How does investment banking and analyst association affect stock ratings? : evidence from M&A transactions

(1)

How does investment banking and

analyst association affect stock

ratings?

Evidence from M&A transactions

MSc Thesis Finance

Amsterdam Business School Track: Asset Management Supervisor: Dr. Jan Lemmen

Steve Nugteren, 10466843 July 2017

(2)

2

Statement of Originality

This document is written by Steve Nugteren, 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 reference 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

Financial analysts want to build their reputation for providing accurate stock ratings. Simultaneously, analysts face incentives or strategic motives to publish biased stock ratings. The bias by affiliated sell-side analysts has been extensively studied in previous literature with only marginal research being directed towards the affiliation by analysts from M&A activity. This paper shows that analysts affiliated with M&A advisors are influenced in their stock rating issuance, while trying to exclude selection bias. Specifically, financial analysts publish more bullish stock ratings on the acquirer (bidder) if they advise the acquiring firm, before as well as after exchange ratios are negotiated. Furthermore, analysts advising the target-firm publish more bullish stock ratings on target than non-affiliated analysts. Finally, this study finds that the odds of deal success raises by the issuance of optimistically biased stock ratings. Consequently, it is a relevant incentive that analysts face during equity research. Keywords: Financial analysts, investment banks, equity research, stock ratings/recommendations, mergers and acquisitions, optimistically biased, advisory services, analyst affiliation, incentives

(4)

4

ACKNOWLEDGEMENT

I would like to thank my thesis supervisor dr. Lemmen for his outstanding supervision. He consistently allowed this research to be my own work, while steering me in the right direction whenever needed.

(5)

5

List of tables

Table 1: Theoretical framework. ... 17

Table 2: Overview of the hypotheses ... 21

Table 3: Distribution of the five ratings categories ... 28

Table 4: M&A deal value (in millions) ... 28

Table 5: Method of payment ... 29

Table 6: Status of the deal ... 29

Table 7: Summary statistics: Acquirer-advising analysts ... 32

Table 8: Results acquirer-advising analysts, covering both acquirer and target firm: Before the exchange ratios are negotiated ... 34

Table 9: Results acquirer-advising analysts, covering both acquirer and target firm: After the exchange ratios are negotiated ... 36

Table 10: Robustness check for specifications L2, L6 and L8: Acquirer-advising analysts ... 37

Table 11: Summary statistics: Target-advising analysts ... 38

Table 12: Results target-advising analysts, covering both acquirer and target firm: Before the exchange ratios are negotiated ... 39

Table 13: Results target-advising analysts, covering both acquirer and target firm: After the exchange ratios are negotiated ... 40

Table 14: Robustness check for specifications L10 and L14: Target advising analysts ... 41

Table 15: Deal success results, reporting on the acquirer: Affiliated analysts ... 43

Table 16: Deal succes results, reporting on the target: Affiliated analysts ... 45

Table A: Term definitions ... 53

Table B: Shapiro-Wilcoxon test ... 53

(7)

7

1 Introduction

The actions of financial analysts have long been followed with caution by researchers as well as practitioners. Most research on the matter state that the information from these analysts are beneficial to the market efficiency, enabling investors to valuate firms more accurately (Jegadeesh, Kim, Krische & Lee, 2004). Merkley, Michaely and Pecelli (2017) demonstrate that having more financial analysts improves information quality, which can lead to more efficient dissemination of information. The lion’s share of the analysts work for investment banks that have several sources of profit. Investment banks generally act as equity underwriters, lenders, asset managers and provide equity research (Haushalter & Lowry, 2010). Their actions could influence the profits of other departments. As a result, the judgement of analysts could be affected. For instance, Womack (1996) documents that analysts could avoid giving a negative rating about firms which they cover, because doing so might harm the relationship with the investment banking department. Additionally, management could reduce or even stop the flow of information as a response to the issuance of unfavorable ratings. The heart of the conflict is the view that financial analysts provide bullish research coverage to gain favor with the existing clients of their firm or to win future investment banking business from covered firms (Corwin, Larocque & Stegemoller, 2017).

This is highly relevant, since analysts exert a significant force on the current stock market. The stock rating (recommendation) of an all-star analyst can temporarily cause the share price of a company to fluctuate upwards and downwards, regardless if the fundamentals of the firm changed (SEC, ‘’Investor Publications’’, 2010). Not only investors are influenced by the ratings of analysts. Corporates value sell-side equity research to sell their securities and raise the liquidity of stocks, with the majority of the leading investment banks spending over $100 million dollars on equity and market research per year (Groysberg, Healy & Maber, 2011).

In April 2003, ten leading firms of Wall Street entered into the the Global Settlement with federal regulators on the concern of conflict of interests confronted by financial analysts (Agrawal and Chan, 2008). These firms were obligated to fulfill a payment of $1.4 billion as a result of the accuses that analysts misled investors with bullish stock ratings to gain favor with other potential investment banks (Clarke, Khorana, Patel & Rau, 2011). Moreover, the Global Settlement prohibited financial analysts from going on a road show with investment bankers (SEC, “Fact Sheet Global Settlement”, 2003). The prestige of financial analysts from the sell-side, in particular those at prominent investment banks, has taken serious damage in the last

(8)

8

two decades. Optimistically biased research was issued by biased financial analysts and possibly helped attracting new investment banking clients.

In an attempt to verify whether the optimistic stock ratings were a widespread phenomenon, numerous studies have researched the activities of investment banks that might lead analysts towards biasing their stock ratings. One stream in literature suggests that equity underwriting relations lead analysts to provide more favorable stock ratings compared to analysts that are not associated (Lin & McNichols, 1998). Malmendier and Shanthikumar (2014) state that analysts who are employed by or are under the influence of investment banking pressure can have strategic motives to publish more bullish ratings or similarly, more bearish ratings. According to Hong and Kubik (2003), brokerages have rewarded optimistic analysts who publish relatively bullish ratings compared to the overall consensus with more favorable job separations. Additionally, they state that job separations depend less on accuracy and more on bullish expectations. This matters to investors, since Barber, Lehavy and Trueman (2007) show results where the average daily abnormal return generated through buying all bullish ratings of independent research firms tops that of investment banks and brokerages significantly. The findings of all the literature do not consistently reach the same conclusion, creating a theoretical puzzle. Alternative research has reported the absence of the accused biasedness of stock ratings, published by biased-analysts. Particularly, the paper of Cowen, Groysberg and Healy (2006) fails to find association between optimistically or pessimistically biased stock ratings and affiliated analysts. In fact, Jacob, Rock and Weber (2003) indicate that investment banks provide the superior forecasts, despite financial analysts facing incentives to optimistically bias earnings predictions.

In conclusion, on the one hand financial analysts face strategic motives to issue biased stock ratings. On the other hand, analysts are incentivized to build a reputation by providing accurate ratings. Consequently, self-interested analysts have to weigh future career plans against any short-term rewards from favors gained from cooperating with investment banks (Ljungqvist, Marston, Starks, Wei & Yan, 2007).

Although there are many incentives or strategic motives, this paper hypothesizes that one of the biggest incentives arises from M&A affiliation. The core business of investment banks includes supporting acquiring and target firms in mergers and acquisitions transactions. The fees generated through M&A advisory services provide the largest part of the revenue for investment banks. Kolasinski and Kothari (2008) find that analysts face conflicted interests as a result of M&A activity. These conflicts could influence analysts, resulting in biased stock ratings. This field had previously not been examined in terms of the analyst conflict. Their

(9)

9

paper showed that analysts employed by investment banks for M&A services upgrade stock ratings more often than the non-affiliated. As a result, the stock price of the acquiring firm could be appreciated and benefit the acquirer (bidder). Moreover, analysts advising the targets upgrade ratings on target firms, before as well as after the ratios are negotiated. Finally, they upgrade the ratings on the acquiring firms after the exchange ratios are negotiated. The paper concludes that M&A affiliation causes a bias in stock ratings.

However, the regulatory landscape has altered since the implementation of the Global Settlement and the disclosure of requirements of NASD Rule 2711 and 472. The Global Settlement strengthened the requirements stating that investment banking departments and financial analysts are divided by Chinese walls (Haushalter & Lowry, 2010). The results of Kadan, Madureira, Wang and Zach (2009) indicate that the new regulation had a significant influence. Their paper reports that after the change of regulation, bullish ratings have become less frequent and more informative. According to Bradshaw (2011), literature finds itself at an interesting juncture of time with a relatively new trend towards the payment for analyst coverage. The regulation that followed brought the sell-side industry under scrutiny and should have reduced the incentives that financial analysts face.

The research will test if an association between M&A affiliation and the issuance of stock ratings is still a widespread phenomenon. Therefore, the paper merges a sample of stock ratings from financial analysts with a sample of M&A transactions in 2010-2016. To give a concise answer to the research question, multiple hypotheses are formulated. First, the paper investigates whether affiliated analysts are more bullish or bearish relative to consensus in order to gain favor with the investment banks. Analysts are divided as affiliated or non-affiliated. As a follow-up, affiliated analysts are subdivided into analysts that advise the acquirer of the target. The samples have to be divided one last time, since analysts can be confronted with contrasting incentives before as well as after the exchange ratios of the M&A transactions are negotiated. Therefore, the samples are split based on the timing. Hypotheses are formulated to indicate the predicted sign of affiliation on the rating of stocks. Thereafter, the hypotheses are tested by employing the ordered logit model, probit regressions as well as linear least squares.

Specifically, one of the main incentives faced by financial analysts is investigated. The association between publishing bullish or bearish ratings by affiliated analysts on stocks and the odds of deal success is analyzed. Specifically, the influence of bullish and bearish stock ratings on deal success will be determined. This is partly in line with the paper of Becher, Cohn and Juergens (2015).

(10)

10

This study focuses on the conflicts of interests that financial analysts employed at investment banks and brokerages face. Only marginal research has been directed towards the affiliation by analysts from M&A activity. Additionally, the marginal research directed towards M&A affiliation does not always succeed at excluding selection bias. For instance, the research in a previous master thesis (Sonneveld, 2016) fails at excluding any form of selection bias. Sonneveld finds significant results, but these could easily be explained by selection bias. Consequently, the drivers of the bias are ambiguous, whereas the results can be explained through reversed causality. One way this paper tries to exclude selection bias is by retrieving stock ratings that are issued within 90 days before the announcement date of the M&A transaction. According to Kolasinski and Kothari (2008), this ensures that it is highly likely for companies to have retained the advising firms in the M&A deal. This is highly relevant, since the findings could otherwise indicate a reversed relation, i.e. firms choose M&A advisors that issue the most optimistic stock ratings on their company. Therefore, failing to eliminate selection bias does not prove the conflict of interest arising from M&A affiliation. The paper provides additional evidence on the behavior of affiliated analysts in the M&A context, in a time period after regulatory changes such as the Global Settlement, NASD 2711 and 472.

The findings are relevant for multiple reasons. The conflicts which arise by advising during an M&A transaction are assumed to be more pervasive phenomena than conflicts originating from Initial Public Offerings (IPOs) or Seasonal Equity Offerings (SEOs), considering companies are more likely to carry out an M&A transaction than issuing equity (Haushalter & Lowry, 2010). From 1994 onward, revenue generated from advising mergers and acquisitions surpassed equity underwriting fees in the United States, which has grown significantly larger over the years (Kolasinski & Kothari, 2008). Since M&A fees generate the most revenue of all investment banking divisions, conflicts of interests might arise more strongly from this department compared to divisions which generate less. Moreover, raising capital through equity are relatively rare developments for a company and often singular, opposed to M&A transactions. Aggressive firms often acquire multiple firms each year. Therefore, an opportunity exists to handle multiple acquisition deals in exchange for optimistically or pessimistically biased equity research coverage. Pressure by investment banking managers to generate M&A fees fuels a potentially considerable cause of analyst conflicts. Above all, the results have implications for investors as well as regulators. Investors should consider the potential presence of bias in sell-side reports of financial analysts. The

(11)

11

results show the bias is still present, which indicate regulators could think of new ways of reducing incentives analysts may face.

The findings of this study implicate that stock ratings issued by analysts advising the bidders are more bullish than non-affiliated analysts before and after the exchange ratios have been negotiated. Interestingly, the results are in line with previous results from the paper by Kolasinski and Kothari (2008). They find that analysts advising the acquirer publish optimistically biased ratings relative to non-affiliated analysts before the exchange ratios are negotiated. This paper finds that financial analysts associated with the target firm publishes optimistically biased ratings on targets firms before as well as after the ratios are negotiated. Interestingly, the results of this research are partly in line with the results found by the paper written by Kolasinski and Kothari (2008), which suggests that the conflicts of interest are still present after the regulatory changes. Aside from investigating the existence of analyst bias from M&A affiliation, deal success is examined as incentive for biasing stock ratings, in line with the paper of Becher et al. (2015). This is tested by examining the association between optimistically biased or pessimistically biased ratings of affiliated analysts and the odds of deal success. The findings suggest that bearish ratings fail to raise the odds of deal success. This might help explaining why this thesis fails to find any bearish bias on stock ratings. Contrarily, the findings imply that optimistically biased ratings from M&A advising analysts correspond to increased odds of deal success. Therefore, deal success is an important incentive to optimistically bias stock ratings.

This paper tries to contribute to two streams of literature. First, to increase the understanding in which ways investment banks potentially contest for this business and second it investigates whether the behavior of analysts around M&A deals are still a widespread phenomenon. Although there are numerous papers on the behavior of analysts around equity issues, relatively little research is found on the behavior of analysts around mergers and acquisitions. Second, it investigates one of the main theoretical incentives of the conflicts of interest.

This paper provides both a qualitative and quantitative study that attempt to answer the research questions. First, the theoretical framework will be sketched and the incentives that financial analysts face will be discussed in chapter 2. This chapter is followed-up by formulating hypotheses and describing the expected outcome of all research questions in chapter 3. In chapter 4, the methodology is explained. Subsequently, in chapter 5 the data is outlined and the data collection and management process is explained. Chapter 6 depicts and interpreters the output of the employed regressions on the association between stock ratings

(12)

12

and M&A relations through advisory services. Furthermore, odds of deal success are investigated as incentive to bias stock ratings. Finally, the last chapter outlines the main conclusions which can be drawn from the results and provides a discussion and suggestions for future research.

(13)

13

2 Literature review

In this paragraph, the main reasons why financial analysts can become biased through advising an acquirer or target are explained. Subsequently, the process how analysts can be tempted to taunt is described. The regulatory landscape will be sketched and the theoretical framework will be presented. As a follow up, a table will be constructed to sum up all findings.

2.1 Introduction to M&A affiliation

Investment banks generate revenue through multiple sources, with M&A advisory fees making up the largest part. Consequently, conflicts of interest could arise from different departments which are generating less revenue within the investment bank. Moreover, firms usually perform numerous M&A deals which provides the opportunity to do multiple transactions in return for bullish stock coverage. Other sources of income from investment banks, such as initial or seasonal equity offerings, are relatively uncommon developments in the life of a company. Therefore, the conflicts of interest are possibly larger for M&A transactions. This creates affiliation (association) for financial analysts. Financial analysts who provide equity research for investment banks and publish ratings for the client are compensated in the form of a fee. The compensation is often a proportion of the sum paid for the acquisition, generated from guiding the investment banks during the deal. Generally, the proportion of the sum is determined by the frequency as well as the size of the transaction. Ultimately, generating these fees is contingent upon the transaction to be successful. Therefore, analysts are incentivized to bias stock ratings considering the probability of a successful transaction could raise. The findings of Becher et al. (2015) show that deal success is the main incentive investment bankers face.

2.2 Theoretical framework

In the academic literature, research has analyzed diverse aspects of the association between investment banking and the issuance of stock ratings. Analysts may not always publish objective ratings for various reasons.

Lin and McNichols (1998) were one of the first to examine analyst bias. They investigated the influence of underwriting relationships on stock ratings and forecasts. The paper finds that analysts which are lead and co-underwriter publish considerably more favorable growth forecasts and ratings relative to non-affiliated analysts. Moreover, the

(14)

14

findings of their research suggest investors assume it is more likely for analysts to advise a ‘Hold’ rating, where ‘Sell’ is warranted. In their research, the reaction of the market is measured by calculating the buy-and-hold-returns subtracting the corresponding period buy and hold returns for the portfolio of companies which are matched on size decile. Moreover, the findings of Michaely and Womack (1999) indicate that stocks recommended by analysts involved in underwriting activity do worse compared to ‘’Buy’’ ratings by non-affiliated stockbrokers. They state that affiliated analysts are being excessively optimistic to support the investment banking business. This is consistent with the findings of Chan, Karceski and Lakonishok (2003). Their paper argues that analysts might publish an optimistic rating on a company to gain favor with managers who could bring in future investment banking deals. Similarly, Jegadeesh et al. (2003) observe that analysts ratings are biased in such way that corresponds with economic incentivized behavior, encountered by sell-side brokerages.

Becher and Juergens (2015) show that bullish stock ratings published by biased analysts on the acquiring firm raise the odds of the transaction to succeed. An appreciation of the currency of the bidder (higher share price), decreases the absolute price that is necessary to buy the target firm. Similarly, bearish ratings on the target lead to a lower price paid for the target. An appreciation of the acquirer’s share price, as well as a depreciation of the target’s shares are both positively associated with the odds of deal success.

Groysberg et al. (2011) points out that compensation does not rely on the accuracy of earnings forecasts. Hence, the paper argues that analyst compensation is designed to reward activity that is beneficial towards the revenue of investment banks and brokerages. This corresponds with the findings of Hong and Kubik (2003) who find that relatively optimistic analysts are rewarded. They conclude that job separations are less dependent on being accurate and more so on being optimistic. Additionally, Hong et al. (2000) state that younger analysts are penalized more severely for inaccurate forecasts or reckless forecasts.

Aside from investment banks, brokerage firms also face the conflicts of interest. Gu, Li and Yang (2012) show for stocks in which fund companies have taken large positions, analysts issue more bullish stock recommendations when their brokerages receive trading commission fees from these companies. Moreover, Firth, Lin, Liu and Xuan (2012) conclude that the rating of an analyst is more bullish to consensus when the stock is included in the portfolio by the clients of the analyst’s brokerage.

This is relevant to investors. Barber et al. (2004) investigated the abnormal returns of stock ratings issued by investment banks compared to firms that perform independent research. The results show an outperformance by the independent research firm bull ratings.

(15)

15

The literature is in disagreement about the existence of the bias, since not all research reach similar results. Contrarily, Cowen et al. (2006) do not observe any bias in the ratings of affiliated analysts. They state that analysts who are hired by investment banks publish forecasts that are indistinguishable from those published by analysts from brokerage firms and autonomous research companies. Correspondingly, the findings of Agrawal and Chen (2008) disagree with the belief that affiliated analysts are able to mislead investors with optimistically biased security ratings. Jacob et al. (2003) argue that investment banks tend to have more resources and are can subsidize the research function due to investment banking revenues. Therefore, their research finds that earnings expectations of investment bankers tend to have a lower forecast error and are more bearish compared to expectations of autonomous research firms. In addition, Boni and Womack (2002) document that independent research firms are not able to provide the compensation packages that investment banks can. Thus, they would never be able to retain or compete for the best analysis.

The research of Malmendier and Shanthikumar (2014) investigates the incentive misalignment by a proxy of affiliation. Their main evidence is split into two ways. They proof that analysts with investment banking pressure and affiliation can have strategic reasons to overly issue bullish ratings, however their earnings forecast seems to be unbiased. Therefore, they conclude that financial analysts speak in two tongues, as the signs do not correspond with each other.

Mikhail, Walther and Willis (1997) document an improvement in forecast accuracy of analysts, as analysts gain more firm-specific experience. Hence, they implicate that the expertise of an analyst’s experience can be utilized in order to enhance the accuracy of the prediction of earnings. The research of Clement, Koonce, Patel and Rau (2007) argues that a link between an analyst’s previous working experiences and his current job is necessary, to influence the performance of his coverage. Lastly, Clement and Tse (2005) find that the more analysts that follow a company, the higher the likelihood of herding.

The incentives that sell-side analysts face should have been reduced after the change in the regulatory landscape since 2003. Kadan et al. (2009) performed an event study on the Global Analyst Research Resettlement, related to the regulations on sell-side research. They document that after the change in regulation, optimistic ratings have become less persistent, while neutral and bearish ratings became more frequent and less informative. Finally, they state the probability of publishing optimistic ratings does no longer depend on analyst affiliation, contradicting the predictions of this study.

(16)

16

This research tries to add to the existing discussion in the literature on the influence of analysts advising M&A deals on stock ratings. Specifically, the paper tries to shed light upon the conflicts of interests that analysts from investment banks face in the new regulatory landscape. This research relates to all prior research on analyst affiliation and in particular that on M&A affiliation. Predominantly, it follows the research of Kolasinski and Kothari (2008) and combines the methodology of Lin and McNichols (1998) and Becher et al. (2015).

(17)

17

Table 1: Theoretical framework

Authors Driver of analyst bias Conclusion

Lin and McNichols (1998) Initial or seasonal equity offerings Affiliated analysts publish more bullish stock ratings, relative to the non-affiliated. Michaely and Womack (1999) Initial or seasonal equity offerings Affiliated analysts publish more bullish stock ratings, relative to the non-affiliated.

Chan et al. (2003) Investment banking activities Shifts in the distribution of earnings surprises, especially for growth firms. Less non-negative

surprises.

Barber et al. (2004) Investment banking activities and brokerage

incentives

Autonomous research firm’s bull ratings perform better than investment banks’ ratings.

Cowen et al. (2006) Equity underwriting activities Affiliated analysts do not publish biased stock ratings.

Agrawal and Chen (2007) Investment banking activities Affiliated analysts publish more bullish stock ratings, relative to the non-affiliated.

Hausholter and Lowry (2008) Mergers and acquisitions transaction fees Bullish bias in ratings of affiliated analyst when firms engage in M&A activity.

Kolasinski and Kothari (2008) Mergers and acquisitions transaction fees When advising the acquiring firm (target), the analyst will more often upgrade the stock rating on the acquiring firm (target). Additionally, analysts advising the target firm are more likely to upgrade ratings on the acquiring firm after the exchange ratios are negotiated.

Groysberg et al. (2011) Revenue-generating activities of investment

banks or brokerage firms

Compensation does not rely on earnings forecast accuracy. Analyst compensation is designed to praise activity which is beneficial toward brokerage and investment-banking revenues.

Firth et al. (2012) Revenue-generating activities of investment

banks or brokerage firms

Analyst’s ratings on stocks are more bullish if the stock is held in the portfolio of the brokerages.

Gu, Li and Yang (2012) Investment banking activities Earnings skewness and analyst forecast bias are positively correlated, only a skewness-induced bias

exists. Malmendier and Shantikumar

(2014)

Initial or seasonal equity offerings Affiliated analysts publish bullish stock ratings, compared to non-affiliated analysts. Reversed influence found for the predictions of earnings per share (EPS).

Becher et al. (2015) Revenue from M&A activity Bullish advise on the acquiring firm, as well as bearish advise on the target raise the odds of a

successful transaction.

Merkley et al. (2017) Herding A shock in the amount of financial analysts that cover a market influences analyst competition and

(18)

18

3 Hypotheses

In this chapter, the paper formulates hypotheses based on the theoretical framework and the drawn expectations. Each hypothesis describes the expected sign of the bias corresponding to the economic incentive exerted on the analysts.

3.1 Prediction of M&A bias

A) Advising and covering the acquiring firm

Optimistic research could be beneficial to the client in at least three ways (Kolasinski & Kothari, 2008). First, if the transaction is paid in a stock-for-stock deal, bullish ratings on the acquiring firm could raise the share price and therefore appreciate the currency (share price) of the bidder. Thus, making the deal cheaper for the acquiring party. In turn, this could increase the amount of fees received by the analysts. Moreover, bullish coverage on the acquiring firm might raise the odds of deal success, since the probability of reaching shareholders’ approval of the bidder could rise. Since the M&A fees are conditional on deal success, financial analysts face an incentive to issue optimistically biased ratings. Lastly, optimistic coverage can appreciate the share price of the firm covered favorably impacting management compensation (Womack, 1996). In their gratitude, they might reward bankers with future M&A deals.

Incentives alter due to the timing. Issuing bearish stock ratings after the exchange ratios are negotiated would decrease the share price of the acquiring firm, leading to a relatively reduced target’s share price in stock-for-stock transactions. The effect would be marginal (Kolasinski & Kothari, 2008). Nevertheless, analysts might be influenced by this. Contrarily, pessimistically biased research could negatively impact management compensation and thus negatively influence the relationship for future M&A deals. Consequently, this paper hypothesizes that opposite forces will cause no bias after the exchange ratios of the transaction are negotiated. Kolasinski and Kothari (2008) have mentioned both effects, yet they do not examine it in their study. The following hypotheses are formulated:

H1.1: Analysts advising the acquirer publish more bullish ratings on bidders than

(19)

19

H1.2: Analysts advising the acquirer issue unbiased ratings, after the exchange ratios of the

transactions are fixed.

B) Advising the acquiring firm and covering the target

Bearish coverage on a target could decrease the target’s share price, which reduces the number of shares or the amount of cash the bidder has to give up in return for the transaction. According to Kolasinski and Kothari (2008), analysts face incentives to publish bullish ratings after the exchange ratios are negotiated. Favorable prospects of target-firms could have a higher probability to be approved by bidder stakeholders. This would increase the odds of the deal to succeed. Hypotheses 4.1 and 4.2 represent the aforementioned expectations: H2.1: Analysts advising the acquirer publish more bearish ratings on targets than

non-affiliated analysts, before the exchange ratios of the transactions are fixed.

H2.2: Analysts advising the acquirer publish more bullish ratings on targets than

non-affiliated analysts, after the exchange ratios of the transactions are fixed.

C) Advising and covering the target

Bullish coverage by analysts may raise a target’s stock price. As a result, the absolute price paid by the bidder raises, which is beneficial for both the target manager as the M&A advisor. In stock-for-stock transactions, target managers want to obtain the maximum value in stock of the acquiring firm as they can get which is swapped for target stock. In addition, bullish ratings increase the odds of shareholders’ approval of the target.

It could be argued that when the exchange ratios are negotiated, analysts who advise the target face less incentives to benefit the firm as the advisory payments could be fixed by now. However, the payment is contingent on the success of the transaction. Bullish ratings would increase the odds of the success of the M&A transaction by shareholders from the bidders as well as the targets. Another possibility is that if the target-affiliated analysts establish a reputation for generating bullish reports, future manager would be more likely to hire them (Hong and Kubik, 2003).

Financial analysts face incentives to bias ratings optimistically before and after the exchange ratios are negotiated. Therefore, the hypotheses are:

(20)

20

H3.1: Analysts advising the target publish more bullish ratings on targets than non-affiliated

analysts, before the exchange ratios of the transactions are fixed.

H3.2: Analysts advising the target publish more bullish ratings on targets than non-affiliated

analysts, after the exchange ratios of the transactions are fixed.

D) Advising the target firm and covering the acquiring firm

Analysts advising the target are incentivized to advise bearish reports on the acquiring firm, before the exchange ratios of the transactions are negotiated. These ratings could depreciate the share price of the acquiring firm, facilitating the demand for additional shares. When the exchange ratios are fixed, this incentive vanishes. Target stakeholders will be future stakeholders of the acquirer, which incentivizes analysts to issue bullish ratings on the acquiring firm. Furthermore, optimistic research might increase the probability of shareholder’s approval, increasing the chance that analysts generate M&A earnings. This is hypothesized as follows:

H4.1: Analysts advising the target publish more bearish ratings on bidders than

non-affiliated analysts, before the exchange ratios of the transactions are fixed.

H4.2: Analysts advising the target publish more bullish ratings on bidders than

non-affiliated analysts, after the exchange ratios of the transactions are fixed.

To give an overview of all hypotheses, an overview of the analyst bias is given in table 2. The table is partly consistent with the table of hypotheses from Kolasinski and Kothari (2008) and the Master’s thesis of Sonneveld (2016). However, the hypotheses that deviate from these papers are based on recent literature which alters expectations and introduces forces working in opposite directions. Since the research of Sonneveld includes the influence of selection bias in the hypotheses, they are fundamentally different from this research.

(21)

21

Table 2: Overview of the hypotheses

Predicted biases on the issuance of stock ratings by affiliated analysts. An analyst is considered affiliated when he advises the bidder or target with the transaction. The rating can be published before as well as after the exchange ratios are negotiated, which influences the predicted bias. The ratings are published on the acquirer or target.

Analyst affiliation Reporting on Timing of the rating Prediction

Acquirer Acquirer Before exchange ratios are negotiated Optimism

Acquirer Acquirer After exchange ratios are negotiated No bias

Acquirer Target Before exchange ratios are negotiated Pessimism

Acquirer Target After exchange ratios are negotiated Optimism

Target Target Before exchange ratios are negotiated Optimism

Target Target After exchange ratios are negotiated Optimism

Target Acquirer Before exchange ratios are negotiated Pessimism

Target Acquirer After exchange ratios are negotiated Optimism

E) M&A affiliation and deal success

Financial analysts face incentives to generate M&A revenues. Revenue from M&A activity is only generated whenever the transaction succeeds. Therefore, it may be that optimistically or pessimistically biased ratings by analysts affiliated through M&A advisory services raise the odds of a successful transaction. Accordingly, the hypothesis is formulated as:

H5: Analysts advising acquirers and targets publish bullish and bearish stock ratings that raise the odds of a successful transaction.

(22)

22

4 Methodology

This research employs the ordered logit model, the probit model as well as linear least squares. In this chapter, the empirical strategy to examine to what extent M&A affiliated analysts bias stock ratings in 2010-2016 is described.

4.1 The set-up

In this section, ordered logistic regressions analyses are employed. Analyst ratings are the dependent variable, whereas an ordinal variable is constructed for the linear logistic regressions. The ordinal variable is ordered in a reversed order, using a five-category system. A financial analyst publishes a “strong buy” rating when the variable indicates the value 1; 2 illustrates the “buy” rating; 3 illustrates the “hold” rating; 4 illustrates the “sell” rating and finally, 5 illustrates the “strong sell” rating. Ratings published on the acquiring or target firm are divided. Moreover, advisors to the acquirer or target are indicated by the binary indicator-variable. The binary variable indicates 1 in the presence of affiliation originated from M&A transactions, while simultaneously covering the firm. Control variables are added that affect the category of the stock rating. Finally, fixed effects are included.

4.2 Excluding selection bias

Considering the lack of statistical power when measuring the up- or downgrade of stock ratings in a time-span of 2 years, the regression deviates in part from the papers of Kolasinski and Kothari (2008) and Becher et al. (2015). Instead of measuring the change in ratings, the paper focuses on the absolute stock ratings. This method is partly in line with the research of Lin and McNichols (1998) and Malmendier and Shanthikumar (2014). In order to exclude selection bias, the dataset only contains ratings published within 90 days of the announcement date of the transaction. By excluding all other ratings, M&A advisors are expected to be retained by the investment bank. Otherwise, investment banks might pick investors as advisors that are bullish on their own stocks. Excluding these prevents the influence of reversed causality.

4.3 Estimating the models

A) Advising the acquiring firm and stock ratings

Now, the ratings published on the acquirer by analysts advising the acquirer and non-affiliated analyst are analyzed, before as well as after the announcement of the M&A transaction are negotiated. The first model tests whether acquirer-advising analysts have an increased

(23)

23

likelihood to publish bullish ratings relative to non-affiliated analysts. Below a formal presentation of the model:

!"#$%& = ) + +,"-.$%&+ +/0123456 + +7-"83 + +9012.":;6

+ +_<012.":;6 ∗ ">>_$%&+ +_?86"! + +_@4A-;3#!8 + B

(1)

Where !"#_$%& is the ordinal variable for the ratings published on the bidding firm. "-._$%& is a binary indicator which indicates 1 if the rating is published by an analyst advising the acquiring firm. As the ordinal variable is constructed in a reversed order, the coefficient of the indicator is predicted to be below zero, indicating an optimistic bias before the exchange ratios of the transactions are negotiated. When the exchange ratios are negotiated, the coefficient is expected to be indistinguishable from zero. This implies that a bullish bias is expected before the exchange ratios are negotiated and an unbiased result is predicted after the exchange ratios are already negotiated.

The regression controls for the company size (SIZE). Size is defined as the log of the total market value of a company’s outstanding shares in 2010-2016. The average over the years has been calculated. This variable is included to proxy the unpredictability of the proceeds of the future (Kolasinski & Kothari, 2008). Larger firms provide more certainty than smaller firms. Therefore, the variable is expected to be positively associated with stock ratings. Consequently, the coefficient is hypothesized to be below zero. Moreover, the analysis controls for the number of days between the issuance of the stock rating and the date of the deal (DAYS). Analysts are presumed to be increasingly influenced in their ratings on firms when the rating reaches nearer to the completion of the deal. Hence, an increase in the variable DAYS is negatively associated with stock ratings. Consequently, the coefficient of the DAYS variable should have an opposite sign to "-._$%&. Another variable which is controlled for is the sum paid by the acquirer for the target firm. The incentive for analysts advising the acquirer/target during a deal to bias a rating is presumed to be more persistent for larger transactions and therefore the sign is expected to correspond to the coefficient of the indicator. In addition, an interaction term consisting the log of transaction fee defined as (VALUE) and the indicator "-._$%& for the acquiring firm is included. The analyses are controlled for fixed effects. First, shocks on industry level are controlled for by including two-digit SIC codes. This is done by converting the four-two-digit SIC code to a two-two-digit code and creating 62 dummies, one for each SIC code in the sample. Additionally, year fixed effects are controlled for by including the year of every specific deal event, respectively. Therefore,

(24)

24

the regression accounts for unobserved heterogeneity in all the transactions over the period 2010-2016 and firm industry.

Subsequently, the regression is performed again with the ratings now reported on the target. The “RAT” variable is now adjusted:

RAT_IJK = α + β_,ADV_JPQ+ β_/logSIZE + β₇DAYS + β₉logVALUE + β_<logVALUE ∗ ADV_JPQ+ β_?YEAR + β_@INDUSTRY + ε

(2)

B) Advising the target and stock ratings

Next, the paper examines the relationship between analysts that are advising the target and the issuance of stock ratings on target firms. They are incentivized to act bullish on the target, before as well as after the exchange ratios are negotiated. Therefore, the following regression is estimated:

!"#_{_$`} = ) + +_,"-._{_$`}+ +_/0123456 + +₇-"83 + +₉012.":;6 + +_<012.":;6 ∗ "-._{_$`}+ +_?86"! + +_@4A-;3#!8 + B

(3)

Where the model is identical to regression (1), except that the target advising indicator is used ("-._{_$`}) instead of the acquirer and the ratings are published on the target.

Second, the model is analyzed again for analysts advising the target, however the stock ratings are now issued on the acquirer. A negative bias is expected before the exchange ratios are negotiated, which is predicted to switch to an optimistically bias after the exchange ratios are negotiated. The regression is modified to the formal model below:

!"#_$%& = ) + +_,"-._{_$`}+ +_/0123456 + +₇-"83 + +₉012.":;6 + +_<012.":;6 ∗ "-._{_$`}+ +_?86"! + +_@4A-;3#!8 + B

(4)

C) Firm coverage from affiliated analysts and the odds of deal success

The M&A fees of financial analysts are contingent on deal success. Thus, one major incentive to issue optimistically or pessimistically biased stock ratings is to raise the odds of a successful transaction. This section will investigate if the odds of deal success increase after biased stock ratings. Unlike the five ratings categories that are ranked, deal success only has two possible outcomes and they are not ordered. Furthermore, the data is not normally distributed, since 93% of all deals are completed (left-skewed). Therefore, probit regressions

(25)

25

are employed. First, bullish ratings published by affiliated analysts during M&A deals will be examined. Second, bearish ratings from affiliates in the M&A deal will be examined. Below is the formal model:

-cd0 3effcgg = ) + +,!dhij2klmmnop+ +/0123456 + +7012.":;6 + +986"! + B

(5) -cd0 3effcgg = ) + +_/!dhij2_krstnop+ +_/0123456 + +₇012.":;6 + +₉86"! + B (6)

Where “Deal Success” is a constructed binary indicator-variable. If the variable is 1, it indicates that the transaction has been successful and it is zero otherwise. The deal succeeded in 93% of all cases. As the data shows a strong negative skewness and there is a binary outcome (0 or 1), probit regressions instead of ordinal logistic regressions are employed. The variable “Rating” is split between bullish ratings and bearish ratings. The bullish rating indicator indicating 1 if the rating is a “buy” or “strong buy” rating. The bearish rating indicator illustrates 1 if the rating states “sell” or “strong sell”. The methodology deviates from the methodology employed by Becher et al. (2015) and Sonneveld (2016), in particular since preannouncement ratings are issued up and until 90 days before the merger whereas Becher et al. (2015) uses 50 days. Additionally, it does not measure changes of ratings, corresponding to the methodology used throughout this paper. The statistical power would be too low. In line with the research from Becher et al. (2015), linear least squares regressions are employed. In this paper, this method is employed as a robustness check. Lastly, the research on deal success has been conducted differently than Sonneveld (2016) as the dependent variable is constructed differently and three other variables are added in line with the research of Becher et al. (2015). Additionally, it tries to exclude selection bias by excluding rating that fail to be within ninety days before the merger. The regressions control for the size and deal value. Negative coefficients from the regressions would imply that when a negative (bullish) bias is found, bullish ratings raise the odds of a successful transaction and vice versa. This would provide evidence that raising the odds of a successful transaction is relevant to analysts in their decision to issue overly bullish or bearish stock ratings.

(26)

26

5 Data and summary statistics

This paragraph describes how the data has been collected and how it is managed to reach the unique and final datasets used for the methodology in chapter 4. Subsequently, the distribution of the data is presented in tables. Finally, multiple sample selections are described.

5.1 Data collection and management

Data on all mergers and acquisitions activity are obtained from Thomson One (previously Securities Data Corporation, SDC) from 2010 until the end of 2016. All acquiring or target firms established outside the U.S. are excluded from the dataset. Furthermore, the research necessitates that the acquirer and target are publicly listed, since accounting figures are generally difficult to retrieve for private entities. Data on stock ratings are collected from the Thomson Reuter I/B/E/S and the First Call Database (discontinued in 2010) from WRDS on listed American companies in the time span 2010 up and until 2016. It is a possible concern of the research design that analysts could be biased in a pursuit to win future deals. Thereby, the ratings of analysts whose firm wins M&A advisory service would wrongfully be compared with analysts whose firms fail to acquire M&A business. To solve this selection bias concern, the analysis is limited to be within ninety days of the transaction, consistent with the study of Kolasinski and Kothari (2008). In this time period, M&A investment bankers are expected to be held by the client. Consequently, investment banks are certain about whether they won M&A advisory service by a covered company. Ratings published by the acquiring firm and target firm are split. Moreover, an indicator-variable is constructed indicating affiliation whenever the firm of the financial analyst advises the acquiring or target firm, while his firm simultaneously publishes stock ratings for the acquirer/target. The sample of M&A deals solely consists of statutory M&A, excluding stock buy-back programs, acquisitions of specific assets, recapitalizations, shares which are distributed to the parent company through spin-offs, split-offs and carve-outs, tender offers as well as acquisitions of minority interest in controlled subsidiaries from the dataset. Often it is unclear how analysts are incentivized in transactions such as these and are therefore removed from the sample. Deals which are not completed, such as withdrawn or pending deals, are deleted from the sample for the first section. In the second section, the dataset includes this data to investigate deal success.

The analysis necessitates merging the Thomson One merger data and the I/B/E/S rating data. Unfortunately, the data-file of the stock ratings does not provide a matching

(27)

27

opportunity with the data-file of the mergers and acquisition activity of investment banks, as the transcript file has been lost at I/B/E/S. Therefore, the companies of all financial analysts had to be identified, collected and individually matched with the corresponding investment banks in the Thomson One database. Since the firm names for the I/B/E/S ratings file were often unclear, (e.g. “FBOSTON” corresponds with Credit Suisse; “SMITH” corresponds with Citi Bank) individual analysts had to be investigated to find the corresponding firm they were working for. LinkedIn profiles of analysts and analyst coverage pages from acquiring and target firms were used for the collection. After finding the correct names and matching these manually, it could be determined which ratings were issued by affiliated analysts.

5.2 Constructing the variables

Control variables are collected and constructed and held constant during the employed regressions. Data on the variable SIZE are retrieved from Compustat, collecting the total market value of the firm’s outstanding shares. The average of the firm between 2010 and 2016 is calculated. To test if the variables are normally distributed, Shapiro Wilcoxen tests were performed. Whenever the test rejected the normal distribution, the logarithm is taken. Moreover, data is gathered on M&A transactions from Thomson One to construct the VALUE variable. If data was missing, it is excluded from the dataset. Subsequently, an interaction term is included which is the indicator variable for advising the acquirer/target, multiplied by the M&A transaction value. Similarly, the Shapiro Wilcoxen test indicates the log variable can be taken to deal with the non-normal distribution.

Multicollinearity had to be handled in the employed regressions. The variance inflation factor is measured to quantify the severity of multicollinearity. The indicator variable for advising a deal is highly correlated with the transaction value variable VALUE. Consequently, the advising indicator and the transaction value are mean centered. Failing to deal with multicollinearity significantly influenced the results. These results have not been tabulated by this research.

As a robustness check, the variable EXPERIENCE is constructed. This is constructed as the announcement date of the rating from a financial analyst minus the date of the analyst’s first rating, found in I/B/E/S. After performing the additional Shapiro Wilcoxen test, the logarithm is taken.

(28)

28

5.3 Distribution of the data

In this section, the data is depicted into tables to present an overview of the sample. Table 3 presents how the rating variable is distributed into the five reversed ordered-categories. The variable is not evenly distributed, as 93.34% is a strong buy, buy or hold rating. Only 6.66% of all stock ratings are bearish, containing the “sell’’ and “strong sell” rating.

Table 3: Distribution of the five rating categories

Total number of ratings, separated into analyst rating scores 1 to 5 for the M&A deals between 2010 and 2016. The order is reversed from high to low. Only 6.66% of all ratings are negative .

Rating Frequency Percent Cumulative

1 (strong buy) 4,938 19.15 19.15 2 (buy) 8,635 33.49 52.65 3 (hold) 10,492 40.70 93.34 4 (sell) 1,369 5.31 98.65 5 (strong sell) 348 1.35 100.00 Total 25,782 100.00

Table 4 contains data on the sum paid by the acquirer of 3,729 transactions. Not all deals have information available on transaction value, of which the logarithm has been taken.

Table 4: M&A deal value (in millions)

Transaction value on all merger and acquisition deals in the sample. Transactions between non-listed firms or firms established outside the United States were excluded.

Variable N Mean Standard Deviation Minimum Maximum Transaction value 3,729 433 17,775 0.276 577,759

Table 5 depicts the frequencies for the payment method of the transaction during the deal. In 37.6% of all deals, only cash is used as a method of payment and in 23.2% stocks are used. Note that 22.62% of all deals have no available information on the method of payment, or are paid differently than in cash or stock.

(29)

29

Table 5: Method of payment

Distribution on the method of payment for the deal. Identified method of payments include cash, mix of stock and cash or transactions paid in all stocks. Out of 16,981 deals, 3,841 deals have been paid in an alternative way, or no information was available on the method of transaction.

Method of payment Frequency Percent Cumulative

All Cash 6,382 37.58 37.58

Mix of Stock and Cash 2,818 16.60 54.18

All Stock 3,939 23.20 77.38

Other / No information 3,841 22.62 100.00

Total 16,981 100.00

Table 6 summarizes the status of each deal covered in this study. Deals are considered uncompleted if the status is intended, pending or withdrawn. 93.06% of the deals are completed within the time frame.

Table 6: Status of the deal

Summary statistics on the status of each deal covered in the sample between 2010 and 2016. A deal is considered to be uncompleted if the status is intended, pending or withdrawn.

Status Frequency Percent Cumulative

Completed 6,702 93.06 93.06 Intended 4 0.06 93.11 Pending 195 2.71 95.82 Withdrawn 301 4.18 100.00 Total 7,202 100.00 5.4 Sample selection

A) Acquirer advisors and the influence on acquirer ratings

The M&A transactions are merged with ratings based on firm codes. Whenever the financial advisor to the acquirer was unknown, it was excluded from the dataset. All accessible analyst stock ratings published on the acquiring as well as target firms within ninety days of the transaction announcement are collected from I/B/E/S. The acquirer advisors’ names are matched individually with the investment banks of the stock ratings. If the rating was published by the firm which was simultaneously working together with the acquiring firm, the indicator variable “Acquirer advisor” indicates 1. The above criteria leave a dataset with 186,702 stock ratings, of which 11,042 are published by analysts advising the acquirer.

(30)

30

The duration between the date of the issuance of a stock rating and the date when the transaction was announcent is calculated (DAYS) as a control variable. All ratings published after the M&A deal went effective are excluded from the sample, leaving 113,354 stock ratings of which 6,805 are affiliated.

Finally, the sample is split between days before and after the exchange ratios are negotiated. The day that the transaction was announced is used as proxy to the date on which the exchange ratios are negotiated, as previous literature has found this is correct in 78% of all transactions (Mitchell, Stafford & Pulvino, 2004). As a result, 7,023 stock ratings are left before the determination of the exchange ratios and 7,612 stock ratings after the determination of the exchange ratios.

B) Target advisors and the influence on target ratings

The second dataset is the combination of the Thomson One file merged on the stock ratings based on the codes of the target firm, by analysts advising the target. Target companies with incomplete figures from the sample and all observations after the M&A transaction are removed from the dataset. The names of the advising firms from the stock rating dataset are individually matched with the investment banks and brokerage houses in the M&A dataset. Subsequently, an indicator variable is constructed which illustrates the analyst advises the target company by the value 1. This results in a dataset of 26,235 stock ratings, of which 2,181 are published by affiliated analysts.

C) Acquirer advisors and the influence on target ratings

In this section, the dataset containing all ratings on target firms is used. If data was missing on the acquirer advisors, they were excluded from the sample. Next, a binary indicator-variable is constructed, equivalent to 1 whenever an acquirer-advisor publishes stock ratings on the target. Subsequently, the ratings issued on the same date or after the M&A transaction are excluded. 22,090 Ratings are left, of which 1,000 are published by analysts who are covering the target company while advising the acquiring firm.

Repeatedly, the sample is split in two. As a result, a sample of 3,411 stock ratings is left before the exchange ratios are negotiated and another sample of 2,511 stocks ratings after determining the exchange ratios.

D) Target advisors and the influence on acquirer ratings

The database containing acquirer ratings is used repeatedly for this analysis. Subsequently, an indicator variable is constructed, that indicates 1 if an analyst is covering the acquiring firm,

(31)

31

while his firm advises the target in the M&A transaction. The sample leaves 104,089 stock ratings and after dropping ratings that are not published within 90 days of announcing the transaction, a dataset before the exchange ratios are negotiated has 6,293 ratings left and the other dataset has 7,175 ratings left after the exchange ratios are fixed.

E) Bullish and bearish stock ratings and deal success

The last hypothesis is tested by using the datasets from previous sections. First, the association between bullish stock ratings and the odds of deal success is examined. Regressions are employed before and after the exchange ratios are negotiated, for both parties in an M&A transaction. Moreover, the association between bearish stock ratings and the odds of deal success is investigated. Two indicator variables are created. “RAT bullish” is equivalent to 1 when the stock rating is either “strong buy” or “buy”, while “RAT bearish” is equivalent to 1 when the stock rating is either “strong sell” or “sell”.

(32)

32

6 Empirical results

In this paragraph the results which are found by employing the methodology described in chapter 4 are presented. The ordered logit model, probit regressions, as well as the linear least squares method are performed to test the hypotheses. In total, four datasets are used to test the impact of affiliation originated by M&A advisory. As a follow-up, the research tests if transaction success is more likely due to stock ratings published by M&A advising analysts. The association is tested by a probit regression.

6.1 Results: the influence of mergers and acquisitions on stock ratings A) Analyst advising acquirers

This section tests the influence of being employed by the bidding firm originating from M&A services and the issuance of stock ratings, published on the bidding as well as the target firm. Table 7 presents summary statistics on all the variables.

Table 7: Summary statistics Analysts advising acquirers

Table 7 reports summary statistics on variables used in the regressions on analysts with ties to the acquirers. ratings” is the ordinal dependent variable of ratings published on the acquiring firm. “Acquirer-advisor” is an indicator-variable, indicating 1 if the rating is published by an analyst that advises the acquirer and zero if non-affiliated. “Company size” is the mean of the total market value of a company’s outstanding shares in 2010-2016, reported in millions. “Transaction value” illustrates the sum transferred by the acquiring firm for the transaction, reported in millions. Experience is the duration expressed in days between the first rating reported in I/B/E/S by an analyst and the date of his rating.

Variable Observations Avg. Std. Dev. Minimum Maximum

Acquirer ratings 11,641 2.35 .900 1 5

Acquirer-advisor 11,641 .059 .041 0 1

Company size 11,641 24,496 50,429 .276 407,408

Transaction value 11,641 5,204 12,993 .172 145,785

Experience of analyst 11,641 3,190 2,193 0 8,562

Table 8 summarizes the findings of the employed regressions. The ordinal variable for stock ratings on the acquiring firm is the dependent variable. In line with the formulated hypothesis, it is expected that the coefficient is below zero before the exchange ratios are negotiated.

(33)

33

Hence, a negative coefficient implies higher odds for a bullish stock rating if the ratings are published by analysts advising acquirers. The regressions are adjusted with centered variables to correct for multicollinearity. Model L1 and L2 test the association of analysts advising acquirers and the ratings reported on the acquirer, whereas Model L3 and L4 investigate the association with ratings reported on the target.

For regressions L1 and L2, the models show significant results. Both models find coefficients below zero for the dependent variables. This implicates that stock issuers advising the bidding firm publish more bullish stock ratings on the bidders before the exchange ratios are negotiated. This is in line with the expected hypothesis and is statically significant at a 1% level. Since this is a logistic regression analysis, the coefficients in a logistic regression are a logarithmic transformation of the odd ratio. As the order of categories is reversed, the sign for the logarithmic transformed is reversed as well to make the interpretations more intuitive. The odd ratio for the coefficient of -0.367 (0.367) is 1.44, meaning that the odds of an analyst that advises the acquirer publishes a more bullish category is 44% higher than a non-affiliated analyst.

Regressions L3 and L4 show that analysts advising the acquirer, while reporting on the target firm publish more bullish ratings on the target, before the exchange ratios are fixed. However, after including control variables and fixed effects, the outcome is not statistically significant. The sign of the result does not correspond with the predictions. However, due to large standard errors, the statistical power is low and the result is not distinguishable from zero. The research fails to find a significant influence of the analysts advising the acquirer reporting on the target firm, before the exchange ratios are negotiated. Analysts advising the acquirer do not give a bearish bias to stock ratings on the target firm to decrease its stock price.

(34)

34

Table 8: Results acquirer-advisors, covering both acquirer and target firm Before the exchange ratios are negotiated

Ordinal logistic regressions are performed to test the association of analysts advising the acquiring firm and publish stock ratings, covering the acquiring as well as the target firm. The ratings are all published before the exchange ratios are negotiated. The control variables include: log of company size, log of days between ratings and the deal and log of transaction value. An interaction term is constructed “Acquirer-advisor * Log of transaction value” which illustrates the indicator variable for acquirer-advisors, multiplied by the logarithm of the deal value of the M&A transaction. Centered variables are constructed by subtracting the mean of the variable from its own value. Specifications L2 and L4 control for industry with a two-digit SIC code and year fixed effects. Reported in brackets are the standard errors.

Acquirer-advising analysts (L1) (L2) (L3) (L4) Reporting on Acquirer Reporting on Acquirer Reporting on Target Reporting on Target Centered Acquirer-advisor -0.397*** -0.367*** -0.645*** -0.336 (0.001) (0.126) (0.236) (0.302)

Log of company size 0.054*** 0.135***

(0.017) (0.039)

Log of days between transaction and rating 0.002*** 0.017***

(0.000) (0.002)

Centered Log of transaction value -0.007 -0.091*

(0.017) (0.042)

Centered Acquirer-advisor* Log of transaction value -0.137** -0.015

(0.063) (0.042)

Time fixed effects No Yes No Yes

Industry fixed effects No Yes No Yes

Observations 7,023 7,023 3,143 2,028

Pseudo R-squared 0.006 0.010 0.009 0.043

Second, the ordinal logistic regressions will be performed after the exchange ratios are negotiated. The outcome of the performed regressions are p in table 9. Acquirer-advising analysts could be confronted with mixed incentives, as forces working in opposite directions were present. Table 8 depicts the outcome of the employed analyses. Model L5 shows a negative coefficient that is significant and not in line with expectations. Including control variables and control for fixed effects does not alter the results as the coefficient stays negative and is statistically significant at a 1% level. The result is economically significant and implies that analysts who are cooperating with the acquiring firm have a ordered log odd of 0.334 of being in a more bullish category. This corresponds to an odd ratio of 1.40, which suggests affiliated analysts have 40% more chance of issuing a more optimistic rating category. This seems slightly lower than before the exchange ratios are fixed. Incentives for analysts advising the acquiring firm are hypothesized to be lower after the exchange ratios are fixed. However, it seems that the incentives are still present which was not in line with the

How does investment banking and analyst association affect stock ratings? : evidence from M&A transactions