst and the error term is normally distributed. As dependent variable, I now include an interaction dummy variable if the IPO was backed by a venture capitalist. To test the impact of high management ownership on the recommendations received, I divide my sample into quartiles. Just like Martin (2008), I base these quartiles on the percentage of pre-IPO management ownership, obtained from the Thomson One equities database. I define the highest quartile as “high management ownership” and interact this variable with both expiration of the lockup period and analyst affiliation. Based on my hypothesis, I expect both the VC-backed dummy coefficient and the high management ownership coefficient to be positive and significantly differently different from zero.

In order to investigate the impact the lockup expiration has on the recommendation surprise, I run the following OLS regression:

Recommendation surprise = β0 + β1 lockupexpiration + standard firm controls + ui (4)

Here recommendation surprise is the dependent variable. Based on my hypotheses presented in the previous section. I expect the coefficient of lockup expiration to be positive and significant.

In the appendix I include a list of all the variables used in the regressions with their definitions. This is in order to make the results easier to read and to understand what each variable measures.

(21)

21

6. Results

In this section I present my results, thereby following the hypotheses presented in section 4. First I present the results related to the impact of lockup expiration on the recommendations issued by analysts. Hereafter I present the results related to the impact on recommendation optimism. Finally I present the results related to the impact of the lockup expiration on the level of strategic distortion, based on the two-tongues metric developed by Malmendier and Shanthikumar (2014).

6.1 Recommendation level and lockup expiration

The table below presents an overview of analyst recommendations, forecasts and the level of strategic distortion during and after the lockup period for my entire sample. For now, I focus on the recommendations part of this table.

INSERT TABLE 3 HERE

In total, my sample includes 6251 recommendations during and after the lockup period. A large part of these recommendations (79%) were issued during the lockup period. The mean recommendation during the lockup period is 4.09, higher than the mean recommendation of 3.79 I find for the 50 day period after the lockup expiration. This is in line with my first hypothesis:

Hypothesis 1: Analysts recommendations during the lockup period are significantly better than recommendations after the lockup expiration

In order to get more insight in the recommendations issued by security analysts, I group the recommendations by type of analyst affiliation. The table below presents an overview of the recommendations issued, grouped by type of analyst affiliation.

INSERT TABLE 4 HERE

As can be seen from table 4, a large part of the total recommendations is issued by unaffiliated analysts. The mean recommendation issued for affiliated analysts, labeled as lead managers, of

(22)

22

4.13 is higher than the mean recommendation issued by unaffiliated analysts (4.02). This is in line with my hypotheses, namely that affiliated analysts in general issue higher recommendations. These results are also in line with prior research by Degeorge, Derrien and Womack (2007) and Michaely and Womack (1999). These researchers also find a higher bias in the recommendations issued by affiliated analysts, as compared to unaffiliated analysts.

Both affiliated and unaffiliated analysts, on average, issue higher recommendations during the lockup period than after the expiration of the lockup period. For affiliated analysts the mean recommendation issued during the lockup period is 4.14 compared to 4.06 after the lockup period expiration. For unaffiliated analysts, the mean recommendation issued during the lockup period is 4.08 compared to 3.78 after the expiration of the lockup period. Moreover, for both sub-periods I find higher recommendations for the affiliated analysts than for the unaffiliated analysts.

Remarkable to see here is the relatively small drop in the level of recommendations for the affiliated analysts compared to the drop for unaffiliated analysts. Actually, the difference between the mean recommendation level between affiliated and unaffiliated analysts increases after the lockup expiration (from 0.06 to 0.28).

To formally test the first two hypotheses, I run an ordered probit regression, as presented in table 5.

INSERT TABLE 5 HERE

Remarkable to see is the low R-squared for all three regression models. However, one has to keep in mind that it’s very hard to find a precise model for analyst recommendations. Looking at existing literature in this field, similar regressions as presented by Martin (2006) show comparable R-squared levels. Therefore, I do not see this as a problem for the strength of my results. By including more control variables, the adjusted R-squared of the regression models increases slightly.

The variable of interest here is the lockup expiration dummy. As can be seen in the table, in all three regression models I run the regression coefficient of this lockup expiration dummy is negative and significantly different from zero at the 1% (for model 1 and 2) and 5% level (model 3). This is in line with my first hypothesis, namely that analysts will revise their recommendations downwards as soon as the lockup period expires. The probability of receiving

(23)

23

a strong buy recommendation after the lockup expiration decreases by respectively 13.76% (model 1), 13.26% (model 2) and 12.55% (model 3).

The above findings are very similar to prior research. Jens Martin (2006) even finds a drop of 31% in receiving a strong buy recommendation after the expiration of the lockup period. Michaely and Womack (1999) argue that analysts deviate from their natural role as providers of unbiased information about investment decisions. This is exactly what I find here, and I even prove that this lasts till the expiration of the lockup period.

The lead manager dummy is positive, indicating that analysts affiliated with the lead manager indeed issue higher recommendations. However, what is remarkable to see is that in two of the three models presented the interaction dummy between lead manager and lockup expiration is positive, although not significant. I expected this coefficient to be negative, based on existing literature in this field. Martin (2006) actually finds the opposite in his research. He finds a significant decrease in the recommendation level for affiliated analysts after the lockup expiration. However, I have to note that he uses the second regression model presented here in his research, in which I find a negative coefficient as well.

Based on existing literature in this field, I am not able to form a good explanation for my findings. It might be the case that earnings announcements follow closely to the expiration of the lockup period, for a relatively large part of the firms in my sample. Positive earnings announcements give analysts (especially affiliated analysts) reason to issue high recommendations. However, I have not the means to formally check whether this is the case and leave this open as an opportunity for future research.

6.2 Strategic distortion around lockup expiration

In this section I will look at the impact of the lockup expiration on the level of strategic distortion displayed by security analysts. As mentioned before, I use the two-tongues metric in order to measure the level of strategic distortion.

The two-tongues metric, as defined by Malmendier and Shantikumar (2014), measures the difference between an analysts’ recommendation optimism and the scaled forecast optimism, for a specific firm the analyst is covering. Figure 1 shows the construction of this two-tongues metric. The graph at the top shows the distribution of the recommendation optimism. Recommendation optimism is defined as the difference between a recommendation and the consensus recommendation as of that day. The consensus is calculated as the average

(24)

24

recommendation over a one month period. Recommendations are on a 1 to 5 scale. As can be seen from the first graph in figure 1, most of the observations lie between -2 and +2.

The second graph displays the earnings forecast optimism. Earnings forecast optimism is defined as the difference between the earnings per share forecast and the consensus earnings forecast as of that day, normalized by the share price and multiplied by 100. Again the consensus is calculated over a one month period. Normalization takes place to make the magnitude of the earnings surprise comparable to the recommendation surprises.

The bottom graph in figure 1 shows the distribution of the two-tongues metric. This metric is defined as recommendation optimism minus earnings forecast optimism. Again, most of the observations lie between -2 and +2. All of these distribution related findings are very comparable to the findings of Malmendier and Shanthikumar (2014).

As can be seen from table 3. the mean two-tongues metric is higher during the lockup period than after the lockup period. This is in line with my hypotheses. Recommendations are higher during the lockup period. However, for the earnings announcements this does not hold. As explained by Malmendier and Shanthikumar (2014), analysts issue upward biased recommendations to trigger small investor purchases. However, in their quarterly earnings per share forecasts analysts are more realistic. Therefore, these forecasts do not change because of the lockup period expiration.

In order to test the impact of the expiration of the lockup period on the level of strategic distortion, I run an OLS regression with the two-tongues metric as dependent variable. In table 6 I present the results of this regression. Again, I run three different regression models, to control for certain time periods.

In all three regression models, I find negative coefficients for the lockup expiration variable, although not significant. This indicates that the level of strategic distortion indeed slightly decreases after the expiration of the lockup period. This also is in line with the prior research conducted by Malmendier and Shanthikumar (2014). They argue that strategic distortion results in small investor purchases and thus is positive for the company. This is even more important during the lockup period.

(25)

25

Remarkable to see again is the positive and very significant (at the 1% level) of the coefficient on the lead manager x lockup expiration dummy. This indicated that affiliated analysts show larger strategic in their behavior after the lockup period than during the lockup period, which is exactly the opposite as what I expected. However, this is in line with the findings I present regarding the level of affiliated analysts’ recommendations.

The two-tongues metric consists of various components. As explained by Malmendier and Shanthikumar (2014), most of the strategic distortion is reflected in the recommendation surprise. Therefore, I examine the impact of the lockup expiration on the recommendation surprise level more closely. Table 7. presents two ordered probit regression models in which the recommendation surprise variable is the dependent variable.

Again I find a negative coefficient for the lockup expiration dummy, although not significant. Besides this, I again find a positive and very significant coefficient for the lockup expiration x lead manager dummy. This actually means that the level of strategic distortion increases for affiliated analysts after the lockup period. Based on the existing literature, I cannot find an explanation for this. Again, it might be the case that positive earnings announcements follow closely to the lockup expiration. Malmendier and Shanthikumar (2014) argue that affiliated analysts tend to adjust their earnings per share forecasts downward prior to earnings announcements in order to please company management. Thereby, they increase their recommendations at the same time, leading to an increase in the strategic distortion metric. However, I have not the means to check whether this is the case and leave this open for future research.

(26)

26 6.3 Specific groups of insiders

In this section I present the results regarding the recommendation levels for venture capital backed firm and firms with high inside ownership. I run an ordered probit regression in which I look at the impact of the lockup expiration for those two groups in specific.

Table 8. presents the results from this regression. I indeed find preferred treatment for firms with high insider ownership, indicated by the significant downward revision of analyst recommendations at the expiration of the lockup period. This is the case for both affiliated and unaffiliated analysts. These findings demonstrate that companies with high insider ownership actually profit from the leverage they have over the investment banks, which is in line with my hypotheses.

For venture capital backed IPOs, I find mixed results. For unaffiliated analysts I indeed find a slight decrease in the level of recommendations after the lockup expiration. However, this decrease is not significant. For affiliated analysts I even find a positive coefficient, indicating that they increase their recommendations after the lockup expiration for this subgroup. However, again this coefficient is not significant. These findings indicate that venture capital backed companies don’t profit from the leverage they have over the investment banks. This provides new insights compared to the prior work of Jens Martin (2006) who actually finds that these firms do profit from this leverage. However, this trend may have changed over time.

(27)

27

7. Robustness checks

In this section I provide robustness checks in order to increase the validity of my results. The first robustness check focuses on the potential 180 days after the IPO effect, which might explain the downward revision in analyst recommendations. In the second check, I examine the level of strategic distortion during a symmetrical -60 to +60 day around the expiration of the lockup period. Ordered probit and OLS regressions are used in order to execute the robustness

checks.

Most of the companies in my sample (approximately 80%) have a lockup period of 180 days. Therefore, one might argue that this specific time period causes the change in analyst behavior, and not the expiration of the lockup period. In order to check for this, I include a 180 day dummy variable in my regression. This dummy variable is one if the recommendation has been issued more than 180 days after the IPO. Table 9 shows the results of this ordered probit regression.

INSERT TABLE 9 HERE

As can be seen, the lockup expiration dummy remains very significant, even at the 1% level. This indicates that the lockup expiration is of significant importance regarding the change in analyst recommendations. However, the 180 days after IPO dummy is not significant. Therefore, I conclude that the change in analyst recommendations is not caused by the 180 days after the IPO effect. The expiration of the lockup period indeed is the explanation for this downward revision in the recommendations of analysts.

The second robustness check I present here is related to the change in strategic distortion around the expiration of the lockup period. The two-tongues metric for strategic distortion (Malmendier and Shanthikumar, 2014) includes consensus levels for both recommendations and earnings forecasts. As explained before, I calculate consensus as average over a one month period going backwards from the issue date. However, in the first days following an IPO very little recommendations and forecasts are issued. Therefore, consensuses are based on very little observations, making them relatively volatile.

In order to overcome this issue, I investigate the change in strategic distortion for a symmetrical -60 days to +60 days around the expiration of the lockup period. Important to note here is that data for the complete period, ranging from IPO to 60 days after the expiration of the

(28)

28

lockup period, is used to calculate consensuses. The results of the regressions I run for this specific period are presented in the table below.

INSERT TABLE 10 HERE

Using the alternative symmetrical +60 to -60 days period around lockup expiration, I find similar results as before. Again, the lockup expiration dummy variable coefficient is negative and significant. This indicates that there is a decline in the level of analyst recommendations.

Again, the lockup expirationxleadmanager dummy variable coefficient is positive and significant. This indicates an increase in the level of recommendations issued by analyst affiliated with the lead manager after the expiration of the lockup period.

These results show that my prior findings are not caused by the volatile and small recommendations issued in the first 60 days after an IPO.

(29)

29

8. Conclusion

In this paper, I examine the behavior of analysts around the expiration of the IPO lockup period. Where prior research papers mainly focus on analyst behavior in terms of IPO underpricing, I examine analyst behavior during the entire lockup period, until the expiration of the lockup period.

My sample consists of firms going public on the NYSE, NASDAQ or AMEX during the years 1996 through 2014. Hereby, I make a distinction between the two ways in which analysts can provide information: recommendations and earnings (per share) forecasts. I combine the two in the level of strategic distortion metric, measured with the two-tongues metric (Malmendier and Shanthikumar, 2014).

Using ordered probit regressions, I find that both affiliated and unaffiliated analysts issue upward biased recommendations during the lockup period. For firms with high pre-IPO insider ownership, I find the bias to be even larger. Those results are in line with the prior findings of Martin (2008). I expected the same to be the case for VC-backed IPOs. However, my results show that they do not profit from the leverage they have over the investment banks. The recommendations issued for VC-backed IPO firms are not significantly higher compared to other firms in my sample.

Analysts are subject to pressure to issue upward biased recommendations. Based on prior literature, I argue that this pressure comes from pressure issued by insiders. These insiders put pressure on the analysts to stimulate the stock price till the end of the lockup period, since that is their first real opportunity to sell their shares. After the end of the lockup period I find a significant downward revision in the level of recommendations issued by analysts. However, for analysts affiliated with the lead manager I actually find an upward revision in the recommendations they issue compared to the recommendations they issued during the lockup period. This might be due to negative earnings announcements following closely to the expiration of the lockup period, although I do not have the means to check for this.

In line with prior research by Malmendier and Shanthikumar (2014), I find no bias in the earnings per share forecasts during the lockup period. Analysts issue biased recommendations, but in order to remain the trust from investors, issue realistic earnings per share forecasts.

Although the bias in stock recommendations is statistically significant higher during the lockup period, the same does not hold for the level of strategic distortion. I even find that analysts affiliated with the lead manager show a statically significant higher level of strategic

(30)

30

distortion in their behavior after the lockup period expiration. The same does hold for the recommendation surprise variable.

My findings are robust for two robustness checks. I find that the 180-days after the IPO effect is not causing the change in analyst behavior. Moreover, I control for the fact that very little recommendations are issued closely following the IPO, by examining a symmetrical -60 to +60 days period around the expiration of the lockup period.

The behavior I find for affiliated analysts is not what I expected based on prior literature. As suggested, this could be due to earnings announcements close to the expiration of the lockup period. However, I have not the means to formally check for this. Therefore, I leave this open as an opportunity for future research.

(31)

31

References

Aggarwal, R. K., Krigman, L., & Womack, K. L. (2002). Strategic IPO underpricing, information momentum, and lockup expiration selling. Journal of financial economics, 66(1), 105-137.

Boni, L., & Womack, K. (2002). Solving the sell-side research problem: Insights from buy-side professionals. Working Paper, University of New Mexico and University of Toronto Rotman School of Management.

Boni, L., & Womack, K. L. (2006). Analysts, industries, and price momentum. Journal of Financial and Quantitative Analysis, 41(01), 85-109.

Bradley, D. J., Jordan, B. D., Yi, H. C., & Roten, I. C. (2001). Venture capital and IPO lockup expiration: An empirical analysis. Journal of Financial Research, 24(4), 465-493.

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.

Brau, J. C., & Fawcett, S. E. (2006). Initial public offerings: An analysis of theory and practice. The Journal of Finance, 61(1), 399-436.

Brav, A., & Gompers, P. A. (2003). The role of lockups in initial public offerings. Review of Financial Studies, 16(1), 1-29.

Chen, Hsuan-Chi, and Jay R. Ritter. "The seven percent solution." Journal of finance (2000): 1105-1131.

Degeorge, F., Derrien, F., & Womack, K. L. (2007). Analyst hype in IPOs: Explaining the popularity of bookbuilding. Review of Financial Studies, 20(4), 1021-1058.

Field, Laura Casares, and Gordon Hanka. "The expiration of IPO share lockups." The Journal of Finance 56.2 (2001): 471-500.

(32)

32

Gompers, P., & Lerner, J. (1998). Venture Capital Distributions: Short‐Run and Long‐Run Reactions. The Journal of Finance, 53(6), 2161-2183.

Keasler, T. R. (2001). The underwriter’s early lock-up release: Empirical evidence. Journal of

Economics and Finance, 25(2), 214-228.

Kelly, B. T., & Ljungqvist, A. (2007). The value of research.

Lin, H. W., & McNichols, M. F. (1998). Underwriting relationships, analysts' earnings forecasts and investment recommendations. Journal of Accounting and Economics, 25(1), 101- 127.

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

Loh, R. K., & Mian, G. M. (2006). Do accurate earnings forecasts facilitate superior investment recommendations?. Journal of Financial Economics, 80(2), 455-483.

Loughran, Tim, and Jay R. Ritter. "Why has IPO underpricing changed over time?." (2003).

Malmendier, U., & Shanthikumar, D. (2007). Are small investors naive about incentives?. Journal of Financial Economics, 85(2), 457-489.

Malmendier, U., & Shanthikumar, D. (2014). Do security analysts speak in two tongues?. Review of Financial Studies, 27(5), 1287-1322.

Martin, Jens. "Prop ups during lockups." AFFI/EUROFIDAI, Paris December 2008 Finance International Meeting AFFI-EUROFIDAI. 2011.

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

Michaely, R., & Womack, K. L. (2005). Brokerage recommendations: Stylized characteristics, market responses, and biases. Advances in Behavioral Finance II, 389-422.

(33)

33

Ritter, J. R., & Zhang, D. (2007). Affiliated mutual funds and the allocation of initial public offerings. Journal of Financial Economics, 86(2), 337-368.

Reuters: http://www.reuters.com/article/2015/03/18/us-alibaba-group-stocks idUSKBN0ME18R20150318#DMjDvpMsS046WT8v.97

Teoh, S. H., & Wong, T. J. (2002). Why New Issues and High‐Accrual Firms Underperform: The Role of Analysts' Credulity. Review of Financial Studies,15(3), 869-900.

(34)

34

Figures and Tables

Figure 1. Two-tongues metric

These graphs illustrate the construction of the two-tongues metric for strategic distortion, which is equal to recommendation optimism minus scaled forecast optimism. The graph at the top left shows the distribution of the recommendation optimism variable, calculated as recommendation minus consensus recommendation, which is equal to the mean recommendation over a one month period, as of that day (Malmendier and Shantikumar, 2014). The figure at the top right shows the distribution of the winsorized earning surprise variable. I winsorized the variable at the 1% and 99% level. Earnings surprise is defined as earnings forecast minus the consensus earnings forecasts as of the day, where the consensus is calculated as mean forecast over a one month period as of that day. This earnings surprise is then normalized by the stock price as of that day and multiplied by one hundred. The graph at the bottom shows the distribution of the two-tongues metric of strategic distortion. I winsorized this metric at the 1% and 99% level.

0 .5 1 1 .5 2 D e n si ty -4 -2 0 2 4 twotonguesmetric

Recommendation Optimism Minus Earnings Forecast Optimism

0 1 2 3 4 D e n si ty -3 -2 -1 0 1 2 recsurprise Recommendation Optimism 0 .5 1 1 .5 2 2 .5 D e n si ty -4 -2 0 2 4 epssurprise

(35)

35

Table 1. Descriptive sample statistics

The sample consists of firms conducting an IPO between 1996 and 2014, and were subsequently listed on the NYSE, NASDAQ or AMEX. I exclude IPOs by closed-end funds, American depository receipts (ADRs), real estate investment trusts (REITs), unit offerings, financial companies and utilities. Moreover, I also drop all offerings for which the offer price is below $5 as well as firms for which information regarding the lockup period is not available. Proceeds are shown in USD million and include overallotment options. The length of the lockup period is measured in days. The percentage of insider ownership shows the percentage of the company that was owned by management, pre-IPO. The VC backed variable indicates whether the IPO was backed by a venture capitalist.

Obs Mean Median Minimum Maximum

Proceeds 885 110 60 8 16007

Lenght of lockup

period 961 173 180 30 730

VC backed 475

(36)

36

Table 2. Correlation matrix of the main variables included in the regression models

This table presents correlations between the main variables included in the ordered probit- and OLS regressions. A definition of each of the variables included in the table can be found in the appendix, at the end of this paper.

Correlation table analyst

recommendations

uw rank score lockup expiration NYSE NASDAQ lead manager

dummy

number of underwriters

proceeds vc backed bubble crisis

analyst recommedations 1.000

uw rank score -0.0492 1.000

lockup expiration -0.1468 -0.0321 1.000

NYSE -0.1107 0.1249 0.0330 1.000

NASDAQ 0.0914 -0.1178 -0.0313 -0.9448 1.000

lead manager dummy 0.0425 0.1672 -0.0864 0.0149 -0.0162 1.000

number of underwriters -0.1359 0.0129 -0.0185 0.1615 -0.1457 -0.0603 1.000

proceeds -0.0695 -0.0262 -0.0205 -0.0018 0.0054 0.0345 0.9089 1.000

vc backed 0.0573 -0.0373 -0.0484 -0.1907 0.2133 -0.0062 0.0214 0.0846 1.000

bubble 0.1797 -0.0076 -0.0715 -0.1799 0.1899 -0.0462 -0.1422 -0.0809 0.1995 1.000

(37)

37

Table 3. Overview of recommendations and forecasts during and after the lockup period

This table presents an overview of the recommendations and forecasts during and after the lockup period. Recommendation and forecast data is obtained from the IBES database. I include recommendations issued and forecasts from the moment of the IPO until 60 days after the lockup expiration. The sample consists of firms conducting an IPO between 1996 and 2014, and were subsequently listed on the NYSE, NASDAQ or AMEX. I exclude IPOs by closed-end funds, American depository receipts (ADRs), real estate investment trusts (REITs), unit offerings, financial companies and utilities. Moreover, I also drop all offerings for which the offer price is below $5 as well as firms for which information regarding the lockup period is not available.

During lockup period After lockup period

All analysts 4936 1315

Mean recommendation 4.09 3.79

Mean recsurprise 0.0021954 -0.0079698

Mean EPS forecast

surprise -0.0134718 -0.0351052

All analysts 5140 1111

Mean two tongues

(38)

38

Table 4. Overview of recommendations and forecasts during and after the lockup period grouped by analyst affiliation

This table presents an overview of forecasts and recommendations grouped by analyst affiliation. Recommendation and forecast data is obtained from the IBES database. I include recommendations and forecasts issued from the moment of the IPO until 60 days after the lockup expiration. The sample consists of firms conducting an IPO between 1996 and 2014, and were subsequently listed on the NYSE, NASDAQ or AMEX. I exclude IPOs by closed-end funds, American depository receipts (ADRs), real estate investment trusts (REITs), unit offerings, financial companies and utilities. Moreover, I also drop all offerings for which the offer price is below $5 as well as firms for which information regarding the lockup period is not available.

During lockup After lockup Total Lead manager

Observations 593 72 665

Mean recommendation 4.14 4.06 4.13

Mean EPS forecast -0.11

Unaffiliated analysts

Observations 4343 1243 5586

Mean recommendation 4.08 3.78 4.02

(39)

39

Table 5. Order probit regression models examining the change in analyst recommendations at the expiration of the lockup period

This table presents an ordered probit regression with analyst recommendation as dependent variable. Model 1 is contains standard firm control variables. Model 2 additionally controls for the bubble period. Model 3 controls for the bubble period as well as for the financial crisis. Lockup expiration is a dummy variable equaling one if the recommendation was issued after the lockup period. Lockup expiration x lead manager is an interaction dummy variable, equaling one if the recommendation was made by an analyst affiliated with the lead manager and if the recommendation was issued after the lockup period. Underwriter rank scores are determined based on the underwriter ranking of Ritter (2003). Lead manager is a dummy variable equaling one if the recommendation was issued by an analyst affiliated with the lead manager of the underwriting syndicate. Proceeds are measured in USD million and include overallotment options. Number of underwriter measures the total number of underwriters (including the lead-managers). VC backed is a dummy variable equaling one if the IPO was backed by a venture capitalist. % of insider ownership measures the percentage of the company that was owned by the management, pre-IPO. Bubble period is a dummy variable equaling one for the years 1999 – 2000. Crisis is a dummy variable equaling one for the years 2006-2007. Recommendation data is obtained from the IBES database. Include recommendations issued from the moment of the IPO until 60 days after the lockup expiration. The sample consists of firms conducting an IPO between 1996 and 2014, and were subsequently listed on the NYSE, NASDAQ or AMEX. I exclude IPOs by closed-end funds, American depository receipts (ADRs), real estate investment trusts (REITs), unit offerings, financial companies and utilities. Moreover, I also drop all offerings for which the offer price is below $5 as well as firms for which information regarding the lockup period is not available. Standard errors are shown between brackets, under the coefficients.

Analyst recommendations and strategic distortion around lockup period expiration