Analyst Target Price Optimism around the World

Transcription

1 Analyst Target Price Optimism around the World Mark Bradshaw Boston College Alan G. Huang University of Waterloo Hongping Tan SUNY-Buffalo and University of Waterloo August 2012 Abstract Using a unique analyst-location data covering 11,436 analysts from 41 countries, we find that analyst optimism in target prices is exacerbated by conflicts of interest and investment banking pressures but is mediated by country-level institutional infrastructure. This relation holds for ex-ante and ex-post analyst optimism measures, for analyst traits related to conflicts of interest, for firm traits related to external financing needs, and for country-level characteristics of investor protection, financial transparency, and culture. The authors gratefully acknowledge helpful comments from Xi Liu, Masahiro Wanatabe, from conference participants at the 2012 Eastern Finance Association meetings, the 2012 Northern Finance Association meetings, and from seminar participants at University of Toronto and University of Waterloo. We thank the Social Sciences and Humanities Research Council of Canada for financial support. All remaining errors are our own. 1 Electronic copy available at:

2 Target prices convey sell-side analysts assessment of the future value of underlying stocks and are presumably the culmination of analysts research efforts. In recent years, analysts have increasingly issued target prices alongside earnings forecasts and stock recommendations in their equity research reports. Yet, the credibility and usefulness of target prices has long been dubious. In spite of target price hits from time to time, 1 media and investment managers frequently accuse target prices as merely sales hype. In a work that contributed to a Pulitzer Prize in 2002, the New York Times journalist Gretchen Morgenson criticized target prices as being based more on fantasy than reality and concluded that Price Targets are Hazardous to Investors Wealth. 2 Prior research demonstrates significant optimistic bias in target prices relative to current trading prices. Asquith, Mikhail and Au (2005) and Brav and Lehavy (2005) both document an average return implied by analyst target prices of 32.9% for the period of 1997 to 1999, while Bradshaw, Brown and Huang (2012) document an implied return of 24.0% from analyst target prices for the period of 2000 to In contrast, Mehra (2003) finds that from 1802 to 1998, the real annual U.S. equity return is only 7.0%, and that the equity returns in other developed countries are even lower. Data provided by CRSP shows that the nominal NYSE/Nasdaq/Amex market return over the period of 1997 to 2009 is 8.1%. Despite such optimism in analyst target prices, the literature has not offered much insight on the determinants of target price optimism. An emerging literature investigates the determinants and investment value of target prices and their relation to other analyst forecast outputs such as earnings forecasts and stock recommendations. 3 This literature establishes that target prices are useful. For example, target prices trigger large stock price movements, even after controlling for other information 1 One example of target price hits is Henry Blodget s famous $400 target for Amazon.com in On December 16, 1998, Henry Blodget issued a target price of $400 per share for Amazon.com when the stock closed at $ in the previous day. The stock rose over 19% that day, and blew through the $400 price target in about three weeks See for example Bandyopadhyay, Brown, and Richardson (1995); Bradshaw (2002); Brav and Lehavy (2003); Asquith, Mikhail and Au (2005); Gleason, Johnson, and Li (2008); Huang, Mian, and Sankaraguruswamy (2009); Da and Schaumburg (2011); and Da, Hong, and Lee (2010). 2 Electronic copy available at:

3 included in analyst reports (Asquith et al. 2005). Furthermore, possibly due to the lack of data, however, few studies examine target price in a cross-country context. 4 This paper fills in the void in the literature by examining target price optimism in an international context. Using a unique analyst-location data that covers 11,436 analysts located in 41 countries, we find that analyst optimism in target prices is exacerbated by conflicts of interest but is mitigated by good country-level institutional infrastructure. Our results, while confirming the classic conflicts of interest argument that analysts tend to be overly optimistic when there exist potential benefits for biased research, establish that there are channels that mitigate analyst bias. In particular, we demonstrate that analysts in countries with good institutional infrastructure as characterized by strong investor protection, transparent financial environment, and strong cultural forces show less optimism in target prices. Analysts target prices may be optimistic compared with both the current and future stock prices. Accordingly, we design optimism measures that examine how a particular target price forecast compares with not only the concurrent stock price but also the realized price in the next 12 months. For the former, we use the implied return of the target price and its rank; and for the latter, we examine whether and for how long the target price is met over the next 12 months. These measures reflect either the ex-ante optimism exhibited by the analyst, or the materialized, ex-post optimism of the analyst. Our data allows us to identify a number of analyst attributes that are directly related to the information advantage on the covered firms and the potential conflicts of interest between the analysts and their covered firms. In particular, we identify analysts who work for pure brokers or brokers that have previous investment banking ties with the covered firms, or reside locally. These analysts may provide less biased target price forecasts due to their information advantage arising from underwriting relationship and geographic proximity (e.g., Malloy 2005; Ke and Yu 2006; Bae, Stulz and Tan 2008). These analysts, however, may be subject to greater conflicts of interest, originating from both personal and business considerations. For example, an analyst may seek to maintain a good relationship with the target firm s management to generate underwriting business and to enhance her career opportunities. In fact, the 4 Target price forecasts are available for academic subscription through the I/B/E/S starting in

4 literature on stock recommendations shows evidence that analysts who are employed by banks with business ties with the target firms tend to provide more optimistic recommendations (Dugar and Nathan 1995; Lin and McNichols 1998; Michaely and Womack 1999; Krigman, Shaw, and Womack 2001). Thus we have a trade-off between the effect of information advantage and conflicts of interest on target price optimism; and the net impact of these two forces is an empirical question. Aside from direct investment banking pressure described above, we also consider indirect investment banking pressure. Analysts can identify firms that have greater needs of investment banking business and can try to please these firms with overly optimistic target prices. In fact, Bradshaw, Richardson, and Sloan (2006) document that external financing needs are positively related to overoptimism in analyst forecasts of earnings, stock recommendations, and target prices. Consistent with the classic conflict of interest, indirect investment banking pressure in firms with greater external financing needs induces target price optimism. In line with this argument, we also examine whether external financing needs induce inflated target prices. We further investigate whether country characteristics of investor protection, financial transparency, economic development and culture mitigate target price optimism. On the one hand, due to better investor protection and legal enforcement, and more transparent information environment, analysts in countries with better institutional infrastructure would refrain from conflict-of-interest behaviours, since engaging in such activities would likely be subject to more severe penalties. Thus analysts in these countries would provide less biased and more accurate target prices. On the other hand, countries with better institutional infrastructure usually have more developed financial markets and fiercer competition in financial services. Rather than aiming at providing accurate forecasts, analysts in these countries may be motivated to fulfill other business mandates through biased research. Thus, ceteris paribus, analysts from countries with better institutional infrastructure would provide more optimistic forecasts. Since these two forces offset each other, it is unclear whether better country-level institutional infrastructure would help mitigate target price optimism. 4

5 We find that conflicts of interest and investment banking pressures dominate information advantage and exacerbate target price optimism, and that institutional infrastructure reduces optimism by imposing higher costs on target price biases. Distinctly, our study includes a heterogeneous cross-section of both developed and developing economies and a detailed list of country characteristics, and uses a unique data set to identify a host of variables to examine potential conflicts of interest. We organize the rest of the paper as follows. In the next section, we briefly review the literature related to analyst target price forecasts. Section II describes our data and proxies for target price optimism. Section III conducts empirical analyses on the determinants and mitigating forces of target price optimism. Section IV presents robustness checks, and Section V concludes. I. Literature review Financial analysts typically provide earnings forecasts, target prices, and stock recommendations in their research reports to convey their assessment of the covered firms. Although target price is one of the three major components of analysts research output, much of the literature has focused on earnings forecasts and stock recommendations. It is only until recently that academic research has begun examining the determinants and investment value of target prices. Bandyopadhyay, Brown, and Richardson (1995) find that target price forecast revisions co-move with earnings forecast revisions. Bradshaw (2002) demonstrates that analysts issue target prices to support their stock recommendations. Brav and Lehavy (2003) and Asquith, Mikhail and Au (2005) show that analysts target prices have investment values incremental to their stock recommendations. Gleason, Johnson, and Li (2008) show that target price accuracy increases with the use of rigorous valuation techniques rather than a heuristic one, especially for analysts who are adept at accurate earnings forecasts. Huang, Mian, and Sankaraguruswamy (2009) find that portfolios based on changes in both consensus recommendations and target prices are more profitable than those based merely on changes in recommendations or target prices. Da and Schaumburg (2011) document profitable trading strategies based on industry relative valuations implicit in analyst target prices, suggesting that the informativeness of target prices mainly derives from analysts ability to assess the relative performance of stocks within a specific industry. Da, Hong, and 5

6 Lee (2010) find that the investment value of target price derives from earnings forecasts and the implied discount rates embedded in the forecasts of P/E ratios. While most studies in this literature conclude that target price is informative and value relevant, there also exists some doubts over the investment value of target prices. Bradshaw, Brown and Huang (2012) find that target price accuracy in the U.S. is limited and argue that the lack of accuracy is possibly due to the fact that target price forecasting is largely an unmonitored activity. These authors document some ability of analysts to provide more accurate prices across time; however, this could simply reflect persistent differences in optimism. Bonini, Zanetti, Bianchini, and Salvi (2010) find that target prices are systematically biased. Perhaps due to the lack of large sample data, few studies examine target prices in a cross-country context. Bonini et al. (2010) find that analyst target price forecasts are systematically biased by factors such as boldness, firm size, and market momentum in the Italian market. Bilinski, Lyssimachou, and Walker (2011) report that analyst characteristics and affiliation affect accuracy, boldness, and revision frequency of target prices in 16 countries. Kerl (2011) finds that target price accuracy is negatively related to analyst optimism and firm risks in the German market. Since these papers examine analyst target prices in different regions and countries, their results are not directly comparable due to the use of different samples and the differences in institutional environment in which the firms and analysts are situated. For example, Bilinski, Lyssimachou and Walker (2011) report that bank affiliation increases target price accuracy in their sample of 16 developed countries, while Kerl (2011) finds that such an affiliation has no effect on target price accuracy. While Bilinski et al. (2011), Bonini et al. (2010), and Kerl (2011) also examine target prices in an international context, our paper differs from these papers in two important ways. First, our focus is on target price optimism, while these other papers focus on target price accuracy. Although ex-ante optimism is arguably negatively associated with ex-post forecast accuracy, it needs to be so theoretically: For example, extreme low target prices could be associated with low ex-post accuracy. In fact, we use target price level as an optimism measure (dependent variable), whereas target price level is treated as an 6

7 independent variable in these studies. Second, our sample countries include not only developed markets but also developing markets. This enables us to examine the attenuation effects brought about by country characteristics. The more heterogonous country characteristics and more complete coverage also increase the generalizability of our results. Our paper is also closely related to the literature that relates information advantage and conflicts of interest to analyst forecast performance. Bae, Stulz and Tan (2008) find that analysts resident in a country make more precise earnings forecasts for firms in that country than non-resident analysts; moreover, the local analyst advantage is high in countries where earnings are smoothed more, less information is disclosed by firms, and firm idiosyncratic information explains a smaller fraction of stock returns. Chen and Martin (2011) report that analysts are more accurate after the underlying firms borrow from their affiliated banks. The increase in forecast accuracy is more pronounced for borrowers with greater information asymmetry, and for borrowers with bad news and high credit risk. Their results suggest that there is information spillover from commercial banking division to equity research division within financial conglomerates. Duan, Hotchkiss, Jiao (2011) find that the selling by mutual funds whose families provide financial services to underlying firms employee pension plans is more likely to be motivated by an information advantage than their buying, suggesting that mutual funds obtaining information advantages through pension business ties. A large literature demonstrates analyst optimism in stock recommendations due to the underlying incentives on the part of either analysts or their employers. The evidence is mixed for earnings forecasts (See Dugar and Nathan 1995; Lin and McNichols 1998; Michaely and Womack 1999; Krigman, Shaw, and Womack 2001; Boni and Womack 2002; and Cliff 2006; Chan, Karceski, and Lakonishok 2007; Ljungqvist et al. 2007; Barber, Lehavy, and Trueman 2007; Agrawal and Chen 2008). Mehran and Stulz (2007) focus their discussion of the economics of conflict of interest in financial institutions on analyst optimism in earnings forecasts and stock recommendations. They point out that reputation, labour market, competition, institutional investors, and legal and regulatory actions could mitigate the adverse impact of conflicts of interest. Ke and Yu (2009) find that the disconnection between individual analysts stock 7

8 recommendations and earnings forecasts is greater for analysts who face investment banking pressure, when analysts follow firms with heavier insider selling or higher institutional ownership, when the analysts brokerage house relies more on trading commissions, or in periods with more extreme investor sentiment. We extend the analysis of analyst conflicts of interest and information advantage to target price forecasts, and use the cross-country setting to investigate the effectiveness of institutional framework at the country level in disciplining analysts from making biased forecasts. II. Data and Key Variables In this section we describe our data sources, the criteria for constructing our samples, and computational details of proxies for target price optimism. We also show summary descriptive statistics for the samples we use in our analyses. Table I describes our variables and data sources. [INSERT TABLE I HERE] A. Sample We begin our sample selection by identifying analysts target price forecasts that are revisions to their own forecasts issued over one week to six months earlier based on the I/B/E/S price target detail file during , for both U.S. and non-u.s. firms. 5 From Compustat, we obtain daily stock prices, adjustment factor, market capitalization and annual financial data. We use the latest closing price within three days before the announcement date of an analyst s target price as the concurrent market price and use it to compute the implied return of the target price. We adjust for discrepancy in the underlying currency between Compustat and I/B/E/S using the daily exchange rate from Compustat. We obtain stock prices from the I/B/E/S monthly summary file to verify the consistency in stock prices between Compustat and the I/B/E/S. The closing dates for the monthly stock prices from the I/B/E/S summary file fall usually in the middle of the month. We keep only those target prices whose latest market prices (after 5 The target price data in I/B/E/S prior to 2002 is dominated by U.S. sample. Since there is a need to maintain a degree of cross-country variations in our study, we drop the pre-2002 data. Adding the pre-2002 data, however, does not change the conclusions of the paper. 8

9 considering adjustment factor) from Compustat within three days before the issuance dates are between 70% and 400% of the monthly stock prices from the I/B/E/S. 6 Extant literature (e.g. Brav and Lehavy 2003; Asquith et al. 2005) usually separates target price forecasts into buy / hold / sell subsamples based on the concurrent stock recommendations issued by the same analysts. Restricting the sample of target prices to those with non-missing concurrent stock recommendations, however, result in a substantial reduction of sample size. In our final sample, only 72.4% of the target price forecasts are accompanied by stock recommendations. We bypass this restriction by inferring the buy/sell directions based on the target prices benchmarked against the prevailing stock prices on the forecast dates. For completeness, we include stock recommendations in robustness checks. We use the annual volumes of Nelson s Directory of Investment Research for to identify the country locations of financial analysts. 7 The Nelson s Directories provide information on nearly 1,700 research firms with approximately 40,000 equity analysts covering publicly traded companies located around the world, with the full names and city locations for equity analysts and their associated research firms. We use the I/B/E/S broker translation file to obtain names for analysts and brokers included in the I/B/E/S detail file. We then follow the same procedure as Bae, Stulz, and Tan (2008) and Bae, Tan, and Welker (2008) to match these broker and analyst names to those from the Nelson s Directory of Investment Research, to obtain country locations for the I/B/E/S analysts. 8 Since we are interested in examining the mitigating effects of analyst traits and analyst-country characteristics on the determinants of forecast optimism in target prices, we keep only those target prices issued by analysts that we can identify their country locations. We further require each firm country to have at least 10 firms, and each analyst country to house at least five analysts to make meaningful comparison of 6 These cutoffs correspond roughly to the bottom and top five percentiles in the distribution of the ratios between the prices from Compustat and the I/B/E/S. 7 Nelson Publishing, Inc. stopped producing its Directory of Investment Research after We implicitly assume that analysts do not change their locations afterwards. 8 We obtained our broker translation file directly from I/B/E/S in September After that date, we supplement using the I/B/E/S recommendation detail file and target price file, which include analyst names and abbreviated broker IDs, but not broker full names. Where possible, we confirm broker names using Nelson s Directory. 9

10 analyst forecast activities across country. Our final sample includes 1,069,879 target prices issued on 16,867 distinct firms by 11,436 analysts located in 41 countries during the period of B. Measures of target price optimism Target prices explicitly convey analysts assessment of the value of underlying stocks, usually over the next 12 months from the date of issuance. 10 An optimistic target price carries the connotation that the target price issued is higher than the current market price. For example, during the Dot-com bubble period of , many analysts were criticized for issuing overly optimistic price targets for Internet stocks. Thus, a natural measure of optimism is the distance between target price (TP) and current market price (P), TP2P, defined as TP/P 1. TP2P is positively related to optimism: A larger value of TP2P indicates greater optimism. TP2P is in essence the implied return of a firm within the next 12 months estimated by an issuing analyst. This implied return measure is, however, not adjusted for potential risks associated with each stock. To ameliorate this concern, we make an industry-adjustment to TP2P following the popular industry-adjusted return method in event studies. In particular, we define TP2PRank to be the rank of TP2P within its 2-digt-SIC industry in any given year. TP2PRank is coded between 0 and 99. To preserve country heterogeneity, we do not rank TP2P by country. TP2P and TP2PRank are ex-ante measures of optimism; namely, the degree of optimism is revealed before the stock price is realized to materialize or de-materialize the optimism. We are also interested in how optimism is materialized ex-post for the time horizon of the issued price target. As previously discussed, analyst target price forecasts are usually for a time horizon of 12 months. If a target price is more optimistic ex-post, the chances are that (1) the target price is less likely to be met during the next 12 months, and (2) the maximum price over the next 12 months is smaller than the target price. 9 Simply aggregating the number of analysts by analyst country would give a total of 12,106 analysts. However, there are analysts who move around countries, so that the distinct number of analysts is smaller. 10 For example, in its ratings disclosure ( Goldman Sachs states: Price targets are required for all covered stocks. The return potential, price target and associated time horizon are stated in each report... When a security is rated as Attractive, [t]he investment outlook over the following 12 months is favorable relative to the coverage group s historical fundamentals and/or valuation. The same outlook period applies to other ratings. 10

11 Correspondingly, we define Ratio12 as the percentage of trading days in the next 12 months that the stock price is lower than TP; and TPNOTMET as a dummy variable that equals one if the maximum stock price over the next 12 months is lower than TP, and zero otherwise. TPNOTMET is an inverse measure of target price optimism used in the literature (e.g., Bradshaw, Brown and Huang 2012). Although we use 12 months as our primary time horizon to compute optimism measures, we note that target prices are frequently updated. In our sample, the average (median) interval for a forecast revision is 89 (71) days. We therefore also consider forecast optimism over three- and six-month horizons as robustness checks. In untabulated results, our conclusions remain qualitatively the same with these alternative time horizons. These results are available upon request. Finally, in the earnings forecast literature (e.g., Hong and Kubik 2003, Cowen et al. 2006, Ke and Yu 2006), optimism is commonly defined as relative ranking of earnings forecast within-firm. Following the methodology used in Hong and Kubik (2003) and Ke and Yu (2006), we construct a within-firm proxy for optimism in analyst target price forecasts as follows. First, we compute the average target price by each analyst who covers a given firm-year. Second, we rank all analysts based on the average target prices for the firm-year and create a variable denoted as rank ijt,where i indexes firm, j indexes analyst, and t indexes year. rank ijt is bounded below at 1 and above at the number of following analysts. Finally, we construct a ranking score normalized by the number of analysts across firms: TPRank ijt = 100 (rank ijt 1) / (Number of analysts it 1) TPRank has a value of 0 to 100 for each firm-year. As Hong, Kubik and Solomon (2000) acknowledge, however, the relative optimism measures could be less reliable for firm-years with few analyst coverage. We therefore require at least three analysts to cover a given firm-year. Since the construction of TPRank requires firm-year averaging by analyst, we only use the measure for the sample that collapses the full sample by taking the mean values of the variables to the firm-analyst-year level (which allows a maximum of one observation per analyst per firm-year). 11

12 In sum, we consider the ex-ante optimism measures of TP2P, TP2PRank and TPRank, and the expost optimism measures of Ratio12 and TPNOTMET. All these measures are positively related to optimism. Table II reports their summary statistics and sample correlations. In Panel A, TP2P has a mean value of Compared with the overall equity market return of around 10% per annum in the past century, this high value of TP2P indicates a widespread of target price optimism around the world. Expost, over the next 12 months, 63% of the time the target price remains above the market price, and 27% of the time the target price is never met. These metrics indicate that optimism is non-trivial even after the issuance of the target price. As expected, Panel B of Table II shows that all of the optimism measures are significantly positively correlated. [INSERT TABLE II HERE] To show the distribution of target price optimism, Figure 1 plots the percentage histogram of and Ratio12. Panel (a) of the Figure shows that the majority of target prices have a value of greater than zero. In fact, 81.5% (61.5%) of the target prices have implied returns that are positive (greater than 0.10). This result is consistent with the literature in stock recommendations that buy recommendations dominate. If we treat positive implied returns as being optimistic and negative implied returns as being pessimistic, Panel (a) of Figure 1 shows that analysts are much more likely to be optimistic than to be pessimistic. Panel (b) of Figure 1 shows the distribution of Ratio12. Notably, we observe that more than 30% of the values of Ratio12 fall in the left-tail bin of [0.967, 1.00]. The left-tail distribution of Ratio12 reflects the fact that many target prices are seldom met during the next 12 months; in fact, earlier Table II reported that TPNOTMET (i.e., dummy that target price is never met) has a mean value of Overall, the numbers underlying Panel (b) show that 64% of the time Ratio12 is greater than 0.50; in other words, about 2/3 of the time the stock price is more likely to stay below the target price over the next 12 months. [Figure 1 about here.] Table III shows the summary statistics of our final sample by country (Panel A) and by year (Panel B). The final sample includes 41 countries, with the U.S., Canada, the U.K. and Japan being the top four countries with the largest number of sample firms. The U.S and the U.K. are also the top two countries with the largest number of analysts in residence, followed by Canada, Korea, Japan, and France. 12

13 The countries with the highest TP2P are Russia, Turkey, Pakistan, Brazil and Argentina, each with more than 30% of implied returns. New Zealand has the lowest TP2P, followed by China and France. The average TP2P across our sample countries is 21%. We note that the magnitude of target price optimism for our sample is consistent with what has been reported for individual countries in the literature. For example, in our sample of , TP2P and TPNOTMET for the U.S. have a mean of 0.20 and 0.26, respectively. Bradshaw, Brown and Huang (2012), using the sample period of , report, respectively, 0.24 and 0.36 for these two values. The somewhat lower TP2P value in our sample is because the period of , which are not included in our sample, witnessed the largest values of TP2P and TPNOTMET in the sample of Bradshaw, Brown and Huang (2012). Panel B of Table III shows the sample distribution by year. For each year during 2002 to 2011, we have an average of 106,988 target price forecasts on 8,179 firms by nearly 5,000 analysts. The number of analysts and firms peaked, respectively, in 2007 and The number of sample countries starts with 24 in 2002, and increases to 41 by We have the highest TP2P in 2008 and the lowest TP2P in 2005, with the average TP2P being 19% over the years. The average Ratio12 is 64%, indicating that 64% of the time the stock prices are lower than their target prices. The average TPNOTMET is 28%, indicating that for 28% of the stocks, the stock prices have never surpassed their target prices. [INSERT TABLE III HERE] Our last check of target price optimism is its time-series variation at quarterly frequency. Figure 2 plots the quarterly cross-sectional means of TP2P and Ratio12 during our sample period. The Figure demonstrates significant time-series variation of the optimism measures. We also plot the quarterly MSCI world-index returns with these optimism measures. Interestingly, the optimism measures move inversely with the markets. In our sample period, 2002 and 2008 experienced severe market downturns. These down markets are associated with high values of TP2P and Ratio12. In the following market rebounds in 2003 and 2009, TP2P and Ratio12 are reversed to be low. In light of this time-series pattern, we include the MSCI index return in our regressions, and in robustness checks we consider the recent economic crisis separately. [Figure 2 about here.] 13

14 C. Analyst and Firm Traits In this paper, we primarily investigate whether optimism in target prices is mediated or exacerbated by three sets of variables: analyst-specific variables, firm-specific variables, and countryspecific variables. Previous literature has documented that analysts are subject to conflicts of interest due to the related business and personal ties with the underlying firms in different contexts, although these ties may also bring analysts relative information advantage. We examine three analyst traits that are related to their information advantage and conflicts of interest: purebroker, underwriter, and local, all defined as dummy variables. Specifically, purebroker takes the value of one if the brokerage house that employs an analyst is a pure broker that has no investment banking business; underwriter takes the value of one if an analyst s employer served as either lead underwriter or co-manager for the covered firm in the past three years based on equity and debt offering from Thomson One Banker; and local takes the value of one if an analyst resides in the same country as the target firm's headquarter country. The impact of the above analyst traits on target price forecast optimism is ambiguous given the offsetting nature due to analysts conflicts of interest and information advantage. While conflicts of interest presumably motivate analysts to inflate target prices and hence elevate optimism, information advantage tends to reduce the uncertainty about the target firms and lower the optimism. For example, local analysts may understand the firm better due to geographical proximity or better knowledge of the firm; however, this information advantage may be offset or even dominated by personal conflicts of interest to secure corporate networking and career opportunities. Likewise, analysts having investment banking ties with the target firms may enjoy information advantage and provide less biased and more accurate target prices; however, such ties may also force these affiliated analysts to inflate the target prices to secure future banking business. Finally, although there is a potential information flow between 14

15 analysts and their brokers, the pressure to generate more trading commission would motivate analysts to issue optimistic target prices. 11 Consistent with the view that conflicts of interest of analysts lead to target price optimism, certain types of firms can induce analysts to exhibit greater optimism. In particular, analysts may identify firms that have greater needs of investment banking business and solicit the business by issuing more favourable target prices. Confirming this observation, Bradshaw, Richardson, and Sloan (2006) document that external financing needs are positively related to over-optimism in analyst forecasts of earnings, stock recommendations, and target prices. They further report that this relation is stronger for firms with equity offering. In addition, Toeh et al. (1998a, 1998b) report that firms manage accruals to increase earnings prior to financing activities of IPO and SEO. In this sense, firms may manage earnings to signal their financing needs. If external financing leads to optimism due to indirect investment banking pressures, so will earnings management. We therefore consider that target price optimism may be exacerbated by the following firm traits: (Bradshaw et al. s (2006) financing needs variable), 12 (Bradshaw et al. s (2006) equity financing needs variable), (Bradshaw et al. s (2006) debt financing needs variable), and earnmgmt_firm (firm-level earnings management level, measured by the absolute value of discretionary accruals from Modified Jones Model). D. Analyst-country characteristics A growing number of studies investigate the analyst forecast activities across countries and find that country characteristics affect the extent of analyst following and the properties of analyst forecasts (Jegadeesh and Kim 2006; Bae, Stulz and Tan 2008; Balboa, Gomez-Sala, and Lopez-Espinosa 2009, Tan, Wang, and Welker 2011). Most of these studies, however, focus only on analyst earnings forecasts, and provide inconclusive results regarding the impact of country characteristics on forecast-related measures such as forecast accuracy. For instance, while Chang, Khanna, and Palepu (2000) find evidence 11 Cowen, Groysberg and Healy (2006) find that pure brokers tend to issue more optimistic earnings forecasts in order to generate trading business. 12 We also use other firm financing variables such as Frank and Goyal s (2003) financing deficit measure and De Franco, Hope, and Larocque s (2011) external financing measure, and find that our results remain qualitatively the same. 15

16 that a country s legal system helps determine the accuracy of analysts, Ang and Ciccone (2001) reach the opposite conclusion. We examine how the institutional framework of a country shapes the action of analysts, which in turn affects the forecast optimism of their target prices. We consider country characteristics that are closely related to three different aspects of a country s institutional infrastructure: investor protection, financial transparency, and economic development and culture. A country with strong investor protection will likely impose higher costs on analysts potential conflict-of-interest activities. For investor protection, we consider the legal origin, the efficiency and integrity of legal system (e.g., impartial courts, judicial independence), the quality of investor protection (e.g., corruption, expropriation, bribes, property rights), and control of self-dealing. These variables measure different aspects that an investor can be protected (or expropriated) by a country s legal system. Our second group of institutional characteristics concerns financial transparency of a country. Transparent financial system gives rise to more accurate information flow, which presumably discourages target price forecast optimism. We include the following financial transparency variables: accounting standard, earnings management, and the R 2 used in Morck et al. (2001). Finally, we add a group of variables that measure the economic and cultural development of a country: GDP per capita, a dummy variable indicating whether a country is a developed country; individualism; media development, and the fraction of a country s equity held by the U.S. investors. Table I earlier lists and defines these country characteristics and cites their data sources. Overall, we hypothesize that countries with better institutional features will display lesser degree of optimism. E. Univariate Comparison of Target Price Optimism Table IV presents a univariate comparison of target price optimism conditioning on the analyst, firm and country characteristics. For analyst and firm traits, we partition the sample into two groups based on the median value of the variable for every analyst country and year. For country characteristic variables (other than binary variables), we partition the sample into two groups based on the median value of the variable for our sample countries for every year. Briefly, we make a number of observations. First, 16

17 optimism is positively associated with analyst traits of underwriter and local, suggesting that conflicts of interest dominate information advantage when underwriter analysts and local analysts set their target prices. To further corroborate the conflicts of interest story, we find that optimism is positively associated with firm financing-related variables of,, and to a lesser degree. In other words, analysts tend to issue more optimistic forecasts for firms with strong financing needs. In contrast, country-level infrastructure generally helps reduce optimism: Better investor protection, higher financial transparency, more economic development, and higher individualism all reduce the degree of optimism. Overall, Table IV provides preliminary evidence that some analyst and firm traits, as well as country characteristics affect target price optimism. [Table IV about here.] III. The Determinants of Target Price Optimism A. Benchmark Regressions In this section, we examine the determinants of target price optimism in a multivariate regression framework. We control for firm and analyst characteristics that previous literature has shown to affect analyst performance. 13 Our controls for firm characteristics include size (logmv), market-to-book (mb), intangible assets (intangible), the number of analysts that cover the firm (nanalyst), stock turnover (turnoverpre12) and return standard deviation (retstd12) over the previous year. The analyst characteristics include firm experience (firmex), general experience (genex), number of firms covered by the analyst (nticker), and the size of the brokerage house (brsize). In addition, since target price is frequently updated, we add the level of target price revision from the previous target price ( TP2TP), and to control for the effect of momentum, we add the cumulative stock return during the target-price revision period (return_rev). Finally, we add the previous 12-month cumulative MSCI World-Index return (MSCI12), since the average degree of optimism is correlated with the overall market return, as demonstrated in Figure 2. We control additionally for time and industry effects by including forecast year, 13 Bradshaw, Brown and Huang (2012) use size and price standard deviation in their study of target price accuracy (which is inversely related to target price optimism) in the US, Clement (1999) shows that earnings forecast accuracy is related to analyst characteristics, and Bonini et al. (2011) add momentum to the Bradshaw, Brown and Huang (2012) specification. 17

18 month and industry indicator variables. We adjust standard errors for two-way clustering at the firm and year levels to correct for cross-sectional and time-series dependence. Our regression model takes the following general form: (1) where the dependent variable is a proxy for target price optimism. All the variables are defined in Table I. In Equation (1), we suppress subscripts with the understanding that each of the observations involves analyst i s forecast on firm j at day t. The independent variables are measured using information up to the forecast date. Table V reports the results of benchmark regressions on the determinants of target price optimism. For brevity, we do not report coefficients on industry and time dummies. Among the firm characteristics, we first note that firm size is positively associated with ex-post optimism measures of Ratio12 (Column (3)). This finding is consistent with Bradshaw, Brown and Huang (2012), who report that size is negatively related to forecast accuracy. However, size is negatively associated with the ex-ante optimism measures of TP2P and TP2PRank. Since the difference between the ex-ante and the ex-post optimism measures lies in the scaling stock price, one potential explanation for this discrepancy is that the result is driven by the differing association between size and the stock price scalar ex-ante and ex-post of target price issuance; in other words, the scalar effect. We find that this is indeed the case. For the ex-post measures, the correlation between firm size and 1/end-of-12-month price is only -0.02, whereas for the ex-ante measures, the correlation between firm size and 1/con-current market price is The much increased negative correlation of the scalar effect in the ex-ante measures leads to a negative coefficient on firm size in Columns (1) and (2). Among other firm characteristics variables, market to book and number of following analysts are negatively associated with optimism, and intangible assets, stock turnover, and return standard deviation are positively associated with optimism. These results indicate that analysts are less optimistic on growth stocks (perhaps because these stocks already possess high 18

19 prices) and stocks with more peers following (perhaps to stay conservative), but are more optimistic on stocks with more intangible assets, more-liquid stocks, and more-volatile stocks (perhaps because optimistic targets are easier to be met for these stocks). [Table V about here.] In the remaining variables, we observe that the stock price momentum during the target price revision leads to less target price optimism. Further, experience with the firm by the analyst, the number of firms followed at the same time by the same analyst, and the size of the brokerage house that employs the analyst all leads to less optimistic price targets. Since less target price optimism is arguably inversely related to accuracy of analysts, these results are consistent with Clement (1999) that these analyst characteristics inversely affect earnings forecast accuracy. Last, consistent with Figure 2, the overall past 12-month market return is negatively related to ex-ante optimism (MSCI12 in Columns (1) and (2)). However, due to the long-run price reversal over 2-5 years (DeBondt and Thaler 1985), the past 12-month market return becomes positively related to ex-post optimism (MSCI12 in Column (3)). In sum, Table V indicates that our optimism measures are related to a number of benchmark firm characteristics, stock price performance, and analyst characteristics. Based on these benchmark regressions, in subsequent sections we will examine the impacts of conflicts of interest and institutional characteristics on optimism. B. The Impact of Conflicts of Interest, External Financing, and Country Characteristics on Optimism As previously discussed, we are interested in analyst traits that are related to conflicts of interest, firm traits that are related to external financing, and three categories of country characteristics investor protection, financial transparency, and economic development and culture. For each trait, we run the following OLS regression: (2) where denotes an analyst, firm or country characteristic that we are interested in. We include all of the controls in Equation (1) and is the vector of coefficients for those controls. Since many of these characteristics are highly correlated, we add them one at a time to avoid confounding effects. 19

20 Table VI presents the results. For brevity, we do not report the results on the control variables, although we can report that they are virtually identical to those in Table V. We also remind readers that in Table IV, we have presented the univariate comparison of target price optimism based on these traits. It should be pointed out that the coefficient on the variable in Equation (2) essentially captures the same effect as those of the univariate results, with the difference being that Equation (2) is estimated with continuous variables (where applicable) and with controls. [INSERT TABLE VI HERE] Let us discuss the results on analyst traits first. The coefficient on purebroker is significantly positive in two out of four optimism measures; this significance indicates that analysts employed by pure brokers tend to issue over-optimistic price targets. This result is consistent with Cowen, Groysberg and Healy (2006), who find that pure brokers tend to issue more optimistic earnings forecasts in order to generate trading business. The coefficient on underwriter is reliably positive and significant across all of the optimism measures, indicating that analysts employed by the target firms underwriters and comanagers tend to issue overly optimistic target prices. Again, this is consistent with the conflicts of interest hypothesis that analysts who have either direct or indirect business ties tend to issue optimistic target prices. The coefficient on local is significantly positive in one out of four optimism measures. A deeper look at local analysts yields some interesting results. We partition local analysts further into those hired by foreign brokers (expatriate local) and those hired by local brokers (pure local), and create a dummy variable for each type of local analysts denoted as expalocal and purelocal, respectively. We find that the coefficient on expalocal is negative and significant for two out of the four optimism measures, whereas purelocal is significantly positive for three of the four optimism measures. These results suggest that pure locals are more inclined to higher conflicts of interest, perhaps due to the pressure of local social network. Collectively, the effects of these two types of local analysts offset each other, rendering the effect of local to be less significant. Overall, the results with these analyst traits support the conflicts of interest hypothesis over the information advantage hypothesis. 20

21 Turning to financing-related firm traits, we find that both and earnmgmt_firm are strongly positively related to optimism. Between the two components of, is significantly related to all four optimism measures, whereas is significant in three optimism measures. Thus, consistent with Bradshaw et al. (2006), sources of financing plays a role in elevating analyst optimism in our case firms in need of equity financing induces more optimism than firms in need of debt financing as judged by statistical significance. Collectively, these results suggest that indirect investment banking pressures, as evidenced by firms external financing needs, induces target price optimism. Having shown that conflicts of interest, and direct and indirect investment banking pressures may exacerbate analyst optimism, we now turn to institutional characteristics that we argued earlier could mediate optimism. The evidence in Table VI shows that this is indeed the case. Let us discuss investor protection variables first. The first piece of evidence is that common law origin mitigates optimism in three out of four optimism measures. Secondly, we present a number of overall measures of a country s rule and law system: efficiency and integrity of the legal environment (judicial), overall score of legal system and property rights (legsys), the law and order tradition (rule). We find that these overall measures of rule and law are predominantly significantly negatively related to our ex-ante and ex-post optimism measures. We then examine a number of specific aspects of the judicial system, namely, the integrity of the legal system (integrity), court impartiality (judimp), and judicial independence (judindp); and a number of quality measure of investor protection, namely, the degree of non-corruption of the government (corruption), the difficulty of expropriation such as outright confiscation or forced nationalization (expropriate), the inverse of bribery extent (bribe), the protection of property rights (propprot), and the anti-self-dealing index that measures the legal protection of minority shareholders against expropriation by corporate insiders. In sum, the coefficients on these country-level judicial traits are negative and significant, suggesting that higher degree of investor protection helps mitigate forecast optimism. Turning to financial transparency, we find that forecast optimism is inversely related to the degree of accounting disclosure (accnstd), and positively related to the degree of country-level earnings 21

22 management (earnmgmt). The coefficients on Morck et al. s (2000) R 2 are negative but insignificant in the first two columns in which we use ex-ante optimism measures, but positive and significant when we use ex-post optimism measures. Since better accounting disclosure, lower level of a country s earnings management, and a lower R 2 all indicate more financial transparency, these results overall suggest that financial transparency mediates optimism in target prices. Finally, economic and cultural developments by and large help mitigate target price optimism. For the former, we observe that optimism is less severe in developed countries and in countries where U.S. holds a larger percent of its equities. The coefficient on GDP per capita (gdpp), however, is negative but insignificant in three out of four regressions. For cultural variables, we observe that optimism is significantly negatively associated with individualism. The results on media are mixed. In sum, the results in Table VI suggest that analysts conflicts of interest lead to target price optimism. However, some country-level institutional infrastructure such as investor protection, financial transparency, and culture forces could discourage analysts from liberally inflating target prices. The evidence suggests that these institutional infrastructure traits are effective in disciplining analysts from biased research. IV. Robustness Checks A. Mediating Effects at the Firm-Analyst-Year Level In our previous regressions, we pool all observations together. Although we use Petersen s (2009) two-way cluster adjustment of standard errors at the firm and year levels, our results may still tilt towards analysts who provide more forecasts during the year for each firm or for each country. To address this concern, we collapse the observations by the mean values of the variables to the firm-analyst-year level (which allows a maximum of one observation per analyst per firm-year). Doing so reduces the number of observations from the full sample observations of 1,069,879 to 361,980 observations; in other words, each analyst on average issues three target prices per year for a firm. Note that in this sample, we are able 22

23 to use the within-firm ex-ante optimism measure of TPRank; hence we have six optimism measures instead of five optimism measures in previous tables. In untabulated results, we can report that the benchmark regression produces qualitatively the same results as those of the full sample in Table V. We therefore focus on the mediating effects. Table VII presents the results. A simple comparison of Table VII with the full-sample results in Table VI reveals that the mediating effects remain of the same sign and order of magnitude for the five optimism measures presented earlier. Thus, our conclusions that conflicts of interest exacerbate optimism while institutional infrastructure mitigates optimism remain unchanged for the firm-analyst-year sample. [INSERT TABLE VII HERE] The firm-analyst-year sample also offers a chance to examine the performance of the within-firm measure of optimism TPRank. Notably, the mediating effects contained in TPRank are much weaker than the other optimism measures. Given that the range of TPRank remains the same across firms, the result is not surprising: After all, one should discriminate conflicts of interest or infrastructure differences across firms rather than within firms. In this sense, the results on TPRank provide a validity check of the mediating effects of our other optimism measures. B. Mediating Effects in Firms Covered by Both Local and Foreign Analysts The mediating effects of institutional characteristics originate from countries where analysts domicile. A sharper focus of the institutional mediating effects would be to isolate the covered firms, so that one could better differentiate whether optimism across analysts is indeed induced by institutional features. To this end, we examine the mediating effects in firms that are covered by both local and foreign analysts. This subsample leaves us about half of the observations. Implicitly, local (foreign) analysts are subject to local (foreign) laws and rules. Compared to the full-sample results, the institutional mediating effects should be more pronounced in this subsample. Table VIII presents the results. We observe that almost all of the institutional mediating effects in Table V are preserved in Table VIII. Furthermore, the institutional mediating effects in firms covered by both local and foreign analysts are somewhat stronger than those of the full sample in Table V, in that the 23

24 magnitudes of the coefficient estimates tend to be larger. These results offer further support to the hypothesis that better institutional infrastructure mitigates analyst optimism. [INSERT TABLE VIII HERE] For completeness, Table VIII also reports the mediating effects of analyst and firm traits and finds that the results are similar to what was reported for the full sample. Of particular note, local now load significantly and positively for all four optimism measures (as opposed to for just one measure in Table V). As with the institutional characteristics, the local analysts now have a sharpened contrast with foreign analysts once we impose the same firms for them to cover. The result on local in Table VIII thus offers stronger evidence that local analysts are subject more to conflicts of interest than to information advantage. C. Mediating Effects for Non-US Analysts Our full sample is dominated by U.S. analysts, who provide roughly one third of the observations. To examine whether our results are merely driven by U.S. analysts, we remove U.S. analysts from the sample and examine the subsample of non-us analysts only. In untabulated results, we can report that the main inferences are qualitatively the same as those of the full sample. We therefore conclude that our results are not driven by U.S. analysts. 14 D. Mediating Effects of Buy, Hold, and Sell Samples In this subsection, we focus on a more restrictive sample where analysts issue both target prices and stock recommendations. In tiered stock recommendations, buy recommendations are more common and tend to be viewed as over-optimistic. In our sample, requiring non-missing stock recommendations reduces the sample size by 28%. In this reduced sample, buy ( Buy and Strong Buy recommendations), hold, and sell (the Sell and Strong sell ) recommendations each accounts for, respectively, 53%, 36%, and 11% of the total observations, consistent with the previous literature in that sell recommendations are less common. 14 We partition the sample by firm locations and find similar results. 24

25 It is, however, not uncommon to see analyst s target price greater than the current market price even though the analyst issues a sell recommendation. For example, on January 31, 2012, Goldman Sachs issued a sell recommendation on Barclays with a target price higher than Barclay s closing price of January 30, In fact, among the sell recommendations in our sample, 26% of sell recommendations are accompanied with target prices greater than current market prices. Thus, optimism in target prices widely exists even in the pessimistic sell recommendations. We separately examine the mediating effects for the buy, hold and sell subsamples. Table IX presents the results. Overall, we observe that most of the mediating effects regarding analyst, firm, and country characteristics remain unchanged for these subsamples, indicating that the mediating effects exist across the board over different categories of analyst recommendations. We also make a few noteworthy observations. First, for the sell recommendations, the investment-banking related mediating effects are weaker: underwriter, FIN, EQUITY, and earnmgmt_firm no longer have significance on TP2P. This is consistent with the fact that there are fewer investment banking opportunities in the sell firms, and thus less incentive for analysts to issue overly favorable target prices. As expected, the signs of these investment-banking mediating effects are reversed to be positive for the hold and buy recommendations. Second, the sign of TP2P is at times negative for pure brokers and local analysts in the buy and hold samples. These are limited evidence that pure brokers and local analysts are less optimistic. We conjecture that for stocks with buy recommendations, the need to generate brokerage business is less acute, and local analysts can better use their information advantage. Last, in the buy sample, the mediating effect of GDP is no longer negative. Despite these anomalous results, Table IX suggests that our conclusions about the mediating effects of analyst, firm, and country characteristics, in particular, those of the country characteristics, remain unchanged in buy, hold, and sell subsamples. [INSERT TABLE IX HERE] E. The Recent Financial Crisis In Table X, we consider the impact of recent financial crisis on target price optimism. We partition the sample into crisis period and non-crisis periods, where the crisis period is based on the 25

26 NBER recession period of December 2007 to June We observe that for the non-crisis periods, none of our results regarding the analyst, firm, and country mediating effects is changed (except that media is insignificant). [INSERT TABLE X HERE] It is possible that financial crisis may alter the behaviors of financial analysts. For the crisis period, the mediating effects related to financing-related firm traits and country characteristics are not changed, except for the underwriter variable, which remains positive and significant. The mediating effects related to analyst traits are mixed: The signs on purebroker become significantly negative, whereas the sign on local is inconsistent. Nonetheless, our conclusions remain qualitatively the same for both non-crisis and crisis periods for the country-level institutional traits. V. Conclusion Using a unique analyst-location data that include 11,436 analysts covering 16,867 firms, we examine the determinants of target price optimism across 41 countries around the world. We propose both ex-ante and ex-post optimism measures that capture the implied return of the target price and whether or for how long the target price is met over the forecast horizon. We examine the impacts of analyst, firm and country characteristics on target price optimism. We identify local analysts and analysts that are affiliated with lead underwriters, two properties that are implicative of both conflicts of interest and information advantage. We find evidence that conflicts of interest dominate information advantage in analysts issuance of target prices. Consistent with these results, we find that target prices issued by analysts employed by pure brokers are more optimistic, suggesting incentives to generate trading; and that firms with greater external financing needs induce more analyst optimism, suggesting indirect investment banking pressures. We find that analyst optimism is attenuated by capital market development and institutional infrastructure. Capital market development is usually accompanied by both competition and law-andorder. The former promotes optimism while the latter mitigates optimism. We find that country-level characteristics of investor protection, financial transparency, economic development, and culture all 26

27 attenuate rather than exacerbate target price optimism. Importantly, country-level institutional infrastructure seems to be effective in disciplining analysts from liberally inflating target prices arising from potential conflicts of interest. In sum, our findings suggest that analyst optimism in target prices is exacerbated by conflicts of interest and investment banking pressures but is mediated by country-level institutional infrastructure. 27

28 References Arand and Kerl Sell-Side Analysts' Target Prices in the Presence of Conflicts of Interest Asquith, P., M. Mikhail, and A. Au Information content of equity analyst reports. Journal of Financial Economics 75: Bandyopadhyay, S., L. Brown, and G. Richardson Analysts Use of Earnings Forecasts in Predicting Stock Returns: Forecast Horizon Effects. International Journal of Forecasting, 11 (3): Barniv, R., O.K. Hope, M. Myring and W.B. Thomas Do analysts practice what they preach and should investors listen? Effects of recent regulations. The Accounting Review 84(4): Barniv, R., O.K. Hope, M. Myring and W.B. Thomas International evidence on analyst stock recommendations, valuations, and returns. Contemporary Accounting Research 27 (4). Bilinski, P., D. Lyssimachou and M. Walker Target price optimism: international evidence. Working paper. Bonini, S., L. Zanetti, R. Bianchini and A. Salvi Target price optimism in equity research. Journal of Business Finance and Accounting 37: Bradshaw, M The Use of Target Prices to Justify Sell-Side Analysts Stock Recommendations, Accounting Horizons, Vol. 16, No. 1, pp Bradshaw, M., and L. Brown Do sell-side analysts exhibit differential target price forecasting ability? Working paper, Harvard Business School and Georgia State University. Brav, A., and R. Lehavy An empirical analysis of analysts' target prices: short-term informativeness and longterm dynamics. Journal of Finance 58: Bushman, R., J. Piotroski, and A. Smith, What determines corporate transparency? Journal of Accounting Research 42: Bushman, R., J. Piotroski, and A. Smith Insider Trading Restrictions and Analysts Incentives to Follow Firms. Journal of Finance 60: Chen, T., and X. Martin, Do Bank-Affiliated Analysts Benefit from Lending Relationships? Journal of Accounting Research. 49 (3): Clarke, J., and A. Subramanian Dynamic forecasting behaviour by analysts: Theory and evidence. Journal of Financial Economics 80: Clarke, J., S. Ferris, N. Jayaraman and J. Lee Are analyst recommendations biased? Evidence from corporate bankruptcies. Journal of Financial and Quantitative Analysis 41: Clement, M Analyst forecast optimism: Do ability, resources, and portfolio complexity matter? Journal of Accounting and Economics 27: Clement, M., and S. Tse Financial analyst characteristics and herding behavior in forecasting. Journal of Finance 60: Cowen, A., B. Groysberg, and P. Healy Which Types of Analyst Firms Are More Optimistic? Journal of Accounting and Economics, 41 (1-2): Da, Z. and E. Schaumburg Relative valuation and analyst target price forecasts. Journal of Financial Markets 14: Da. Z., K. Hong, and S. Lee Where does the investment value of target prices come from? Working paper, University of Notre Dame. Dasgupta, S., J. Gan, and N. Gao Transparency, Price Informativeness, and Stock Return Synchronicity: Theory and Evidence. Journal of Financial and Quantitative Analysis, 45:

29 Djankov, S., R. La Porta, F. Lopez-de-Silanes, and A. Shleifer The Law and Economics of Self-Dealing. Journal of Financial Economics 88: Doidge, C., A. Karolyi, and R. Stulz Why Do Countries Matter so much for Corporate Governance? Journal of Financial Economics 86: Duan, Y., E. Hotchkiss, and Y. Jiao Business Ties and Information Advantage: Evidence from Mutual Fund Trading. Journal of Business and Finance & Accounting, Vol.37, Issue 9-10, pp Dugar, A., and S. Nathan The effect of investment banking relationships on financial analysts' forecasts and investment recommendations. Contemporary Accounting Research 11: Francis, J. and L. Soffer The Relative Informativeness of Analysts Stock Recommendations and Earnings Forecast Revisions, Journal of Accounting Research, Vol. 35, No. 2, pp Francis, J., D. Hanna, and D. Philbrick Management communications with securities analysts. Journal of Accounting and Economics 24: Gleason, C., and C. Lee Analyst forecast revisions and market price discovery. The Accounting Review 78: Gleason, C., W. B. Johnson, and H. Li Valuation Model Use and the Price Target Performance of Sell-Side Equity Analysts. Working paper, University of Iowa. Hong, H., J. Kubik, and D. Solomon Security analysts' career concerns and herding of earnings forecasts. Rand Journal of Economics 31: Huang, J., G. M. Mian, and S. Sankaraguruswamy The value of combining the information content of analyst recommendations and target prices, Journal of Financial Markets 12: Jacob, J., T. Lys and M. Neale Expertise in forecasting performance of security analysts. Journal of Accounting and Economics 28: Kadan, O., L. Madureira, R. Wang, and T. Zach Conflicts of Interest and Stock Recommendations: The Effects of the Global Settlement and Related Regulations, Review of Financial Studies 22: Kerl, A. G Target Price Optimism, BuR - Business Research 4: La Porta, R., F. Lopez-de-Silanes, A. Shleifer, and R. Vishny Law and finance, Journal of Political Economy 106: Lang, M., K. Lins and D. Miller Concentrated control, analyst following, and valuation: Do analysts matter most when investors are protected least? Journal of Accounting Research 42 (3), Leone, A., and J. S. Wu What does it take to become a superstar? Evidence from Institutional Investor rankings of financial analysts. Working paper, University of Rochester. Lin, H., and M. McNichols Underwriting relationships, analysts' earnings forecasts and investment recommendations. Journal of Accounting and Economics 25: Lyssimachou, D., E. Lee, and M. Walker How do analyst characteristics and affiliation influence the quality of their target price forecasts? Working paper, Manchester Business School. Michaely, R., and K. Womack Conflict of interest and the credibility of underwriter analyst recommendations. Review of Financial Studies 12: Mikhail, M., B. Walther and R. Willis Do security analysts improve their performance with experience? Journal of Accounting Research 35: Morck, R., B. Yeung, and W. Yu The Information Content of Stock Markets: Why do Emerging Markets Have Synchronous Stock Price Movements? Journal of Financial Economics 58: O'Brien, P., M. McNichols and H. Lin Analyst impartiality and investment banking relations. Journal of Accounting Research 43:

30 Petersen, M Estimating standard errors in finance panel data sets: comparing approaches. Forthcoming in the Review of Financial Studies. Stickel, S Reputation and performance among security analysts. Journal of Finance 47:

31 31

32 32

33 Table I Variable Definitions This table describes variables used in the analyses. We obtain analyst related data from the I/B/E/S, and financial and stock trading data from Compustat, unless specified otherwise. We winsorize the continuous variables at 1 and 99 percentiles over the full sample. Variable Optimism measures Definition and data source TP2P Target price level, computed as (TP/P 1 ) TP2PRank Ratio12Ratio12 The rank of TP2P, coded from 0 to 99, within its 2-digt SIC-code industry in any given year The percentage of trading days in the next 12 months that stock price is smaller than TP. TPNOTMET A dummy variable that equals one if the maximum stock price over the next 12 months is smaller than TP. TPRank The ranks of target prices for a given firm in a given year, coded from 0 to 100. This measure is based on Ke and Yu (2006). Control variables logmv mb intangible nanalyst turnover12 retstd12 The logarithm of market cap as the product between the stock price and shares outstanding. The market to book ratio of a firm The ratio of intangible asset to current asset The logarithm of the number of analyst following the firm in the previous year Average stock turnover of a firm in the past 12 months Standard deviation of daily stock return for a firm in the past 12 months TP2TP Target price revision computed as (TP TP -1 )/ TP -1, where TP -1 is the previous TP issued by the same analyst return_rev Cumulative stock return during the target-price revision period. MSCI12 firmex genex The previous 12-month cumulative MSCI World-Index Return from Morgan Stanley Inc. Time interval in years since an analyst provides the first forecast for the target firm. Time interval in years since an analyst first forecast in the I/B/E/S detail files for any firm nticker Number of firms that analyst a covers in the year prior to firm f s IPO date. We use I/B/E/S detail files. We use the natural log form of this variable in our multivariate regressions. brsize Brokerage size, defined as the logarithm of the number of analysts working for the I/B/E/S brokerage that an analyst is associated with in a year. crisis Dummy variable that equals one for the recent financial crisis during Dec to July 2009 based on NBER. Analyst traits purebroker underwriter local Purelocal Expalocal A dummy variable that equals one if the brokerage house that employs an analyst is a pure broker that has no investment banking business, and zero otherwise. A dummy variable that equals one if an analyst s employer served as either lead underwriter or co-manager for the covered firm in the past three years based on equity and debt offering from Thomson One Banker, and zero otherwise. A dummy variable that equals one if an analyst resides in the same country as the target firm s headquarter country, and zero otherwise. Analyst location data are from annual volumes of the Nelson's Directory of Investment Research. A dummy variable that equals one if a local analyst works for a local broker. Analyst and brokerage location data are from annual volumes of the Nelson s Directory of Investment Research. A dummy variable that equals one if a local works for a foreign broker. Analyst and brokerage location data are from annual volumes of the Nelson s Directory of Investment Research. 33

34 Firm traits related to external financing and earnings management Financing variable of Bradshaw, Richardson, and Sloan (2006). The measure is equal to change in equity plus change in debt. Change in equity following Bradshaw, Richardson and Sloan (2006). It is computed as sale of common and preferred stock minus purchase of common and preferred stock minus cash dividends paid, scaled by total assets. Change in debt, computed as long-term debt issuance minus long-term debt reduction minus current debt changes, scaled by total assets. earnmgmt_firm Firm-level earnings management level, measured by the absolute value of discretionary accruals from the Modified Jones Model. Country characteristics commonlaw judicial legsys rule integrity judimp judindp corruption expropriat bribes A dummy variable equal to one if the legal origin of a country is common law, and zero otherwise. The raw data are from La Porta et al. (1998). Assessment of the efficiency and integrity of the legal environment as it affects business, particularly foreign firms produced by the country-risk rating agency Business International Corporation. The data are from La Porta et al. (1998). Overall score of legal system & property rights. The data are from the Economic Freedom Dataset by Fraser Institute. Assessment of the law and order tradition in the country produced by the country -risk rating agency International Country Risk (ICR). The data are from La Porta et al. (1998). Integrity of the legal system. The data are from the Economic Freedom Dataset by Fraser Institute. Impartial courts. The data are from the Economic Freedom Dataset by Fraser Institute. Judicial independence. The data are from the Economic Freedom Dataset by Fraser Institute. ICR s assessment of the corruption in government. A higher value indicates less corruption. The data are from La Porta et al. (1998). ICR s assessment of the risk of outright confiscation or forced nationalization. A high level of this variable represents relatively low risk of expropriation. The data are from La Porta et al. (1998). Extra payments/bribes/favoritism. Larger value means lower extent of bribery. propprot Protection of property rights. The data are from the Economic Freedom Dataset by Fraser Institute. anti_dealing Average of ex-ante and ex-post private control of self-dealing from Djankov, La Porta, Lopezde-Silanes, and Shleifer (2008). accntstd Index of accounting disclosure for a country. The data are from La Porta et al. (1998). earnmgmt Index of country-level earnings management from Leuz, Nanda, and Wysocki (2003). A larger value indicates a higher degree of earnings management. R 2 R 2 estimated from the market model for a countryfrom Morck, Yeung, and Yu (2000). GDPP developed usholdgdp idv GDP per capita from World Development Indicator. Dummy variable equal to one if country i is a developed country, and zero otherwise. The data are from Standard and Poor's Global Stock Markets Factbooks A country s U.S. holdings of equity scaled by local market GDP based on US security holding data from the U.S. Treasury Department. Hofstede s (2001) cultural index on Individualism for a country. media The average rank of a country s per capita number of newspapers and televisions during 1993 to 1995 as reported by World Development Indicators. We obtain the data from Bushman, Piotroski, and Smith (2005). 34

35 Table II Summary Statistics and Correlations of Optimism Measures This table provides the summary statistics and correlations of the optimism measures in our final sample. Refer to Table I for the definitions of the optimism measures. TPRank is calculated at the average level of firm-analyst-year (i.e., only one observation is allowed for each firm-analyst pair every year) and hence its number of observation is smaller. The correlations of TPRank with other variables are also calculated at the firm-analyst-year level. Panel A: Summary Statistics N Mean Std Min Median Max TP2P 1,069, TP2PRank 1,069, TPRank 312, Ratio12 976, TPNOTMET 976, Panel B: Correlation Matrix TP2P TP2PRank TPRank Ratio12 TP2P 1 TP2PRank TPRank Ratio TPNOTMET

36 Table III Final Sample by Country and Year This table provides the mean of key variables of our final sample by analyst country in Panel A and by year in Panel B. In Panel A, country firm number refers to the number of firms headquartered in a given country, and analyst number refers to the number of analysts domiciled in a given country. In Panel B, country number refers to the number of countries that the analysts are domiciled. For the definitions of all other variables, refer to Table I. Panel A Sample distribution by country Country Firm Analyst TP2P Ratio12 TPNOTMET Argentina Australia Belgium Brazil Canada 1, Chile China Denmark Egypt England 1,176 1, Finland France Germany Greece Hong Kong India Indonesia Ireland Israel Italy Japan 1, Korea Malaysia Mexico Netherlands New Zealand Norway Pakistan Peru Philippines Portugal Russia Singapore South Africa Spain Sweden Switzerland Taiwan Thailand Turkey USA 5,427 4, Mean

37 Panel B Sample distribution by year Year Obs. Analyst Firm Country TP2P Ratio12 TPNOTMET ,251 3,444 4, ,539 4,101 6, ,334 5,207 7, ,353 5,604 8, ,348 6,035 9, ,965 6,101 9, ,077 5,844 10, ,577 5,401 9, ,356 4,918 9, ,079 2,632 7, Mean 106,988 4,929 8,

38 Table IV Univariate Comparison of Target Price Optimism This table presents univariate comparisons of target price forecast optimism conditional on analyst / firm traits and analyst-country traits. For each trait in the first column, we show the mean differences and their t-statistics in various optimism measures between the high and low groups based on the partition of the trait. The non-country variables (analyst and firm traits) are ordered every year for every analyst location country. The country variables (other than binary variables) are ordered every year for the cross -section of sample countries, with one observation for a country every year. An observation greater (smaller) than the median value based on the above ordering is treated as high (low). The t-statistics for the differences are two-way adjusted at firm and year levels. ***, **, and * denote significance at the 1, 5, and 10 percent levels, respectively. Predicted sign TP2P TP2PRank Ratio12 TPNOTMET Analyst/firm traits purebroker +/? ** (-0.89) (2.26) (-0.18) (0.39) underwriter +/? 0.012** 1.747*** (2.51) (4.75) (0.93) (0.24) local +/? 0.018*** 2.958*** 0.026** 0.022** (3.05) (6.86) (2.30) (2.10) *** 4.296*** 0.029*** 0.040*** (7.80) (12.61) (4.45) (4.71) *** 5.436*** 0.038*** 0.047*** (9.68) (14.29) (5.80) (6.33) *** 1.193*** 0.015*** (4.31) (3.56) (2.62) (1.12) earnmgmt_firm *** 1.337*** (6.22) (5.72) (-0.21) (0.94) Investor protection variables commonlaw * ** (-0.87) (-1.87) (-1.27) (-2.45) judicial ** *** *** (-1.98) (-5.17) (-1.31) (-3.44) legsys *** *** ** *** (-3.81) (-5.10) (-1.99) (-5.25) rule *** *** *** (-4.50) (-8.26) (-1.20) (-3.55) integrity *** (-0.29) (-2.65) (-1.40) (-1.25) judimp *** (-1.28) (-6.58) (-0.47) (-0.91) judindp *** (-0.06) (-7.10) (0.86) (0.75) corruption *** *** * *** (-2.91) (-5.50) (-1.70) (-4.58) expropriat *** *** ** (-2.80) (-5.27) (-0.42) (-2.33) bribes *** (-0.37) (-6.25) (1.04) (1.02) propprot ***

39 (-0.60) (-8.13) (0.77) (0.53) anti_dealing ** *** ** *** (-2.11) (-2.94) (-2.34) (-3.79) Financial transparency variables accntstd *** *** *** (-3.06) (-5.83) (-1.16) (-2.84) earnmgmt ** *** (-0.03) (2.07) (1.22) (3.21) r *** 0.033** 0.062*** (0.99) (3.16) (2.18) (4.46) Economic development and culture variables gdpp *** *** *** (-4.40) (-8.21) (-1.24) (-3.57) developed *** *** ** (-2.91) (-5.16) (-0.85) (-2.36) usholdgdp ** (0.15) (-2.39) (-0.48) (0.01) idv *** * *** (-1.56) (-4.23) (-1.90) (-3.90) media * ** * *** (-1.94) (-2.46) (-1.75) (-7.17) 39

40 Table V Determinants of Optimism: Benchmark Regressions This table reports results of pooling regressions on the determinants of target price optimism for our full sample. All models are estimated using OLS regression except for Column (4) in which we use logit regression. We define the variables in Table I. All of the models include year, month, and industry indicators; for brevity, we do not report these coefficients. We report robust t-statistics adjusted for two-way clustering at both firm and year levels in parentheses. ***, **, and * denote significance at the 1, 5, and 10 percent levels, respectively. (1) (2) (3) (4) TP2P TP2PRank Ratio12 TPNOTMET logmv *** *** 0.008** (-4.80) (-5.14) (1.97) (1.40) mb *** *** * (-2.96) (-4.64) (-1.34) (-1.93) intangible * (-0.80) (-1.08) (1.91) (-0.45) nanalyst ** *** *** (-2.42) (-1.35) (-3.71) (-7.72) turnoverpre *** 1.304*** (3.42) (4.15) (0.35) (-0.73) retstd *** ** (4.85) (2.22) (-0.36) (-0.16) TP2TP 0.471*** *** 0.251*** 1.585*** (6.64) (7.97) (6.24) (5.37) return_rev *** *** *** *** (-7.23) (-9.18) (-9.42) (-11.26) msci *** *** 0.299*** 1.351*** (-5.09) (-5.30) (5.72) (4.18) firmex *** *** (-3.86) (-3.21) (-0.29) (-1.39) genex 0.002*** 0.184*** (4.35) (5.32) (-0.03) (-1.35) nticker ** * (-2.42) (-1.91) (-0.25) (-1.18) brsize *** *** *** *** (-8.07) (-13.63) (-5.91) (-5.95) Constant 0.479*** *** 0.914*** 0.667*** (11.89) (25.69) (21.70) (3.37) Observations 1,046,486 1,046, , ,154 (Pseudo) R-squared

41 Table VI Mitigating Effects of Analyst Traits, Firm Traits, and Country Characteristics The first column lists the analyst, firm, or analyst-country traits. Each row represents a regression, where the trait in the first column is added to the corresponding benchmark regression in Table V. The t-statistics are two-way adjusted for clustering at the firm and year levels. ***, **, and * denote significance at the 1, 5, and 10 percent levels, respectively. (1) (2) (3) (4) Predicted sign TP2P TP2PRank Ratio12 TPNOTMET purebroker +/? ** *** (-0.10) (2.07) (1.42) (3.13) underwriter +/? 0.031*** 3.343*** 0.026*** 0.121*** (7.44) (11.94) (5.41) (5.29) local +/? ** (-1.28) (1.36) (2.42) (1.61) purelocal? 0.018*** 3.062*** 0.018** (3.83) (6.66) (2.52) (1.20) expalocal? *** *** (-6.65) (-10.84) (-1.49) (-0.36) Firm traits *** *** 0.178*** 1.032*** (9.43) (18.37) (7.17) (7.06) *** *** 0.233*** 1.356*** (8.99) (17.42) (6.62) (7.97) *** 8.624*** 0.056** (4.62) (5.38) (2.44) (1.64) earnmgmt_firm *** *** 0.079** 0.721*** (6.35) (7.78) (2.30) (3.68) Investor protection variables commonlaw * * *** (-1.15) (-1.92) (-1.95) (-3.42) judicial *** *** *** *** (-5.20) (-9.31) (-3.94) (-7.26) legsys *** *** *** *** (-3.93) (-6.93) (-2.74) (-3.96) rule *** *** *** *** (-3.16) (-5.49) (-2.85) (-5.18) integrity ** * (-1.21) (-2.29) (-0.74) (-1.76) judimp *** *** *** *** (-6.72) (-10.75) (-4.02) (-5.05) judindp *** *** *** *** (-5.78) (-7.80) (-3.03) (-4.95) corruption *** *** *** *** (-2.98) (-5.92) (-2.70) (-4.37) expropriat ** *** ** *** (-2.20) (-3.77) (-1.97) (-3.85) bribes *** *** ** (-4.68) (-5.98) (-1.63) (-2.25) propprot *** *** ** *** (-3.63) (-5.58) (-2.48) (-3.13) anti_dealing *** *** * * 41

42 (-3.68) (-4.03) (-1.70) (-1.92) Financial transparency variables accntstd *** *** *** *** (-4.56) (-7.00) (-4.01) (-5.86) earnmgmt ** 0.002*** 0.018*** (0.24) (2.27) (3.06) (5.33) r *** 1.389*** (-1.00) (-0.27) (2.63) (4.55) Economic development and culture variables gdpp *** (-0.60) (-1.46) (-0.51) (-2.66) developed *** *** ** *** (-3.53) (-5.70) (-2.21) (-3.58) usholdgdp ** *** ** * (-2.42) (-4.33) (-2.00) (-1.92) idv ** *** *** *** (-2.10) (-4.36) (-3.44) (-5.71) media *** (0.61) (-0.20) (-0.34) (-3.25) 42

43 Table VII Mitigating Effects of Analyst Traits, Firm Traits, and Country Characteristics at the Firm- Analyst-Year Level This table reports the regression results for the Firm-Analyst-Year sample in which all observations are averaged per analyst-firm-year combination. Each row represents a regression, in which each of the traits in the first column is added to benchmark specification of Table V. The Firm-Analyst-Year benchmark regression results remain qualitatively the same as those of the full sample results in Table V and are not reported for brevity. The first column lists the analyst, firm, or analyst-country traits. The t-statistics are two-way adjusted at firm and year levels. ***, **, and * denote significance at the 1, 5, and 10 percent levels, respectively. (1) (2) (3) (4) (5) TP2P TP2PRank TPRank Ratio12 TPNOTMET Analyst traits purebroker * 0.547* (-1.62) (1.93) (1.78) (1.13) (1.63) underwriter 0.038*** 4.597*** 1.196*** 0.033*** 0.134*** (9.74) (12.07) (3.10) (6.84) (6.41) local * 0.761* 1.708*** 0.018*** (-1.79) (1.70) (7.57) (3.21) (1.21) purelocal 0.018*** 3.461*** 1.659*** 0.027*** 0.094* (3.89) (7.30) (6.25) (3.74) (1.91) expalocal *** *** ** * (-6.21) (-10.21) (-1.62) (-2.43) (-1.75) Financing-related firm traits XFIN 0.265*** *** 2.543*** 0.167*** 1.108*** (9.72) (20.82) (3.24) (6.94) (8.44) EQUITY 0.339*** *** 3.601*** 0.215*** 1.421*** (9.68) (17.83) (3.39) (6.22) (8.79) DEBT 0.101*** 9.608*** ** (5.29) (7.08) (0.48) (2.52) (1.53) earnmgmt_firm 0.224*** *** *** 0.799*** (6.07) (7.96) (0.73) (2.62) (4.93) Investor protection variables commonlaw * (1.15) (-0.45) (-1.54) (-1.12) (-1.72) judicial *** *** *** *** (-5.50) (-9.44) (-1.37) (-3.77) (-4.79) legsys *** *** ** *** *** (-5.38) (-7.78) (-2.24) (-3.07) (-6.30) rule *** *** ** *** (-2.89) (-5.38) (-1.39) (-2.42) (-4.37) integrity ** ** (-1.58) (-2.47) (-1.30) (-0.74) (-2.17) judimp *** *** * *** *** (-7.49) (-10.60) (-1.95) (-3.95) (-5.26) judindp *** *** * *** *** (-7.37) (-9.75) (-1.68) (-3.38) (-7.04) 43

44 corruption *** *** ** *** *** (-4.14) (-6.74) (-2.55) (-3.01) (-5.02) expropriat *** ** (-1.61) (-3.26) (-1.34) (-1.33) (-2.54) bribes *** *** ** *** (-6.54) (-7.05) (-1.14) (-2.06) (-4.23) propprot *** *** *** *** (-3.47) (-4.64) (-1.29) (-2.60) (-3.99) anti_dealing *** *** * (-2.98) (-3.21) (-0.88) (-1.75) (-1.22) Financial transparency variables accntstd *** *** * *** *** (-4.72) (-7.28) (-1.77) (-3.88) (-4.88) earnmgmt ** 0.001** 0.013*** (-0.68) (1.30) (2.20) (2.33) (3.10) r ** *** ** (-2.36) (-1.37) (5.25) (1.45) (2.20) gdpp (-0.17) (-0.66) (-0.73) (1.07) (-0.92) developed *** *** ** ** *** (-3.92) (-5.80) (-2.07) (-2.10) (-3.06) usholdgdp *** *** *** ** (-3.22) (-4.96) (-0.50) (-2.62) (-2.44) idv *** ** *** *** (-1.09) (-3.29) (-2.47) (-2.76) (-3.36) media 0.000* ** ** (1.80) (0.78) (-2.52) (0.98) (-2.15) 44

45 Table VIII Mitigating Effects in Firms Covered by both Local and Foreign Analysts This table shows the mediating effects for the firms covered by both local and foreign analysts. The sample is reduced by half from 1,069,879 observations in the full sample to 554,247 observations. Each row represents a regression, in which each of the traits in the first column is added to benchmark specification of Table V. The benchmark regression results using this sub-sample remain qualitatively the same as those of the full sample results in Table V and are not reported for brevity. The first column lists the analyst, firm, or analyst-country traits. The t- statistics are two-way adjusted at firm and year levels. ***, **, and * denote significance at the 1, 5, and 10 percent levels, respectively. (1) (2) (3) (4) TP2P TP2PRank Ratio12 TPNOTMET Analyst traits purebroker *** (-0.61) (1.11) (0.97) (3.07) underwriter 0.024*** 2.741*** 0.028*** 0.103*** (5.15) (8.89) (6.77) (4.01) local 0.012*** 2.157*** 0.015*** 0.059** (3.18) (5.97) (2.67) (1.99) Financing-related firm traits XFIN 0.243*** *** 0.188*** 1.016*** (8.34) (14.62) (6.08) (5.12) EQUITY 0.294*** *** 0.232*** 1.200*** (7.68) (12.80) (5.21) (5.69) DEBT 0.149*** *** 0.090*** 0.587** (3.57) (3.73) (3.11) (2.03) earnmgmt_firm 0.145*** *** *** (4.39) (3.91) (1.24) (2.92) Investor protection variables commonlaw ** *** *** (-2.17) (-2.76) (-1.42) (-2.99) judicial *** *** *** *** (-5.18) (-8.54) (-3.98) (-6.84) legsys *** *** *** *** (-5.01) (-8.53) (-2.98) (-4.84) rule *** *** *** *** (-3.16) (-5.23) (-2.79) (-5.32) integrity ** ** (-0.98) (-2.54) (-0.59) (-2.23) judimp *** *** *** *** (-6.53) (-11.00) (-4.22) (-6.12) 45

46 judindp *** *** *** *** (-6.20) (-9.07) (-3.35) (-6.07) corruption *** *** ** *** (-3.16) (-5.94) (-2.28) (-4.25) expropriat *** *** ** *** (-2.75) (-4.32) (-2.13) (-3.79) bribes *** *** ** *** (-5.36) (-7.43) (-2.46) (-3.71) propprot *** *** *** *** (-6.04) (-9.98) (-3.36) (-5.64) anti_dealing *** *** (-3.65) (-3.32) (-1.13) (-1.05) Financial transparency variables accntstd *** *** *** *** (-5.47) (-8.81) (-3.59) (-5.98) earnmgmt ** 0.001*** 0.017*** (0.21) (2.36) (3.08) (6.37) r ** 1.417*** (-0.16) (0.67) (2.51) (4.81) gdpp *** ** *** (-1.36) (-2.82) (-2.30) (-3.87) developed *** *** ** *** (-3.49) (-5.40) (-2.13) (-3.23) usholdgdp ** *** (-2.41) (-5.11) (-0.54) (-1.55) idv ** *** *** *** (-2.23) (-4.56) (-2.60) (-4.71) media *** (-0.41) (-1.19) (-1.18) (-4.09) 46

47 Table IX Mitigating effects of Buy, Hold, and Sell Samples This table shows the mitigating effects of analyst/firm and analyst-country traits for the buy, hold, and sell samples, respectively. We define buy, hold and sell based on the stock recommendation issued concurrently with the target price by the same analyst. We classify Strong Buy and Buy recommendations as buy, and Strong Sell and Sell recommendations as sell. The first column lists the analyst or analyst country traits. Each row represents a regression, in which each of the traits in the first column is added to benchmark specification of Table V. The first column lists the analyst, firm, or analyst-country traits. The t-statistics are two-way adjusted at firm and year levels. ***, **, and * denote significance at the 1, 5, and 10 percent levels, respectively. Sell sample Hold sample Buy sample TP2P Ratio12 TP2P Ratio12 TP2P Ratio12 Analyst traits purebroker 0.015*** 0.019** ** (2.81) (2.16) (-1.43) (0.42) (-2.21) (0.25) underwriter *** 0.006** 0.011* 0.021*** 0.010* (0.48) (2.77) (2.13) (1.76) (5.57) (1.83) local *** *** (0.12) (0.14) (-4.38) (-1.47) (-3.81) (-0.34) Financing-related firm traits XFIN *** 0.097*** 0.150*** 0.213*** 0.082*** (1.27) (5.25) (6.64) (5.32) (8.58) (4.24) EQUITY *** 0.121*** 0.190*** 0.277*** 0.115*** (0.29) (5.12) (7.37) (4.61) (8.41) (4.53) DEBT 0.055** 0.054* 0.047*** 0.058** 0.067*** (2.14) (1.89) (2.69) (2.12) (3.04) (-0.27) earnmgmt_firm ** 0.087*** 0.084** 0.208*** (0.02) (2.42) (5.28) (2.25) (6.87) (0.39) Investor protection variables commonlaw * ** (-1.93) (-2.16) (0.28) (-1.46) (0.18) (-1.40) judicial *** *** *** *** * (-6.81) (-4.21) (-6.59) (-2.89) (-1.66) (-1.47) legsys *** * *** * *** ** (-4.10) (-1.66) (-5.42) (-1.78) (-2.64) (-2.12) rule *** ** * (-4.09) (-1.54) (-2.36) (-1.72) (-0.50) (-0.90) integrity ** *** (-2.26) (-0.58) (-2.81) (-0.24) (-1.25) (-0.94) judimp *** ** *** *** *** *** (-3.24) (-2.25) (-6.51) (-2.60) (-5.04) (-2.86) judindp *** * *** *** ** (-3.46) (-1.84) (-5.30) (-1.32) (-4.45) (-2.30) corruption ** * (-2.01) (-0.96) (-1.82) (-1.01) (-0.81) (-1.40) expropriat *** ** *** (-5.77) (-2.02) (-3.61) (-1.45) (0.32) (-0.21) bribes *** *** *** (-3.08) (-0.54) (-5.44) (-0.51) (-3.39) (-1.01) 47

48 propprot *** *** ** (-3.02) (-1.41) (-4.63) (-1.48) (-2.11) (-1.54) anti_dealing *** *** *** (-3.64) (-1.48) (-3.23) (-1.08) (-3.06) (-1.07) Financial Transparency variables accntstd *** *** *** ** *** ** (-3.97) (-4.80) (-4.01) (-2.26) (-2.80) (-2.08) earnmgmt 0.001* 0.003*** ** ** 0.001* (1.74) (3.95) (-1.07) (2.08) (-2.17) (1.66) r ** ** *** (1.27) (1.98) (-0.78) (2.57) (-1.59) (2.75) Economic development and culture variables gdpp *** *** *** * 0.000** (-6.24) (-2.91) (-3.47) (-1.20) (1.86) (2.53) developed *** *** (-3.07) (-0.99) (-3.73) (-1.49) (-0.81) (-0.65) usholdgdp * * (1.21) (0.91) (-0.20) (0.40) (-1.96) (-1.92) idv ** ** *** * (-2.15) (-1.99) (-0.75) (-2.60) (0.47) (-1.83) media *** ** * * (-3.88) (-2.36) (-1.89) (-0.93) (1.91) (0.22) 48

49 Table X Robustness of the Mitigating Effects during the Financial Crisis This table shows the mitigating effects for the recent financial crisis period (Dec to June 2009) and the non - crisis period. Each row represents a regression, in which each of the traits in the first column is added to benchmark specification of Table V. The benchmark regression results using each of the sub-samples remain qualitatively the same as those of the full sample results in Table V and are not reported for brevity. The t-statistics are two-way adjusted at firm and year levels. ***, **, and * denote significance at the 1, 5, and 10 percent levels, respectively. Non-crisis Periods Crisis Period TP2P Ratio12 TP2P Ratio12 Analyst traits purebroker *** * (0.88) (1.51) (-4.97) (-1.94) underwriter 0.027*** 0.028*** 0.049*** 0.025*** (7.14) (4.69) (13.22) (10.73) local ** *** (-0.42) (2.25) (-3.22) (0.63) Financing-related firm traits XFIN 0.216*** 0.187*** 0.339*** 0.125*** (14.12) (6.62) (10.26) (4.94) EQUITY 0.284*** 0.246*** 0.435*** 0.160*** (12.68) (6.41) (12.23) (3.20) DEBT 0.081*** 0.056* 0.135** 0.058*** (4.73) (1.86) (2.02) (4.31) earnmgmt_firm 0.160*** 0.079* 0.301*** 0.071*** (10.72) (1.72) (5.57) (5.64) Investor protection variables commonlaw *** (-1.49) (-1.44) (0.16) (-4.14) judicial *** *** *** (-4.96) (-3.66) (-5.89) (-1.31) legsys *** *** ** (-4.02) (-3.29) (-2.57) (-0.10) rule *** *** (-3.78) (-2.91) (-1.31) (-0.02) integrity (-0.96) (-1.10) (-1.10) (1.13) judimp *** *** *** * (-6.84) (-4.30) (-3.59) (-1.72) judindp *** *** *** (-5.00) (-3.15) (-5.47) (-1.52) corruption *** *** (-3.89) (-2.89) (-0.74) (0.25) expropriat ** ** *** (-2.39) (-2.06) (-1.20) (2.67) bribes *** * *** (-3.76) (-1.73) (-3.82) (-0.50) propprot *** ** *** (-3.47) (-2.55) (-3.45) (-0.95) 49

50 anti_dealing *** *** *** (-2.93) (-1.04) (-5.57) (-10.91) Financial transparency variables accntstd *** *** *** *** (-3.73) (-3.38) (-6.53) (-10.08) earnmgmt ** *** (0.99) (2.56) (-0.81) (2.85) r *** ** *** (-0.05) (3.59) (-2.19) (-3.27) Economic development and culture variables gdpp ** (-0.72) (-0.44) (-1.05) (2.14) developed *** ** ** (-3.92) (-2.53) (-2.54) (1.04) usholdgdp ** ** (-2.32) (-2.21) (-0.90) (-0.60) idv *** *** * (-2.89) (-3.78) (0.23) (1.89) media (0.32) (-0.81) (0.36) (1.16) 50