Broker Recommendations and Australian Small-Cap Equity Fund Management

Carole Comerton-Fordea, David R. Gallagherb, Joyce Laic and Terry Walterd

b c d

Discipline of Finance, The University of Sydney

School of Finance and Economics, University of Technology, Sydney School of Banking and Finance, The University of New South Wales School of Finance and Economics, University of Technology, Sydney

This study examines the extent to which the abnormal performance of active Australian small-cap equity fund managers is derived from broker recommendations. This research is motivated given the economic and statistical significance of the alphas reported in previous studies, as well as the small-cap equity market exhibiting a high degree of information asymmetry. Our research links small-cap equity fund performance with how important broker recommendations are in generating risk-adjusted performance. Our empirical evidence concerning the investment value of broker recommendations is supported, where statistically significant cumulative abnormal returns are documented both pre and post broker recommendation dates. We then use a unique dataset of aggregate returns, portfolio holdings and daily trades for our sample of small-cap equity funds to determine the extent to which broker recommendations impact on both the returns and the portfolio management decisions of fund managers. We find that a factor-mimicking portfolio based on broker recommendations results in a 7 percentage point reduction in the magnitude of monthly alphas estimated from a Carhart (1997) four-factor model (from 58 to 54 basis points a month). However, this broker recommendation factor is less important than the momentum factor (which causes a 12 basis point reduction in the monthly alpha) and the Fama-French (i.e., size and bookto-market) factors (which explain a 13 basis point reduction). Using transaction-level data, buy trades following recommendations earn statistically significant cumulative abnormal returns of 1.56 percent after 60 days. Overall, we find evidence suggesting that broker recommendations play an important role in the investment decisions of fund managers, and comprise an economically significant component of the monthly alphas generated by their investment strategies. JEL Classification: Keywords: Active Small-Cap Fund Management; Broker Recommendations; Alpha Corresponding Author: Terry Walter email

1. Introduction
The recent abnormal performance by Australian small-cap equity fund managers prompts the question of whether they are genuinely well-informed, and the extent to which informational asymmetries are exploited by institutional investors. Equity fund managers are known to conduct in-house research, are consumers of sell-side analyst research, and establish relationships with company management. Given this, they may possess an information advantage over other investors, which might have contributed to the significant alphas achieved in recent years. Alternatively, their alphas may be derived by trading on the basis of broker recommendations, which are also known to possess information content. This study explores whether small-cap equity fund outperformance is correlated with managers following the recommendations of brokers, as well as the extent to which broker recommendations influence the investment decisions of Australian small-cap equity funds. The introduction of compulsory superannuation in Australia since the early 1980s has resulted in a dramatic increase in funds under management. This is evident in the $A1.16 trillion in consolidated assets under management as at March 2009 (Reserve Bank of Australia) compared to $A732 billion only five years earlier. The growth in funds under management has resulted in fund managers seeking alternative asset classes, such as small-cap equities to enhance investor returns, based on both their exceptional performance and portfolio diversification benefits. Indeed, Bennett et al. (2003) have documented a shift in the preferences of institutional investors toward small-cap stocks. The appeal of small-caps is twofold. First, the ‘small-firm’ anomaly, first discovered by Banz (1981) and further confirmed by Brown et al. (1983) and Gaunt et al. (2000) in the Australian equities market, provides evidence that small-cap stocks outperform on a risk-adjusted basis. Despite a number of explanations put forward to explain the ‘small-firm’ anomaly, including taxloss selling, additional risk premia and transaction costs, no definitive conclusion has been reached. Second, several studies attribute the performance of small-cap stocks to lower levels of research coverage. This ‘neglected firm’ effect as documented by Arbel et al. (1983) and Irvine (2003) creates both information asymmetries and opportunities for informed fund managers to earn abnormal returns. Our primary motivation for this study is to further understand why Australian small-cap equity fund managers have been able to systematically outperform the market (on a risk-adjusted basis) in recent years. Chen et al. (2009) find evidence of small-cap managers’ stock selection ability with risk-adjusted abnormal returns of between 59.6 and 76.1 basis points per month. Furthermore, Gallagher and Looi (2006) find evidence that fund managers are better at exploiting potential mispricing for the most liquid of small-cap stocks (ranked 101st to 150th in market capitalisation) on the Australian Stock Exchange (ASX). Based on these studies, we pose the question of whether small-cap equity managers earn their alphas through following broker


recommendations, a known source of investment value, or whether they possess a unique information advantage. This study makes a number of contributions to the literature. First, we extend the work of Chan et al. (2006), through the use of robust methods to compute abnormal returns, with a focus on short time horizons for returns around broker recommendation dates. This is in light of findings by Kothari and Warner (2006) that misspecification is common in long-horizon event studies. Furthermore, we use a larger and more representative sample of broker recommendations to confirm the findings of Chan et al. (2006) that broker recommendations possess information content within the Australian market. Second, we extend Chan et al. (2006) to provide a link between broker recommendations and the value derived by small-cap equity fund managers. There exists an extensive literature on the investment value of broker recommendations, including Womack (1996) and Barber et al. (2001). However, no study to our knowledge tests whether information asymmetry is one of the drivers of small-cap equity fund outperformance. We test this by using broker recommendations as a proxy for information, and investigate the extent to which these recommendations affect the investment decisions of small-cap equity managers. Initially, performance is analysed through returns-based models using a representative dataset of 34 active institutional Australian small-cap equity funds. We extend the traditional factor models pioneered by Jensen (1968), Fama and French (1992) and Carhart (1997) which have been widely used in the fund performance literature. We incorporate broker recommendations as an additional factor, and confirm that part of the significant alphas shown in Chen et al. (2009) can be attributed to an investment strategy based on broker recommendations. In addition to returnsbased performance models, transaction-based measures are used to determine the extent to which small-cap equity managers trade on the basis of recommendation levels. This approach represents a significant improvement to holdings-inferred trades, used in previous studies such as Chen et al. (2000) for the U.S. and Pinnuck (2003) for Australia. A number of innovative approaches are used to analyse the transaction-based performance of small-cap equity managers, including their performance sensitivity around recommendation levels, research coverage levels and timing factors. In summary, this study provides a valuable link between broker recommendations and small-cap equity funds, two related areas which have not been jointly explored in the literature. In doing so, we enhance the understanding of the source of small-cap fund outperformance and answer the question whether broker recommendations contribute to the alphas earned, or whether small-cap equity managers are independently well-informed. The remainder of the study is structured as follows. Section 2 reviews the literature with respect to broker recommendations and small-cap fund performance. Section 3 provides a brief overview of the institutional details relating to the Australian fund management industry and the Australian stock market and outlines the data used in this study, while Section 4 provides an outline


of the research design, and this is followed by the empirical results as well as a number of robustness tests. Section 6 concludes.

2. Literature Review
2.1 Broker Recommendations
Brokerage houses expend considerable resources to collect, analyse and publish information on stocks to investors. Sell-side analysts are also known to develop strong relationships with the management of the companies they cover. In spite of regulation which prohibits the selective disclosure of material information by sell-side analysts1, research analysts are often perceived to have an information advantage over other investors given their perceived proximity and access to company management. This may be attributed to either their superior ability in analysing and processing public information, or by acquiring private information before other market participants. The significant stock price reaction upon the announcement of a recommendation is evidence of the information contained within a recommendation. The information provided by sell-side analysts is in the form of either an earnings forecast or recommendation on a stock. According to Elton et al. (1986), the earnings forecasts of analysts are subject to the interpretation of the investor, whereas recommendations provide investors with a clear course of action in whether to buy or sell a stock. As a result, this study is primarily focused on recommendations, rather than the earnings forecasts issued by brokers.

2.1.1 Initiations of Broker Coverage It is expected that a greater level of information asymmetry exists for stocks with lower levels of research coverage. This information asymmetry was first documented by Arbel et al. (1983) who discovered a ‘neglected firm effect’; where firms with minimal research coverage experience higher returns. These returns persist after controlling for stock size, which is important given that smaller firms tend to have lower analyst coverage. In addition, findings to date suggest that as the level of research coverage increases, the marginal contribution of the recommendation is reduced. For instance, Dhiensiri and Sayrak (2005) find that the value of a recommendation revision is inversely related to the number of analysts following a firm. Similarly, Kelly and Ljungqvist (2007) find that price reactions are attenuated with increased coverage, suggesting diminishing returns to coverage at the margin. This further supports the notion that stocks with lower research coverage possess a greater degree of information asymmetry, which potentially provides opportunities for small-cap equity managers to earn excess returns. An ‘Initiating’ recommendation on a stock represents the first coverage of that stock by a broker.
1 Such

Given a lack of prior information, Chan et al. (2006) posit that ‘Initiating’

regulations include Regulation Fair Disclosure (Regulation FD) in the United States and the Continuous Disclosure requirements in Australia. 3 


There is also evidence of this in the Australian market by Chan et al. However. Beneish (1991) and Stickel (1995) find significantly positive (negative) stock price reactions to buy (sell) recommendations.1. Womack (1996) uses the First Call database of recommendations and finds large. Although these authors acknowledge their result is partially attributed to an initiation with favourable information content. Furthermore. as the share price reactions of ‘Virgin’ recommendations are not significantly different from ‘Continuing’ or ‘Initiating’ recommendations. In the Chan et al. as well as a significant but short-lived postrecommendation drift in stock prices consistent with the forecast direction of the recommendation.5 percent can be earned by adopting a trading strategy that involves purchasing stocks with a buy recommendation over those with a sell recommendation. These excess returns are strongest in the month in which the recommendation is made. namely whether there was a recommendation by the same broker in the previous one. Broker recommendations are also found to have predictive power with respect to stock returns. ‘Virgin’ recommendations are a subset of ‘Initiating’ recommendations and are the first coverage of that stock by any broker. Likewise. 4  2   . initial share price reactions on recommendations.82 percent upon their initiation announcement.recommendations have greater information content and price response upon their announcement. Similarly. and ‘Continuing’ if otherwise. Elton et al. The stock price reaction upon announcement of a recommendation is used as evidence of the information contained. For instance. (2006). two or three years. (1986) test the forecast ability of analyst recommendations and find that an additional 4. recommendations are classified as ‘Initiating’ if a particular broker has not published a recommendation in the past one two or three years2. their finding differs to Irvine (2003).                                                          Chen et al. Irvine (2003) finds that the incremental price impact of an initiation of coverage is 1. Demiroglu and Ryngaert (2008) find that ‘neglected’ stocks with no prior research coverage in the past year experience abnormal returns of up to 4. but also persist for up to two months. which shows that returns on initiating buy (sell) recommendations are significantly greater (less) than zero over the six months following their release. An additional finding is that the share price reaction varies with firm size. it suggests the presence of information asymmetry for firms prior to their first initiation of coverage.02 percent greater than the price reaction for a recommendation on a stock that is already covered by analysts.2 The Value of Broker Recommendations The literature generally finds that investment value is contained within broker recommendations. (2006) study. 2. Consistent with this notion. as smaller capitalisation firms experience greater market reactions to recommendations compared to large capitalisation firms. (2006) expect ‘Virgin’ recommendations to have the greatest information content. Chan et al. (2006) use three different periods of time to determine if a recommendation is initiating.

as this may have a negative impact on trading commissions or jeopardise corporate relationships with the client (see McNichols and O’Brien (1997)). (2004) find that the quarterly change in consensus recommendation level is a robust predictor of returns. (2006). This is documented by McNichols and O’Brien (1997) who find that analysts tend to initiate coverage on stocks with favourable prospects. and find the stock price reaction is consistent with the direction of the recommendation. In addition to Chan et al. earn abnormal returns. Barber et al. 5    . there also exists self-selection bias in analyst coverage. (2001) examine an investment strategy on the basis of recommendations. Jegadeesh et al. Thus. 2. One such bias arises from the fact that brokers are usually reluctant to issue a sell recommendation on a stock. (1986) and Womack (1996).Australian evidence also indicates that brokers have stock-picking ability. Aitken et al. economic incentives may still exist for analysts to provide recommendations which are biased in some way. formed around recommendations. They additionally find that stocks in the bottom three deciles by market capitalisation exhibit the greatest returns to positive recommendations. and find that portfolio rebalancing in response to consensus recommendation changes yields annual abnormal gross returns in excess of 4 percent.1.1.4 Recommendation Bias Although ‘Chinese walls’ exist to prevent conflicts of interest in brokerage firms.3 Recommendation Revisions Findings to date suggest that is it the changes. and find that investing on recommendation levels alone is profitable only for ‘glamour’ stocks with high ‘value’ and positive momentum. the evidence revealed by the literature suggests that the recommendations provided by research analyst recommendations have information content which can be used to earn abnormal returns. the number of buy and hold recommendations typically greatly exceeds the number of sell recommendations issued. They also distinguish between stock characteristics. The more recent literature has focused on consensus recommendations. 2. which have a greater impact on returns. as confirmed by Elton et al. while dropping coverage on stocks with less favourable prospects. rather than the absolute level of analyst recommendations. (2000) examine buy and sell recommendations produced by 20 stockbroking firms on Australian stocks. It has also been shown that investment strategies. As well as a tendency to issue positive recommendations. in particular for small-cap stocks. For this reason.

such as holding growth stocks or stocks with high momentum. Using a sample of simulated funds. This may be due to the fact that returns-based measures do not contain fund-specific information other than returns. A drawback to this approach is that hypothetical returns on stock holdings overestimate actual fund returns. Grinblatt.2. However Carhart (1997) finds that this can be explained by the ‘momentum’ effect initially documented by Jegadeesh and Titman (1993). Daniel. The high explanatory power of these models in asset pricing tests provides further justification for their use. The traditional performance evaluation literature examines the actual realised returns of funds relative to particular benchmarks. Moreover. In addition. both authors stress the importance of survivorship bias which may have affected the conclusions made in previous studies.2 Holdings and Transactions-based approaches Performance evaluation techniques are highly dependent on the benchmarks used. Given that funds differentiate their services by systematically following certain ‘styles’. Based 6    . Titman and Wermers (1997) (hereafter DGTW) develop fund return benchmarks based on the size. Beginning with Jensen (1968) the literature generally finds that active managers do not outperform passive benchmarks.1 Returns-Based Approaches Numerous studies evaluate the performance of actively managed funds to determine whether funds earn superior risk-adjusted returns. (1993). performance can be evaluated after taking into account fund style and stock characteristics. Accordingly. in particular in response to the finding by Fama and French (1992) that additional factors such as size and book-to-market explain the cross-sectional variation in stock returns.2. Kothari and Warner (2001) find that the traditional performance evaluation techniques using multi-factor models have poor ability in detecting abnormal performance. 2. as demonstrated by Elton et al. multi-factor performance measurement models have also been applied to investigate fund performance. Similarly. DGTW argue that it is appropriate in determining whether managers have stock selection and timing ability.2.2 Performance Evaluation of Funds 2. which better captures the investment styles of managers. However. Malkiel (1995) confirms that mutual funds do not consistently outperform benchmarks. Carhart (1997) adds an additional ‘momentum’ factor to the Fama and French three-factor model and finds that performance is negatively related to portfolio turnover and expense ratios. Using these benchmarks. several studies document persistence in mutual fund performance. They advocate the use of an event-study approach to evaluate performance by directly analysing the returns on portfolio holdings and trades. given that expenses and transaction costs are not considered. Gruber (1996) evaluates the monthly performance of mutual funds using single index and four index models and concludes that the funds underperform a weighted average of the indices by 65 basis points a year. book-tomarket and momentum characteristics of the underlying stocks held. indicating little value in active investment management. Following this.

2. Thus these studies provide further evidence of the stock selection ability of fund managers and the overall value of active investment management.on the DGTW approach. the literature emphasises the importance of both constructing benchmarks correctly as well as selecting the appropriate measure of normal performance in determining abnormal performance. with a return difference of approximately 2 percent annually. Any inferences on fund outperformance must first involve specification of a benchmark that represents ‘normal’ performance. as well as their methods of construction. of which 60 basis points can be attributable to stock-style characteristics.3 Small-cap Funds The superior performance of small-cap funds has been documented in U.S. Similarly. This approach is used by Chen et al. and based on the CRSP 9-10 Index of U. However. 2. they find that fund managers possess stock selection 7    .3 Importance of Benchmark Specification Use of an appropriately specified benchmark is highly important in portfolio performance evaluation. For example. This has been highlighted by Lehmann and Modest (1987) who find that fund rankings differ substantially with alternative benchmarks. Overall. data and Pinnuck (2003) in an Australian context. Gorman (2003) examines the performance of 165 actively managed small-cap funds in the U. Similarly. as funds hold stocks that outperform the market by 130 basis points annually. (1993) and Grinblatt and Titman (1993) find that mutual fund performance is sensitive to the measurement of performance benchmark. Pinnuck (2003) examines trades and holdings and finds that fund managers are informed when it comes to stock selection. To better reflect the portfolio allocation decisions of fund managers. the use of inferred trades based on holdings data also has some limitations.S. gains 2. Performance evaluation based on stock holdings alone can be used to infer the trades of fund managers to evaluate performance. In Australia. studies such as Keim (1999). small-cap stocks. Similarly. taking into account transaction costs.. these results are consistent with Keim (1999) in that the small-cap funds earn a positive abnormal return of approximately 2 percent per annum. Using the inferred trades from holdings. Wermers (2000) finds evidence of fund manager stock selection ability. between 1982 and 1995. both large and small cap funds have been found to possess stock selection ability.S.2. Elton et al. (2000) using quarterly U. our study focuses on the daily transactions of small-cap funds. Gallagher and Looi (2006) argue that inferring trades based on monthly holdings does not capture intra-month trading for funds. who find that a passive mutual fund operating using patient trade re-balancing. Although they do not specifically examine small-cap funds. they find that actively bought stocks significantly outperform stocks which are sold.2 percent over the 9-10 Index annually.S. Gallagher and Looi (2006) focus on the daily trades of fund managers based on their performance and trading behaviour.

whereas the outperformance of Australian small-cap funds may suggest small-cap fund managers are informed.ability and are able to outperform passive benchmark portfolios. However. and possibly obtain information from company management. The ability of equity fund managers to conduct their own private research. However. In addition. it is worthwhile to test how public information. given the gap in the literature linking broker recommendations and equity funds. Industry surveys reveal that most funds employ their own in-house analysts to provide private research coverage of stocks. Australian studies also find evidence of the superior performance of small-cap equity funds. Chan et al. Similarly. studies have also shown that mutual funds do not consistently outperform the market. More importantly. An alternative view proposed by Irvine et al. (2007) is that sell-side analysts ‘tip’ their institutional clients prior to recommendation release. who find that the costs incurred by small-cap equity funds are substantial and amount to 1. a study by Brown (2008) examines the extent of recommendation motivated trades and finds that mutual funds ‘herd’ into (out of) stocks with consensus upgrades (downgrades). are used in the investment decisions of Australian small-cap funds. they find that fund managers are better at exploiting information for small-cap stocks ranked 101st to 150th in market capitalisation. portfolio holdings and daily transactions of active Australian small-cap equity fund managers and find they possess superior skill and stock selection ability. 2. Chen et al.09 percent for a round trip transaction on average. (2009) examine the aggregate returns. funds themselves have access to company management in the same way that sell-side research analysts do. even after considering transaction costs. (2006). in light of their recent outperformance. such as broker recommendations. (2005) finds that the extent of such herding is greater with increased information uncertainty. which is proxied by the dispersion in analyst forecasts.4 Link between Broker Recommendations and Equity Funds There is limited literature on the value of broker recommendations with respect to equity funds. However. Transaction costs also need to be taken into account when analysing the performance of small-cap funds. These two studies provide evidence that mutual funds rely heavily on the public information provided through broker recommendations. suggests that fund managers may be informed. 8    . This issue is further explored by Comerton-Forde et al. Their joint finding that (i) institutional trading increases significantly in the days prior to recommendation date and that (ii) these trades earn positive abnormal returns provides some support for the tipping hypothesis.

the I/B/E/S database converts these to a 5-point classification system.2 Broker Recommendation data Broker recommendations are sourced from the I/B/E/S database. as too has the mean number of brokers following a Small Ordinaries index stock. A total of 650 stocks were at some stage included in the small ordinaries index. The average number of unique brokers that issued a recommendation on these 650 stocks at some stage during the 15 year period was 7. resulting in a sample of 801 ‘small caps’. which increased from 2.76. with a total of 33. underperform and outperform. For consistency.3. Hold. Based on ASSIRT estimates. 762 unique stocks had recommendations in I/B/E/S.54 in 1994 to 5. overweight. It is evident that coverage has been increasing over time. the total funds under management of the small-cap funds industry at 30 April 2007 is approximately A$4. where the buy/sell ratio is 9. which commences in November 1993 for Australian stocks. Small-cap stocks such as those that are constituents of the Small Ordinaries index are generally associated with lower research analyst coverage levels than large-caps stocks. The direction of this bias to recommend stocks as buys is in line 9    . Table 2 provides the descriptive statistics for the full sample and sub-samples of broker recommendations from the I/B/E/S database. The categories are Strong Buy. These stocks comprise the S&P/ASX Small Ordinaries Index (hereafter Small Ordinaries).1 Small Ordinaries Stocks Small-cap funds in Australia generally target stocks that are constituents of the S&P/ASX300 Index (hereafter ASX 300) but lie outside of the S&P/ASX 100 Index (hereafter ASX 100).13 in 2008. Underperform and Sell. considerable asymmetry exists in the number of buy and sell recommendations. This asymmetry is even greater in Virgin recommendations. This forms the full recommendation sample. Most brokers use an expanded classification system with recommendations such as underweight.561 million. Buy.891 observations. <<INSERT TABLE 1 ABOUT HERE>> 3. Our full sample has a total of 56. The total number of brokers in the market has increased. which we use in this study.2.0 in the full sample. with Strong Buys and Buys combined outnumbering Sells and Underperforms by a factor of 3. We include stocks ranked between 101st and 300th in market capitalisation between November 1993 and December 2008 in our recommendation database. or approximately seven percent of the total Australian equity market capitalisation. in which a total of 64 unique brokers provide recommendations.687 unique recommendations made between November 1993 and December 2008. Table 1 shows broker coverage levels for all stocks in the Small Ordinaries index during the years 1993 to 2008. Data 3. Consistent with the literature. Of these.

and as a result. We use the Portfolio Analytics (PA) database which includes month-end portfolio holdings and daily transactions of a subset of the small-cap equity fund managers.7 percent of the entire small-cap fund universe (by funds under management). the funds had a total of A$0.2 Small-Cap Funds data The Mercer Manager Performance Analytics (Mercer) database includes the monthly returns of 40 active Australian small-cap equity funds.658 recommendations. Hence the PA database accounts for approximately 16. Furthermore. The PA database was constructed using an ‘invitation’ approach to the largest equity investment managers in Australia. The PA database includes month-end portfolio holdings of 13 active Australian small-cap equity funds. This database is free of survivorship bias. Table 2 also presents summary statistics on the market model subsample used to analyse returns around broker recommendations. where firms tend to initiate coverage on firms with favourable prospects as a mechanism to generate trading commissions. on a pre-expense basis. The holdings are from March 1995 to June 2004. <<INSERT TABLE 2 ABOUT HERE>> Observations with missing returns data were removed from the sample. These funds are typically benchmarked to the Small Ordinaries Accumulation index. the mean monthly fund size by net asset value (NAV) throughout the sample period is A$78. Table 2 also presents the summary statistics for the DGTW recommendations sub-sample. which are managed by 11 separate managers. Each manager was requested to provide on a confidential basis. 53. As a robustness check. The resulting subsample of 34 active small-cap funds has consecutive returns that range from 14 to 158 months. Following Chen et al. the recommendation sub-sample consists of 46. This represents a sample of 38. (2009). on the basis of funds under management. information on their largest pooled active Australian equity funds that were open to institutional investors.261 individual holdings of stocks by the small-cap fund managers. As at the end of the sample period in 2004.with economic incentives that exist for analysts to issue favourable recommendations. each fund in our sample is required to have a minimum of 12 consecutive monthly returns between January 1991 and March 2004 to allow model estimation. DGTW-adjusted returns are also used to calculate abnormal returns around the recommendation date.217 broker recommendations.5 million. Continuous trading over the past year is required for momentum characteristics. It is also consistent with the notion that self-selection bias exists. 10    . as both surviving and non-surviving funds are included between 1991 and 2004. Recommendations are required to have returns during the 180-day estimation period to compute market model parameters. resulting in a market model subsample consisting of 3.76 billion in funds under management.

Panel B presents an average breakdown of the stocks held by small-cap funds by index membership. 11    . On average.700 aggregated daily trades over a seven-year period spanning February 1997 to June 2004.3                                                          3 These results are available from the corresponding author on request.535 sell packages. buy packages outnumber sell packages.700 aggregated daily trades comprising the full sample. with 8. The remaining 10. This results in a ‘recommendation subsample’ of 2. To examine the behaviour and performance of small-cap fund managers around broker recommendations.5 stocks in their portfolios. The daily transaction data for the small-cap fund managers include the aggregate daily trades of 12 active Australian small-cap equity funds. <<INSERT TABLE 4 ABOUT HERE>> Table 4 provides summary statistics for the trade packages formed from the daily transactions of the small-cap fund managers in the PA database.060 trade packages are formed using the five-day definition. Within the full sample.525 buy packages and compared to 6. In forming the sub-sample. as momentum characteristics are required to calculate DGTW-adjusted daily abnormal returns around trade packages. Between 1995 and 2004. The sample size is considerably smaller than the full sample. 52 (51) percent of the buy (sell) packages are executed within a single day and approximately 76 percent of all trade packages (i.800 trade packages. hence we also calculated descriptive statistics for this set to ensure no selection bias has occurred. Of the 43.572 buys and 1. Furthermore. Rather than analysing each trade separately. small-cap funds invested approximately 72. 3. 1. the full sample is made up of 43. Of these.<<INSERT TABLE 3 ABOUT HERE>> Panel A of Table 3 provides a summary of holdings in terms of the number of stocks held per fund.2 percent in micro-cap stocks that fall outside of the ASX 300. which are managed by ten unique fund managers. both buy and sell trades) are executed within four days.3 percent is invested in other non-equity assets including cash. we require stocks to be continuously traded over the past year. though in general the value of the packages for this reduced set of recommendations is somewhat larger than those for the full sample. small-cap fund managers hold 45. we form a recommendation subsample of trade packages.. based on stocks with data from our recommendation file.8 percent of funds in the Small Ordinaries Index. The results are quantitatively similar to those in Table 4. On a transaction level. 15.e.7 percent in the ASX 100 and 13. we follow Chan and Lakonishok (1995) and group individual trades into trade packages.228 sells.

recommendations are examined in conjunction with smallcap fund manager data to examine the extent to which fund managers incorporate broker recommendations in their investment decisions. Research Design The research design of this study consists of two parts. For this. 4.3. Hypothesis 5 Research coverage levels are inversely related to the level of abnormal returns earned from smallcap equity manager trades. prior positions in stocks with upcoming recommendations and earn subsequent abnormal returns from these positions. 12    . rights issues or stock splits. These include daily opening and closing prices. stock-specific information such as stock prices is needed to calculate returns. and then explain their rationale. daily high and low prices and also daily trading volumes and values.3 Stock-specific data In addition to analyst recommendations and fund manager data.1 Research Hypotheses Hypothesis 1 Broker recommendations are short-term predictors of stock returns and possess information content. Initiating recommendations have greater information content and associated share price reaction upon announcement. hence the alpha earned by these managers can be (partly) attributed to the information contained in recommendations.e. This dilution factor is used in calculating daily stock returns. Hypothesis 4 Active Australian small-cap equity managers take ‘front running’. 4. through the Securities Industry Research Centre of Asia-Pacific (SIRCA). Hypothesis 2 Small-cap equity managers mimic broker recommendations in their investment decisions. In addition. a daily dilution factor is included for each stock to take into account changes in shares outstanding due to events such as dividends. abnormal returns are examined around recommendation dates to investigate whether broker recommendations have predictive power. Once this has been established. Accounting information such as book-to-market ratios for multi-factor models are sourced from the Aspect Financial database. Firstly. i. Hypothesis 3 Active Australian small-cap equity managers trade on the basis of recommendation levels and subsequently earn significant abnormal returns on these trades. ASX daily price data are sourced from the Stock Exchange Automated Trading System (SEATS). We start by stating our hypotheses (in summarised form)..

2 Value of Broker Recommendations In this section. we use 1993 as a ‘holdout’ year and classify Virgin or Initiating stocks if no prior recommendations were issued during that year.2. These are (i) Strong Buys or Buys. The remaining recommendations are classified as Continuing. After classifying recommendations by level and type. The I/B/E/S classification system is used to group recommendations by level. recommendations are grouped into three categories by level. An Initiating recommendation is the first recommendation made on a stock by a particular broker. 4. two approaches are used to compute abnormal returns . Firstly. Following this.2.t + ei . (2006) focus on six-month returns following recommendation. Abnormal returns (ARs) on a stock around recommendation date are computed as the difference between the actual return and the expected return as predicted by the market model.1 Classifying Recommendations by Level and Type Initially.t = a i + bi Rm . Virgin recommendations are a subset of Initiating recommendations. as well as a control-firm approach first developed by Daniel et al. As the I/B/E/S database commences in 1993 for Australian stocks.2 Abnormal Returns around Recommendation Date – Market Model An event study methodology with a short-horizon window is used to compute the abnormal returns around recommendation date. a market model is used to obtain the coefficient estimates in the 180 day ‘estimation period’ prior to each recommendation date: Ri . we outline our research design used to test whether broker recommendations have information content and investment value. we conduct an event study to investigate the share price reaction and the presence of statistically significant abnormal returns around the recommendation date. as per Chan et al. Thus. which finds that short-horizon methods for event studies are generally more reliable than their long-run counterparts. (ii) Holds. (1997). A Virgin recommendation is the first recommendation made on a stock by any broker. For robustness. and (iii) Underperform or Sell. This approach is chosen following a recent survey by Kothari and Warner (2006).4. as proposed in Hypothesis 1. (2006).t is the return on the Small Ordinaries index on the same day. Therefore while Chan et al. recommendations are classified according to their Type. where other brokers currently cover the stock. we focus on the abnormal returns in the two weeks around a stock’s recommendation date. 4.the traditional market model.t (1) where Rit is the actual return on the underlying stock of recommendation i on day t and Rm. The ARs around each recommendation date 13    .

t − a i − bi Rm .t = Ri . book-to-market and momentum.t is the abnormal return on the underlying stock of recommendation i at time t.t is the return on the Small Ordinaries index on day t.t − rt DGTW (i ) (3) where ARi. To compute abnormal returns around recommendation date.3 DGTW-adjusted Abnormal Returns Abnormal returns based on a market model approach do not control for additional characteristics which may influence returns. mean daily ARs are computed by averaging across recommendations in the sample and the ARs are then summed to form cumulative abnormal returns (CARs). (1996). we include a separation between the estimation (t = -200 to t = -20) and event windows (t = -10 to t = +10).t is the actual return on the underlying stock of recommendation i on day t and Rm. The Pinnuck (2003) approach modifies the Daniel et al. based on estimated coefficients from the ‘estimation’ period. this benchmarking approach can also be applied to ARi . (1997). (2007) based on the Pinnuck (2003) approach. a DGTW control-firm approach to benchmark stock returns is adopted as an additional robustness check. The daily DGTWadjusted alphas can be expressed algebraically as: Although the DGTW method was developed to measure mutual fund performance. t = -10.2.t = ri .t represents the abnormal return on the underlying stock of recommendation i on day t. For instance it is known that considerable momentum is associated with broker recommendations. …. ri. ARi .t . computing abnormal returns on individual stocks. Accordingly. which prevents event-related activity in the stock price from influencing the estimated market model parameters. 4. (2000) and Chan et al. These include well-known market anomalies such as size. +10 (2) where ARi. 14    . we use daily DGTW-adjusted alphas constructed by Fong et el. based on the findings of Hong et al. and are the coefficients Following the event study approach suggested by MacKinlay (1997).t is the actual return on underlying stock of recommendation i on day t and rtDGTW(i) is the return of a characteristic matched benchmark portfolio assigned to the underlying stock of recommendation i across the characteristics of size. (1997) approach for an Australian context and is constructed on a stock universe consistent with our benchmark specification. Based on the entire sample of recommendations. This approach is motivated by the characteristics of the underlying stocks held in fund portfolios examined in Daniel et al. book-to-market and momentum.are computed over a ten day ‘event period’. Ri.

3. which represents return after adjusting for systematic or market risk.t is the pre-expense monthly return of the fund i in month t over the one-month risk-free rate sourced from the Reserve Bank of Australia (RBA) and rm. As such. motivated by Jensen (1968).Following this. In this model.t = a i + biSO rm .3. while biHML and biSMB represent the loading coefficients on the HML and SMB factors respectively. A positive alpha term indicates that a particular fund manager has outperformed the market through stock selection ability. The Fama-French three-factor model is specified as follows. fund manager skill is represented by the alpha term.t + biHML HMLt + biSMB SMBt + ei . 4. However. Furthermore.t ) + ei .t = a i + biSO ( rm . This model can be specified as: ri .t (4) where ri. 4.1 Single factor models – Jensen’s Alpha The single factor model.t (5) where HML and SMB are factor-mimicking portfolios capturing the book-to-market and size factors respectively. Hypothesis 2 is tested by adding a broker recommendation factor to examine the extent to which broker recommendations impact fund returns. the mean abnormal return is computed based on daily abnormal returns averaged across all recommendations in the sample and cumulative abnormal returns (CARs) around recommendation date are calculated. provides a risk-adjusted measure of portfolio performance and is commonly used in performance evaluation literature. Following this. biSO represents the loading coefficient on the market factor and ai is the unconditional Jensen’s alpha for the model. additional book-to-market and size factors are incorporated into the single-factor regression in evaluating fund performance. A number of commonly used single and multi-factor models are employed. Fama and French (1992) find evidence that the cross-sectional variation in stock returns can also be explained by size and book-to-market factors. ri .3 Returns-Based Performance Measures Returns-based performance measures are used to determine whether small-cap equity managers have the ability to outperform passively selected benchmark portfolios. 4.2 Three Factor Model – Fama French (1992) Traditional asset pricing models such as the CAPM assume that expected returns are a positive linear function of excess market returns. 15    .t is the monthly return on the Small Ordinaries Accumulation Index during month t in excess of the one-month risk-free rate on 90-day bank bills from the RBA.

t + biHMLt HMLt + biSMB SMBt + biPR1YR PR1YRt + e1.The HML or ‘High minus Low’ factor is constructed by calculating book-to-market ratios.4 Five Factor Model – Broker Recommendation factor As discussed in Section 2. Firms are sorted on the basis of market value of equity (MVE) from the AGSM SPPR database.3. defined as a firm’s book value of equity from the Aspect Financial database. who finds that the one year momentum anomaly documented by Jegadeesh and Titman (1993) explains the persistence in mutual fund performance documented in prior literature. Following this. ranked on the basis of their prior six-month returns.3. divided by market value of equity sourced from the AGSM Share Price and Price Relative (SPPR) database. and a short position in stocks with newly issued Underperform or Sell recommendations in the month. The model used is similar to the Fama-French three-factor model. with an additional factor PR1YR to capture momentum. where the coefficients on the factor-mimicking portfolios indicate the proportion of mean return explained by the four factors. with firms in the top (bottom) 30 percent classified as being ‘High’ (‘Low’) book-to-market. The Carhart (1997) four-factor model is specified as follows.4 Similarly. To test this. lagged by two months. the HML factor is constructed as being the monthly difference in returns between ‘High’ and ‘Low’ book-to-market stocks. the ability of broker recommendations to predict stock returns has been widely documented. SMB or ‘Small minus Big’ is a factor to account for size differences in firms. Firms above (below) the median in MVE are classified as ‘Big’ (‘Small’) and the SMB factor is the monthly difference in returns between ‘Big’ and ‘Small’ stocks. PR1YR is a momentum factor which represents the difference in returns between the top and bottom third of All Ordinaries stocks. ri . 4. SMB and Momentum factors. the model can be thought of as a performance attribution model. mimicking portfolios are formed on a monthly basis.t (6) where the HML and SMB factors are calculated as per the Fama-French (1992) three-factor model. Firms are sorted by book-to-market ratio. which take a long position in stocks with newly issued Strong Buy or Buy recommendations during the month. 16    . Whether or not small-cap equity managers adopt a trading strategy which mimics broker recommendations is a question posed in Hypothesis 2.3 Four Factor Model – Carhart (1997) The four-factor model used in evaluating fund performance is motivated by Carhart (1997).                                                          4 We acknowledge the help of Philip Brown who supplied the HML.t = ai + biSO rm . 4. we add a fifth broker recommendation factor to the Carhart (1997) four-factor model. According to Carhart (1997). In constructing a broker recommendation factor BMS or ‘Buy minus Sell’.

It is also possible for a single broker to revise the recommendation level on a stock from a buy to a sell and vice versa.t is the return on stock i.t is the return on stock j. The monthly portfolio return on the BMS factor can be expressed as.t = ai + biSO rm . the chosen market portfolio should adequately reflect the investment styles of the fund managers in the sample. rj. the consensus recommendation for that month is used. Indeed.t (8) The HML. and biBMS is its corresponding loading coefficient. BMS represents a ‘Buy minus Sell’ broker recommendation factor. This is to ensure that only pure ‘small-cap’ stocks which fall under the investment mandate of small-cap managers are included. such as Kothari and Warner (2001)5. In particular. For stocks which had ‘conflicting’ recommendations. 4. in light of recent studies which suggest that inferences of abnormal performance are highly sensitive to choice of benchmarks and models.t ) (7) where BMSt is the monthly return on the broker recommendation factor in month t. ri . Within our sample of 4.t + biHMLt HMLt + biSMB SMBt + biPR1YR PR1YRt + biBMS BMS t + ei . which had an Underperform or Sell recommendation issued in month t and M is the total number of stocks with a newly issued Underperform or Sell recommendation in that month. SMB. it is possible for several brokers to issue conflicting recommendation levels on a stock. Lehmann and Modest (1987) also stress the importance of a consistent approach when constructing benchmark factors.5 Justification of Market Benchmarks Specified Specification of an appropriate benchmark and model is extremely important.3. small-cap fund managers are benchmarked against the Small Ordinaries index.770 buy or sell recommendation/months. BMS t = 1 N ∑ (ri . BMS is formed from a recommendation sample whose stocks were members of the Small Ordinaries Index in the month of recommendation. The five-factor model is thus specified as follows. and PR1YR factors are constructed as per the Carhart (1997) four-factor model. Further. Similarly. In any month.In addition. we identified 160 instances of ‘conflicting’ recommendation/months and applied the monthly consensus recommendation level to these cases. The broker recommendation BMS factor has been                                                          5 See also Lehmann and Modest (1987) and Grinblatt and Titman (1993) 17    . ri.t ) − i =1 N 1 M ∑ (−r j =1 M j . an analysis of their holdings shows that over 70 percent of holdings are invested in the Small Ordinaries Index. which had a Strong Buy or Buy recommendation issued in month t and N is the total number of stocks with a newly issued Strong Buy or Buy recommendation in that month. Typically.

4. (2000) argue that trades are more likely to signal private information over the passive decision of holding a stock. (2009). and to disguise the execution of their trades. three key market benchmark specifications are used. as pointed out by Gallagher and Looi (2006).e. Following Chan and Lakonishok (1995). an issue that arises is that the Carhart (1997) mimicking factors are constructed from stocks within the All Ordinaries Index.4 Broker Recommendations and Small-Cap Manager Trades Previous studies have examined fund manager trades by focusing on the inferred trades from quarterly or monthly portfolio holdings. Chen et al. For this reason.1 Trade Package Methodology Consistent with the approach of previous small-cap equity fund studies. the daily transactions of small-cap equity managers to test Hypothesis 3 that broker recommendations influence fund manager trades. Small Ordinaries and Micro-cap stocks (defined as stocks outside of the ASX 300 but within the All Ordinaries indices) with weights of 4/90. trades are grouped based on the trade package methodology of Chan and Lakonishok (1995). To incorporate the use of a wider benchmark and to ensure consistency in benchmark construction. prior positions in stocks with an upcoming recommendation. 73/90 and 13/90 respectively. the All Ordinaries Index is also used as a robustness test. Thus. Comerton-Forde et al. (2006) and Chen et al. This is motivated by the fact that fund managers tend to split up large trades to minimise transactions costs through price impact. The Balanced Index is constructed using the weighted returns on the ASX 100. as well as Hypothesis 4 that small-cap managers are informed and take front-running i.4. In summary. a five-day trade 18    . this approach fails to capture trading activity during the month. Furthermore. Based on both an analysis of the holdings and the way BMS was constructed.. The holdings data also shows approximately 13 percent of holdings by value are micro-caps and approximately 4 percent of holdings are large-cap stocks in the ASX 100. the Small Ordinaries Index is initially used as the market risk factor in returnsbased performance evaluation models.constructed using Small Ordinaries stocks. (1) Small Ordinaries Index (2) All Ordinaries Index (3) A Self-constructed Balanced Index 4. we use a refined level of data. However. such as Gallagher and Looi (2006). However. as this appropriately reflects the underlying stocks small-cap equity managers invest in. we create a customised ‘Balanced Index’ which accurately reflects the proportionate holdings of smallcap equity managers in the sample.

4.package is defined6. 6 19    . That is. we further split the full sample based on whether a manager had an existing position (‘existing position’) in the stock when the trade occurred. buy or sell) until no trades are executed for five consecutive days.4. then we expect to observe trading activity prior to the release of a recommendation. instances of managers taking a ‘front-running’ position prior to recommendation release are identified. Firstly. which includes a manager’s successive trades in a particular stock within the same direction (i. Using the holdings data from the PA database. This is motivated by the idea that managers are more likely to rely on the information content of broker recommendations if they do not have a prior position in the stock. Their finding is that different package-days do not significantly alter their results. The mean daily DGTW-adjusted ARs and the CARs are then calculated for the sample. or an Underperform or Sell (hereafter sell) recommendation in the same month is included in this database. we use the trade package methodology of Chan and Lakonishok (1995). A trade which had a Strong Buy or Buy (hereafter buy). 4. or whether the trade package represented a new position (‘new position’) in the stock for a particular fund manager.4.4. Hold recommendations are                                                          Chan and Lakonishok (1995) test alternative package-day definitions in addition to the 5-day rule from which their results are based. we expect significant abnormal returns to be earned on trades in which a recommendation direction is followed.e. First. we examine the abnormal returns earned on trades which have followed the recommendation direction. we expect fund managers to have a ‘front-running’ position in stocks where information is due to be released to investors through a recommendation.4 The Information Advantage of Small-Cap Fund Managers If recommendations possess information content and if small-cap fund managers are informed as proposed in Hypothesis 4. we propose in Hypothesis 3 that broker recommendations influence the trades of small-cap equity fund managers in two ways. a trade package is flagged if the last day of the package occurs in a 10-day period prior to either a Buy or Sell recommendation.2 Trades on the Basis of Recommendation Level Under the assumption that there is information content in broker recommendations. 4. we partition trade packages on the basis of recommendation level and trade type. managers should be more likely to execute a buy trade on stocks with a Buy recommendation or execute a sell trade on stocks with a Sell recommendation. small-cap equity managers trade on the basis of recommendation levels. Secondly. Again. Following this..3 Returns around Trade Packages Following this. First. We also expect that small-cap fund managers earn significant abnormal returns from these positions. To test the extent small-cap managers trade on the basis of recommendation levels. a database of trade packages which have taken place around recommendation date is constructed. That is. Accordingly. daily DGTW-adjusted abnormal returns are computed in the 60 days before and after the trade package date.

Relative Position – Prior and Post Trade Fund managers are likely to have existing positions in the stocks in which they trade. For instance. can be defined as. Hence it is more meaningful to examine the relative weight of the trade as a proportion of NAV rather than the absolute value of the trade package. Transaction weight Within the sample. The relative weight of the overall position. given the expectation that they do not contain the same level of information as a Buy or Sell recommendation. We define an overweight (relative to the index) metric as. Transactio nWeight ijt = TradePackageValueijt NAV jt (9) Where Trade Package Valueijt is the dollar value of the trade package i made by fund manager j in month t.omitted from this analysis. which has implications on the size of the trades executed. and NAVjt is the net asset value of fund manager j during month t. we are also interested in the relative weight of the overall position in a stock both prior to and following a trade. a number of trade-related metrics are constructed to analyse the relative magnitude of ‘front-running’ trade packages. given that holdings in the PA database are month-end values. RelativePosition( Prior )ijt = HoldingValueijt −1 NAVjt (10) RelativePosition( Post )ijt = TradePackageValueijt + HoldingValueijt −1 NAVjt (11) where Trade Package Valueijt is defined as in Equation 9 and Holding valueijt-1 is the value of fund manager j’s holding on the stock underlying trade package i during month t-1. the small-cap equity funds vary substantially in Net Asset Value (NAV). however other time periods are also used as a robustness check. depending on their market capitalisation. both prior to trade and post trade. 20    . This is to ensure that small-cap fund managers take substantial positions in stocks with an upcoming recommendation. it is expected that positions will be greater in smaller stocks. The transaction weight is defined as. Overweight Position Position weights in a stock may also vary across different stocks. We initially define a 10-day period as an indication of managers taking a front-running position. Hence we also examine the size of a trade package in the context of a stock’s weight in the index. rather than the weight of the trade package alone. After identifying instances of ‘front-running by small-cap equity managers. For this reason.

we initially define sell-side research coverage levels for all the underlying stocks of the trade packages of small-cap funds. 4. we analyse the performance of trade packages with a split around coverage levels. rather than the entire sample. Following this. consistent with the approach detailed above. as suggested by Kecskes and Womack (2007). the frequency of issued recommendations varies between different brokers. Given this. WeightitXSO is the weight of a stock within the Small Ordinaries Index in the corresponding month t. we compute mean daily DGTW-adjusted abnormal returns around each ‘front-running’ trade and average these across the N trade packages in which a front-running position was taken. as proposed in Hypothesis 5. pairwise comparisons are also used to analyse both buy and sell trade packages with a split around recommendation type. We also compute CARs for the 30. Motivated by the approach of Dhiensiri and Sayrak (2005). We use the trade packaging methodology and methods for computing mean daily and cumulative abnormal returns (CARs). an analyst will at least provide a yearly earnings forecast for the stocks they cover. However.1 Value of Brokers’ recommendations 5. If small-cap equity managers are genuinely informed. we define the ‘coverage level’ as the total number of brokers issuing a recommendation on a stock within the same year.5 Impact of Coverage Levels on Transaction-based Performance In this section. Based on the recommendations sample.1. Empirical Results 5. 5. After analysing a number of trade-related metrics.Overweight ijt = RealtivePosi tion ( Post ) ijt − Weight itXSO (12) where Relative Position(Post)ijt is as per Equation 11. again using the last day of a trade package as the reference date. All Continuing recommendations earn statistically significant post21    . then we would expect coverage levels to be inversely related with the abnormal returns earned from their trades. Given that the aim is to determine whether broker recommendation levels influence the trading behaviour and investment decisions of small-cap equity managers. we outline our approach in testing the information advantage of small-cap equity managers by examining research coverage levels in conjunction with transaction-based performance measures.1 Returns around Recommendation Date Table 5 presents the cumulative abnormal returns (CARs) around recommendation date based on a market model approach. (1983) and more recently by Irvine (2003). This is motivated by the ‘neglected firm effect’ first documented by Arbel et al.and 60-day periods. This is based on the notion that managers are better able to exploit information asymmetry and potential mispricing in stocks with less publicly available information.

Initiating hold and underperform/sell recommendations are more negative in the period following the recommendation than those of the Continuing group. hence they are not reported in detail. as suggested by Petersen (2009). 5.55%.0]) and also from day t = 0 to t = +10 (expressed as CAR[0.recommendation CARs that are consistent with the direction of their forecast. our findings based on the market model approach are consistent with those of a similar Australian study by Chan et al. Thus a greater price reaction is expected.59% respectively. This confirms that the result estimated using the market model approach is robust to differences in research design.10]). (2005). Interestingly. cumulative abnormal returns around recommendation date are also computed based on DGTW-adjusted daily alphas. regardless of recommendation level (assuming that a hold recommendation is a disappointment). while the hold and underperform groups have positive. These results are very similar to those for the market model. approach. Initiating and Virgin recommendations are expected to contain greater information content.97% and -1. we conduct pairwise comparisons on the CARs from day t = -10 to t = 0 (expressed as CAR[-10. given there is limited prior information on a stock. with CARs in the 10 days before the recommendation date of -0. recommendations are however entirely inconsistent with expectations. Our statistical tests are essentially identical under either                                                          7 These results are available from the corresponding author on request. consistent with Irvine et al. (2006). We group Strong Buy and Buy recommendations under a ‘Buy’ category and Underperform and Sell recommendations under a ‘Sell’ category. returns.1. Overall the results are similar to the market model approach. We also estimated the statistical significance of the results presented in Table 5 recognising that the standard errors might be biased due to clustering. Virgin buy Virgin recommendations have CARs after 10 days of -1. Consistent with this. albeit insignificant. as proposed in Hypothesis 1.7 The minor differences are that the negative CARs in the period after the Strong Buy/Buy Virgin recommendations are less significantly negative under the DGTW approach. there is some evidence of statistically significant negative returns prior to Initiating and Continuing sell recommendation’s release. <<INSERT TABLE 5 ABOUT HERE>> As a robustness check. especially for the larger Continuing recommendations sample.2 Pairwise Comparisons of CARs around Recommendation Date To further compare the differences between recommendation levels and also recommendation types. while some Initiating Strong Buy or Buy recommendations have significantly positive CARs (at the 5 percent level) in the period after the recommendation. 22    . This finding supports the notion of sell side analysts ‘tipping’ institutional clients. Overall.

58 percent a month.e.<<INSERT TABLE 6 ABOUT HERE>> Panel A of Table 6 reports the pairwise comparisons for CARs computed with a market model approach. indicating a pronounced level of stock selection ability amongst small-cap 23    .. the beta on the market factor indicates that small-cap fund manager returns are highly sensitive to the Small Ordinaries Index. indicating the presence of a small-firm return anomaly. although they are statistically more significant (i. Each model involves a regression of monthly pre-expense returns of a fund in excess of the monthly riskfree rate against one or more factors. but only significant for Initiating recommendations in three cases and Virgin recommendations in one case. buying (selling) stocks with positive (negative) past six-month returns which also contributes to alphas earned. while Panel B reports pairwise comparisons for CARs computed using a DGTWapproach. the majority of which are constituents of the Small Ordinaries.e. alpha drops by 11 basis points to 0. given that fund managers hold diversified portfolios of stocks. This is not surprising. In addition. Panel B suggests that small-cap equity managers also adopt momentum strategies.e. These results follow a consistent pattern to those presented in Table 5 and are also consistent with the findings of Chan et al.. Our results are consistent with Chen et al.. the Fama-French factors) cause small-cap equity manager returns to drop from 82 to 69 basis points per month.. momentum). all four Initiating Buy minus Sell CARs tests are significant). <<INSERT TABLE 7 ABOUT HERE>> The results from the Fama-French (1992) model in Panel C and the Carhart (1997) model in Panel B show that the coefficient on the SMB ‘small minus big’ size factor is significantly positive.e. which partly explains the alphas reported in Panel D. (2006). The difference between buy and sell recommendations is always significant in the market model approach for Continuing recommendations. 7. Table 7 reports the results of a number of returns-based performance evaluation models. although it remains economically (i. After controlling for the additional factor based on the Carhart (1997) four-factor model (i. The DGTW-approach results follow a similar pattern.0% per annum) and statistically significant. The magnitude of Jensen’s alpha provides an estimate of the level of fund manager skill with respect to stock selection ability after controlling for market risk.2 Returns-based Performance Measures Single and multi-factor models are used to evaluate small-cap fund manager performance. the SMB and the HML factors (i.8 percent annually. (2009). 5. or 9. Panel D indicates that small-cap equity funds outperform the Small Ordinaries Index by 82 basis points a month. Similarly. Taken together.

Our findings indicate that BMS continues to be statistically significant after accounting for its correlation with the market factor. even after controlling for market anomalies such as size. partitioned on the basis of recommendation levels.e. managers should be more likely to execute a buy trade on stocks with a Buy recommendation or execute a sell on stocks with a Sell recommendation. the stocks sold by small-cap fund managers after a broker sell recommendation experienced insignificant positive CARs in the order of 0. Based on Hypothesis 3. an examination of the CARs reveals that small-cap fund managers earn highly significant CARs on purchases where the underlying stock had a buy recommendation. The excess return of 58 basis points per month in the Carhart (1997) model is reduced to 54 basis points per month when the BMS factor is added.3 Transactions-based Performance Measures We now turn to an analysis of fund manager data at the trade level to ascertain the impact of broker recommendations on the trading behaviour and performance of small-cap equity managers. In summary. broker recommendations play a strong role in the investment decisions of small-cap equity managers and contribute to their alphas. In contrast. Overall.47 percent in the 60-day period following the last day of the sell trade package. Also.37%.. 5. Panel A presents results for a five-factor model where the BMS broker recommendation factor is introduced. 24  8   . i. That is. book-to-market and momentum factor in Panel A.equity fund managers. Consistent with Hypothesis 3. a buy trade package on a stock with a buy recommendation earns a cumulative abnormal return of 1. because the pre-recommendation CARs are a significantly positive 4.56 percent in the 60-day period following the last day of the buy trade package.                                                          We confirm that multicollinearity is not an issue in the model by conducting further analysis to control for the correlation between BMS and the market factor. For example. These results are available from the corresponding author on request. The addition of the BMS factor also causes an increase in the coefficient on the size.8781 to 0. Panel A of Table 8 presents the results on transaction-based performance measures. This can be explained by the correlation that exists between the BMS factor and the Small Ordinaries market factor8. the findings in Panel A confirm Hypothesis 2. book-to-market ratio and momentum.8132 with the introduction of the BMS factor. Hypothesis 3 predicts significant abnormal returns on trades in which a recommendation direction is followed. The pattern of the CARS for this set of securities suggests that broker buy recommendations are for stocks with positive momentum. the BMS factor explains 7 percent of the excess return in the Carhart model. The loading coefficient on the Small Ordinaries market factor is reduced from 0. it is expected that small-cap equity managers trade on the basis of recommendation levels.

we investigate Hypothesis 4 as to whether managers take ‘front-running’ positions in stocks in which information is due to be released. as expected. the CARs an insignificant 0.<<INSERT TABLE 8 ABOUT HERE>> We investigate whether front-running. over 96% (98%) of the sell trade packages following buy (sell) recommendations have an existing position in the stock. the overweight position for buy trades for which there is a buy recommendation (1. The aim is to test whether managers take substantial positions in stocks prior to recommendation. However. sell trades in stocks with lower coverage levels generated significant CARs posttrade. i. be overweight relative to the weights in the small-cap index) and the majority of trade packages already have an existing position in the stock. occurs by examining the daily trades of small-cap equity managers prior to recommendation date.32) percent.73). 85 (90%) percent of the buy trade packages following a buy (sell) recommendation have an existing position in the stock. taking a prior position in a stock. Given that the average number of stocks held by each manager is 46 (and therefore that most of the existing positions will. than the overweight position for a sell trade following a sell recommendation (0. while the mean transaction weight as a proportion of NAV is approximately 0. indicating that very few of the sell trade packages are short sales..99 25    . The trade-related metrics are similar across all groups with the vast majority of trades taking place in stocks for which the manager has an existing position. Specifically. Panel A reports the CARs around buy packs for the respective coverage level groups. by definition. in contrast. For example. In contrast. the CARs are a significantly negative (-2. We calculate a number of trade-related metrics to analyse the relative magnitude of the trade packages. Following the trade.64) percent after 60 days. managers sell smaller quantities of stock following a sell recommendation than they do following a buy recommendation. Consistent with Hypothesis 5. Buy trades following a buy recommendation are larger. and Panel B reports CARs for the sell packs. Table 9 shows trade packages partitioned by their type (i. than buy trades following a sell recommendation. perhaps because they have a smaller proportion of NAV invested in stocks that receive unfavourable sell recommendations.. as expected.4 Coverage Levels and Transactions-based Performance Table 10 presents the cumulative abnormal returns (CARs) earned around trade packages with a split based on coverage levels. buy or sell) as well as the recommendation level on the underlying stock. Further.e. Specifically.43 (0. For the 356 sell trades made in stocks with a zero coverage level. the magnitude of the trades alone as a proportion of NAV is substantial.03) is greater. <<INSERT TABLE 9 ABOUT HERE>> 5.e.

The Balanced Index comprises stocks which reflect the underlying holdings of small-cap equity managers. who find that recommendations are less informative as the number of analysts following a firm increase. Our findings on sell trades support the notion that small-cap equity managers exploit the lower levels of information available on stocks with low research coverage to earn statistically significant abnormal returns. In contrast and inconsistent with expectations.28%). sell transactions are more informative when there is low coverage. given the very strong propensity for a buy imbalance in broker recommendations. Overall. a number of robustness tests are conducted. Where there are more than 9 brokers covering a stock the post-trade CAR is significantly positive (4. <<INSERT TABLE 11 ABOUT HERE>> 5. <<INSERT TABLE 10 ABOUT HERE>>   5. This result is consistent with Dhiensiri and Sayrak (2005).5 Alternative Market Benchmark Specifications Table 11 presents the results of returns-based factor models where a Balanced Index is used as the market factor. Perhaps there needs to be a strong consensus among several brokers in relation to buys before market participants follow the recommendation.percent after 60 days on a sell trade where the stock had more than nine brokers following it. Panel A reports performance of buy packs through an analysis of the CARs around the last date of a buy trade package. These involve altering the research design to ensure that our results are robust to differences in methodology. while buy packs for which there is low analyst coverage have insignificant returns. 26    . Under this benchmark specification essentially the same results detailed in table 7 are encountered. the magnitude of the CAR following the trade seems to be positively related to broker coverage. In contrast. The various models have high explanatory power.6 Robustness Tests To strengthen the validity of the findings in this study. For buy trades. and the alphas are statistically and economically significant across all model specifications. as well as alternative econometric techniques. the CARs follow a decreasing trend as the number of brokers decrease.

Saar (2001) observes that funds generally do not short sell. For stocks with both buy and sell recommendations. BMS t = b0 + b1 rm . the Small Ordinaries Index is regressed on the BMS factor. the residuals from Equation 13. 27    . a long position is taken in newly issued buy recommendations each month. These results confirm the robustness of our findings.t + e t (13) After estimating this equation. For instance. after controlling for the correlation between BMS and the market factor. we re-construct the BMS factor using long-only portfolios that better reflect the actual portfolio allocation decisions of small-cap funds. This approach is chosen over a regression with calendar year dummies as it prevents over-specification of the                                                          9 10 These results are available from the corresponding author on request. the consensus recommendation is applied for that month and a long position is taken if its consensus recommendation is a buy. returns-based factor models are estimated. The regression results9 indicate that across all the two benchmark specifications. Furthermore. Long-Only Portfolios in Five-Factor Model The broker recommendation factor BMS is formed on the assumption that fund managers are equally able to take short and long positions in stocks. However. These results are available from the corresponding author on request. Based on this. Given that the bust occurred in 2001. broker recommendations continue to play a significant role in the investment decisions of small-cap managers. we divide the sample in half to capture the lead up ‘boom’ period prior to and excluding 2001. Hence. expressed as follows. we control for possible differences in the risk and levels of information asymmetry attributable to this event. the charters of many funds restrict the usage of short sales. We also repeat the process. To construct a long-only BMS factor. which represent the portion of BMS that is uncorrelated with the Small Ordinaries market factor (which we term BMS*) are used in a five-factor model in which BMS* is used in place of BMS in Equation 8. Firstly. substituting the All Ordinaries Index and a Balanced Index in place of the Small Ordinaries Index. as the significance of the BMS factor and the results as a whole remains unchanged. an analysis of the holdings data reveals that short positions are rare or non-existent. the coefficient on the BMS factor continues to be significant. As a robustness test. to avoid risking unlimited losses if the stock price goes up.Controlling for the Correlation between BMS and the Market Factor We control for the correlation between the BMS broker recommendation factor and the market risk factor in the following manner.10 Changing Risk due to the Technology Boom and Bust The sample period used in the returns-based performance regressions encompass both the technology boom and its subsequent bust in 2001. Following this. as well as its subsequent ‘bust’ from 2001 onwards.

Furthermore. alpha is not statistically different to zero in the bust period. We additionally test the robustness of                                                          11 These results are available from the corresponding author on request. (2009) find evidence of the stock selection ability of small-cap equity managers. given the statistically significant alpha of 53 basis points per month prior to 2001 from the five-factor model. a traditional market model approach and a DGTW-approach motivated by Daniel et al. (1997). the results suggest that small-cap managers invested in the technology boom. one source of which is broker recommendations. The results indicate that the magnitude and significance of alpha is highly sensitive to this event. Given that the outperformance of these fund managers has been confirmed. Our study contributes to the literature by jointly exploring the areas of broker recommendations and small-cap equity funds to determine how valuable the recommendations are in the fund management process. These findings are confirmed using two approaches of estimating abnormal returns. we observe that the BMS factor is statistically significant prior to the boom. but not after. although it continues to be statistically significant. 5. The findings on returns-based performance models are consistent with those of Chen et al. If small-cap managers are genuinely informed. with a tilt towards growth stocks prior to 2001. Conclusion Chen et al. They also profited from this. Initially.11 Furthermore. We also show that the price reaction around Initiating recommendations is not significantly different from Continuing recommendations. alphas may be derived from relying on (and having prior access to) other public information. their alphas are sourced from their acquisition of private information. we confirm the findings of Chan et al. small-cap managers were perhaps more cautious and did not rely as much on the information provided by brokers for their investments.models. (2006) in that broker recommendations possess investment value. either through their own private research or through information leakages prior to their release. the addition of a broker recommendation mimicking factor portfolio to the Carhart (1997) model reduces alpha by 7 percent from 58 to 54 basis points a month. this study is unique in that it seeks to understand the drivers behind such outperformance. we show small-cap managers earn economically and statistically significant alphas of 58 basis points a month. 28    . Further. such as the information asymmetry between managers and investors in the market. as they earn risk-adjusted abnormal returns of between 60 and 76 basis points per month. This is observed across all model specifications. as observed by a statistically significant negative coefficient on the HML factor. Using a Carhart (1997) model. This suggests that post-2001. Interestingly. The large magnitude of the alphas earned inevitably prompts the question of whether returns can be explained through the information advantage of small-cap funds. (2009) in that small-cap equity managers possess stock selection ability. Alternatively.

whereas results are reversed for buy trades. Finally. its magnitude varies from 33 to 66 basis points a month. Arbel.. This study has examined broker recommendations as one source of public information which influences the investment decisions of small-cap equity managers. Furthermore. A. In a related paper. For example. Boni and Womack (2004) take an industry perspective when analysing broker recommendation value and find that an analysts’ industry expertise provides incremental investment value. we find that the abnormal returns following small-cap equity manager sell trades are inversely related to the number of analysts following a stock. 57-63. managers have a greater likelihood of trading on the basis of recommendation levels if they do not have an existing position in a stock prior to the trade. (2000). Thus examining both broker recommendations in conjunction with earnings recommendations may provide additional insights on the informational advantage of References Aitken. it is known that brokers and fund managers alike may have an expertise in a particular industry. and hence the information inherent in a buy forecast is discounted unless many brokers agree. and Strebel. (1983). Our transaction-based performance measures involve the examination of DGTW-adjusted cumulative abnormal returns (CARs) around trades. it would be interesting to take industry expertise into account when examining the impact of broker recommendations on small-cap funds. PACAP/FMA conference paper. presumably due a lack of information and company relationships on these stocks. J.. M.alpha across alternative benchmark specifications of the market factor. This confirms findings in the literature that sell trades are motivated by information and allow managers to exploit mispricing when there is lower coverage levels.. Carvell. Lastly. Giraffes. Gallagher et al. Financial Analysts Journal. as brokers revise earnings forecasts and recommendation levels in anticipation or following new information. S. Wong. This highlights the sensitivity of performance to benchmark specification in factor models and further motivates the use of transaction-based performance measures. 29    . and Muthuswamy. Recommendations themselves usually occur around the time of an earnings announcement. The Impact of Brokers' Recommendations: Australian Evidence. Institutions And Neglected Firms. P. Although alpha remains significant in most alternative benchmark specifications. which could result in an information advantage over market participants. whereas buy trades are informative only when there is a strong buy consensus recommendation among several brokers. We find that small-cap equity managers earn statistically significant abnormal returns on trades that follow broker recommendations.. K. (2008) examine whether fund managers realise abnormal returns around earnings announcements. small-cap equity funds. This might reflect the empirical regularity that buy recommendations greatly outnumber sell recommendations. Melbourne. With this in mind. 39(3).

Working Paper. L. W. R.. Lehavy. and Lakonishok. J. Gallagher. N. Gallagher.. 343-368. T. The Cross-Section of Expected Stock Returns. J. Womack (2004). New Evidence on the Nature of Size-Related Anomalies in Stock Prices.. and Lakonishok. H. L.. Brown. K. 56(2). Chen. M. E. Journal of Financial Economics. (2006).. E. Stock Prices and the Dissemination of Analysts' Recommendation. Gruber. (2006). forthcoming. The Relationship Between Return and Market Value of Common Stocks.. (1981). Sias. C. 33-56. K. M. Greener Pastures and the Impact of Dynamic Institutional Preferences. (1995).. C. (2008). K. Boni. Kleidon. (2009). C. M. Chan. J. and Mian. Dhiensiri. G. Working Paper.. Chan. The Journal of Financial and Quantitative Analysis. (2008).. and Wermers. B. M.. and Marsh. (1992). Measuring Mutual Fund Performance with Characteristic-Based Benchmarks.. Working Paper. The Behavior of Stock Prices Around Institutional Trades. (1997). Hwang. (1991). M. P. Comerton-Forde. and Ho. K. T. Initiation of Brokers' Recommendations. C.. Barber. The Profitability of Momentum Strategies. 35(3). Working paper.. Gruber. J. The First Analyst Coverage of Neglected Stocks. 47(2). M. 41(3). (2005). and Trueman. R. and Grossman. (1997).. Titman. J. K.. J. 64(3). Discrete Expectational Data and Portfolio Performance. Y. The Journal of Finance. Working Paper. Australian Journal of management. K. The Journal of Finance. 1203-1238. H. Chen. The Journal of Finance. Jegadeesh. Chan. R.. 427-465 30    . Market Predictors and Stock Returns. M. D. Jegadeesh. The Journal of Business. D.-L. E. The Value Effects of Analysts Coverage Initiations. The Value of Active Mutual Fund Management: An Examination of the Stockholdings and Trades of Fund Managers. The Journal of Finance. (2001). R.. J.. C. 55(6). S. 80-90. Mutual Fund Herding. (1996). N.. L and k. 52. On Persistence in Mutual Fund Performance.. 16(4). M. C. and Sayrak. W. and Overreaction in Stock Prices. B. 3-18. Grinblatt. Working Paper. 16(3). Elton.... W. D. J. Industries and Price Momentum... A. and Wermers. Analyst Recommendations. Carhart. 57-82. and Hlavka.. Can Investors Profit from the Prophets? Security Analyst Recommendations and Stock Returns. S. Financial Analysts Journal. Beneish. Brown. and Walter. (1983). The Journal of Finance. S. and Wermers. 393. 9(1). N. (2005). Analysts. Elton. 213231. (2003). Brown. R. Demiroglu. and Ryngaert. 531-563. H.. Investment Manager Skill in Small-Cap Equities. McNichols. K. N. S. 50(4). T. Daniel. Fama. Das. 1147-1174. R. Review of Financial Studies. (2000).. 12(1). M. Wei.. A. S. and Starks. R. R.-Y. Nahhas. Review of Financial Studies.. 6(1). A. Tuck School of Business. Transaction Costs and Institutional Trading in Small-Cap Equity Funds. D. Bennett.. 10351058. C. Mutual Fund Herding and Dispersion of Analysts' Earnings Forecasts. R. D. L. M..Banz. 122.. Efficiency with costly information: a reinterpretation of evidence from managed portfolios. and Walter. Chan. R. French. (1993).. 52(1). Journal of Multinational Financial Management. The Journal of Finance. (1986). Comerton-Forde. Journal of Financial Economics. C. 699-713.

Bad News Travels Slowly: Size. Working Paper. The Journal of Finance. 47-68. N. The Journal of Finance. and the Profitability of Momentum Strategies. 35(1). C. N. Positive Semi-Definite. 761-781. 167-199. Accounting and Finance. 1985-2010. D. (2003). Event Studies in Economics and Finance. 20(3). B. 55(3). Returns from Investing in Equity Mutual Funds 1971 to 1991. C. Malkiel. A. 389-416. M. C.. B.. (2000). (1987). D. B. K. D. A. Benchmarking benchmarks: measuring characteristic selectivity using portfolio holdings data. Lim. J. P.. The Journal of Finance. G.. and Warner. (2006). 287-300. Self-Selection and Analyst Coverage. and Titman. and Lee. Gruber. and Pinnuck. (1995). P. M. W. and Titman. S. Journal of Financial Economics. 48(5). C.. 23(2). Review of Financial Studies. P... Another Puzzle: The Growth in Activity Managed Mutual Funds. J. D.. 46(1). (1968).. 33-50.. Gallagher. 1083-1124. D. Irvine. A. Mutual Fund Performance Evaluation: A Comparison of Benchmarks and Benchmark Comparisons. The Journal of Finance. (2006). Joonghyuk.. Adds and Drops of Analyst Coverage: Does the Stock Market Overreact? Working Paper. Analyst Coverage. Grinblatt. S. R.. (2004).. (2007). Jensen. and O'Brien. Kelly. 125-147. Gallagher. K. S. 50(2). and Womack. Trading Behaviour and the Performance of Daily Institutional Trades. The Value of Research. J.. and Stein.. Portfolio Rebalancing. K. P. Handbook of Empirical Corporate Finance. and Momentum of Small-cap Mutual Funds. The Journal of Finance. and Puckett. Krische. 233-265. and Ljungqvist. N. (1993). B. 59(3). Heteroskedasticity and Autocorrelation Consistent Covariance Matrix.. Jegadeesh. The Impact of Share Price on Seasonality and Size Anomalies in Australian Equity Returns. A. Tipping. K. Econometrics of Event Studies. R. A Simple. McNichols. Irvine. T. Ed. B. Gray. Journal of Economic Literature. 783-810. 51(2). 51(3). Looi. 741768. Gorman. 48(1).. (2007). Working Paper. 40(1). 66(1).. and West. 431-451. 31    . S. M. S. A. Lipson. J. A. 42(2). Review of Financial Economics (12).. The Incremental Impact of Analyst Initiation of Coverage. C.. (1996). D. 35. Institutional Trading Around Earnings Announcements. (2003). Kothari. Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency. Elsevier-North-Holland. Newey. (1997). Hong. M. (1997). and Looi. Performance Measurement without Benchmarks: An Examination of Mutual Fund Returns. 9. Kothari. M. H. M. 265-295. Econometrica. 56(5). Accounting and Finance. Lehmann. Journal of Corporate Finance. The Journal of Business. In Espen Eckbo. (1993). (2007).. K.. M. L. Journal of Accounting Research. T. Gaunt. Evaluating Mutual Fund Performance. L. and Modest. R. and Warner. (2008). 173-194. (2001). Kecskes. Analyzing the Analysts: When Do Recommendations Add Value? The Journal of Finance. M. The Journal of Finance. Conditional Performance. J.Fong. and McIvor. MacKinlay. (1999). 13-39. Accounting and Finance. D. P. 549-572. C. 55(1). C.. J. The Performance of Mutual Funds in the Period 1945-1964. 65-91. (2000). Keim. and Lee. (1987). D. The Journal of Finance. (2007). A.. Gallagher. 703-708. Jegadeesh. An Analysis of Mutual Fund Design: The Case of Investing in Small-Cap Stocks.

51(5). R. The Anatomy of the Performance of Buy and Sell Recommendations. Mutual Fund Performance: An Empirical Decomposition into Stock-Picking Talent. (2001). 51(1). 811-828. Price Impact Asymmetry of Block Trades: An Institutional Trading Explanation. (2003). An Examination of the Performance of the Trades and Stock Holdings of Fund Managers: Further Evidence. Financial Analysts Journal. Transactions Costs. 14(4). 1153-1181. S. 137-167. 25-39. Style. The Journal of Finance. and Expenses. L. Do Brokerage Analysts' Recommendations Have Investment Value? The Journal of Finance. 38(4). (1995). 16551695. Saar. K. Wermers. Stickel. 55(4). The Journal of Financial and Quantitative Analysis. (1996). M. Womack.Pinnuck. G. 32    . The Review of Financial Studies. (2000).

14 4.15 4.39 5.Table 1 .41 4. brokers 10 15 14 15 23 21 20 20 19 19 19 22 25 26 30 55 Source: I/B/E/S database 33    .22 4.62 4.62 4.11 4.64 3.Broker coverage on small-cap stocks from 1994 to 2008 Descriptive statistics .Broker Recommendation of Small-Cap Stocks Year 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 All Years No.45 4.71 4.54 3. brokers per firm 2.13 7. stocks covered 127 159 155 184 186 194 207 196 183 177 192 201 196 216 220 650 Mean no.44 4.76 Total no.98 5.

25 13.Table 2. An Initiating recommendation is the first coverage by a particular broker.664 375 118 50 2.02 22.44 7.664 19.63 1.921 754 19. Initiating or Continuing and also by both level and type.435 10.976 53.19 42.438 3.145 56.988 7.17 12.31 0.Summary statistics for Broker Recommendations Table 2 provides descriptive statistics for the full sample.50 11. for a stock that is already covered.131 19.303 Recommendations by Level and Type The total sum of recommendations by Type does not equal the sum of recommendations by Level given that Virgin recommendations are a subset of Initiating recommendations. Hold.438 2.22 36.50 0.800 22. The remaining recommendations are classified as continuing.468 46.29 Count 11.079 4.38 12.201 7.20 34.560 666 18.238 1.41 DGTW Subsample Count 9.11 13.33 34. A Virgin recommendation represents the first recommendation on a stock by any broker.72 7.November 1993 .41 4.99 42.25 0.December 2007 Full Sample Count Recommendations by Level Strong Buy Buy Hold Underperform Sell Total Recommendations by Type Virgin Initiating Continuing Virgin Strong Buy or Buy Virgin Hold Virgin Underperform or Sell Initiating Strong Buy or Buy Initiating Hold Initiating Underperform or Sell Continuing Strong Buy or Buy Continuing Hold Continuing Underperform or Sell a. a breakdown is provided by the type of recommendation.68 7.440 825 286 90 4.11 6.41 21.084 4.903 3.33 7. Underperform or Sell.217 543 5.745 3.466 20. Market Model Subsample % 21.55 83.21 % 20.93 0.59 21.090 1.45 0. 34    .710 47.24 34.16 7.18 7.09 5.50 % 20. Broker recommendation descriptive statistics . between November 1993 and December 2008.891 1.742 501 165 72 3.39 1.81 0.33 12.035 2.622 24.78 37. and also due to some Virgin recommendations being excluded during the holdout year.932 3.13 1.091 17.92 42.142 554 16.907 2. market model subsample and DGTW subsample of broker recommendations from the I/B/E/S database. 1993.84 738 6.671 7.82 37.961 12.82 0.32 7. I/B/E/S classifies recommendations by level as either being a Strong Buy.12 85.77 1.29 4. including Virgin.335 6.603 39.664 45.13 6.26 0. Furthermore.42 85.047 11. Buy.658 2.51 13.

15 Other assets 10. Panel A reports the average number of stocks held by each manager per month.453.949 Panel B: Holdings breakdown by Index Membership S&P/ASX 100 Small Ords (%) Micro-Caps (%) (%) Mean Percentage of Holdings 3.478 78. Panel B provides of a breakdown of the index membership of the underlying stocks held based on the ASX100 Index.74 72.31 35    .285 90.530.882 Median 47 44. Dev 23.816 Std.Table 3 – Summary Statistics for Month-end Holdings of Small-Cap Funds Table 3 provides a summary of the month-end portfolio holdings of small-cap fund managers between March 1995 and June 2004.199. Month-end of Small-cap Funds .467 Upper Quartile 63 97.762.March 1995 to June 2004 Panel A: Descriptive Statistics of Month-end Holdings Mean Number of stocks held Value of Holdings $ 45.433 Lower Quartile 31 16.8 13.374. the Small Ordinaries Index and micro-cap stocks which fall outside of the ASX 300.

480 390.700 36    .242 186. only a buy or sell in one package) until no trades are executed for a period of five consecutive days.844 92.099.354 2-4 days 1.171 Total buy packs 8.789 159.540 385.128 238.535 100 365.422 1.517 150.340 >8 days 722 11 895.848 602. The sample is also split into quartiles based on trade package value.549 35.525 100 363.440 5-8 days 1. with a breakdown for the number of days it took to execute the entire trade package.422 308.681 94.538 363.961 >8 days 883 10 1.206 14 497.349 548.940 488.866 107.198 241.446 640. Summary Statistics for Trade Packages Full Sample .005 Panel B: Sells 5-8 days 844 13 505.452 52 201.984 23 321.e.481 71.258.128 60.449 313.409 156.374 290.140.114 371.967 639.512 1.254 581.890 59.860 389.489 35.633 25 339.361 663.428 202. Panel A reports summary statistics for all buy trade packages.539 117.097 643.336 51 227.648 Total sell packs 6.Table 4 – Summary Statistics for Transactions data – Full Sample Trade packages are defined as a fund manager’s successive trades in a particular stock within the same direction (i.375 1.500 1 day Number of packages % of sells Mean pack value $ Std dev pack value Q1 Median Q3 3.056 452.370.028 563.431 2-5 days 1.589 167.088.685 658..596 1. Panel B reports summary statistics for sell trade packages with a similar breakdown.February 1997 to June 2004 Panel A: Buys 1 day Number of packages % of buys Mean pack value $ Std dev pack value Q1 Median Q3 4.

0043*** -0.0130* 0.0053*** 0.0032*** 0. (2) ARi .0004 0.0] CAR[-9.+7] CAR[0.0133*** 0.0058 -0.0062 CAR[-8.0018 0.0] 0.0020 CAR[0.+10] Panel C: Continuing recommendations Strong Buy/Buy 0.0022 0.+3] -0.0036 0.+5] CAR[0.0063*** -0.+2] CAR[0.0010 -0.0002 -0.0084*** -0.0024* -0.0061 -0.0156 -0.0] CAR[0. as well as by recommendation levels.0096** CAR[0.0013 -0.0003 0.0] 0.0] CAR[-10.0088*** -0.0033 0.0162 -0.0] CAR[-3.0104 -0.0004 -0.0] 0.0] 0.0057 -0.0038* CAR[-3.0031*** 0.0038*** -0.0] CAR[-6.0013 -0.0099 -0.0152*** -0.0162*** -0.0016 -0.0102** CAR[0.0011* -0.+3] CAR[0.0026 -0.+9] CAR[0.0039 0.0047 0.0022** -0.0033*** 0.0011 -0.0013* -0. -20 The alpha and beta coefficients from Equation 1 for each recommendation are then used to compute abnormal returns in the -10 to +10 days around recommendation date.0014 0.0] CAR[-1. Panel A: Virgin recommendations Panel B: Initiating recommendations Event CAR[-10.t + ei .0158*** -0.0047*** 0.0082** CAR[0.0009 0.0015 -0.0006 0.0014*** 0.0013*** -0.+4] CAR[0.0072*** -0.0075*** -0.+8] CAR[0.0027*** 0. +10 The mean daily abnormal returns are computed by averaging across recommendations and CAR[t. where t = -200.0164*** -0.0066 -0.0097** CAR[-10.0008 -0.0036*** 0.0140** 0.0017 -0.0035*** 0.0012 CAR[0.+9] -0.0067 -0.0077* CAR[-9. respectively.0040 CAR[0.+2] -0.0] CAR[-5.0006 0.0029*** -0. 37    .0039 CAR[0.0] CAR[-1.0008 -0.0108** 0.+8] CAR[0.0] CAR[-9.0103*** -0.0007 0.0106 -0.Returns around Recommendation Date – Market Model Cumulative Abnormal Returns (CARs) Table 5 presents the CARs around recommendation date.0070 0.0025*** -0.0026 CAR[-6.0041*** -0.0] CAR[-7.0010 0.0046 -0.0000 -0.+6] CAR[0.0] -0.0027** 0.0159*** -0.0022 -0.0033*** 0. Initiating and Continuing recommendations.+10] CAR[0.0062*** -0.0] CAR[-5.0001 -0.0036 -0.0012 0.0022 0. ….0002 0.0003 0.0] CAR[-8.0039* -0.0018** 0.0037*** -0.0104*** Strong Underperform/ Strong Underperform/ Event Hold Event Hold Buy/Buy Sell Buy/Buy Sell 0.+1] CAR[0.T] = 0 and H1: CAR[t.0034 0.0103** CAR[0.0014 0.0108 -0.T] is the sum of mean daily abnormal returns between days t and T. The returns for the 180 day period from t =-200 to t = -20 prior to each recommendation date are regressed against the return on the Small Ordinaries Index.0101 -0.+8] -0.0036*** Hold -0.0] 0.0067*** -0.0] CAR[-8.0] CAR[-3.+1] 0.t − a i − bi Rm .+2] CAR[0.0095*** -0.0051 -0.0062*** -0.+5] -0.0041*** -0. where t = -10. All tstatistics and significance levels are calculated using Newey West (1987) standard errors which adjust for heteroskedasticity and autocorrelation in the residuals.0013 0.0035 -0.0029 -0.0059*** -0.0050*** -0.+5] CAR[0.0040* CAR[-4.0207 -0.0003 -0.+6] CAR[0.+3] CAR[0.0013 -0.0023* -0.0098 -0.+6] -0.t = ai + bi Rm .t .0046 -0.+4] -0.0131*** CAR[0.0051*** 0.0001 -0.0020 0.0047 0.0016 0. 5% and 10% (two-tail) level.0025 CAR[-5.0036*** 0.0036 0.0072 -0.0106 -0.0004 -0.0013 0.0143* 0.0] CAR[-4.0036* -0.+7] -0.0032*** -0.0097 -0.0011** -0.0063*** -0.0034*** -0.0087*** -0.0169*** -0.0015 0.+10] ***.0] 0. H0: CAR[t.+9] CAR[0.0035*** 0.0022** 0.+1] CAR[0.0046*** 0.0] CAR[-2.0030** 0.0036*** 0.T] ≠ 0.0] CAR[-6.0] 0.0022 0.0060*** -0.0015 0.0007 CAR[0.0094*** -0.0010 -0.0] CAR[-7. (1) Ri .0000 -0.0118*** -0.0026 CAR[-2.0] CAR[-2.0043*** 0.t .t = Ri .0019** -0.0023 -0.0072 -0. with a breakdown for Virgin.0089 -0.0041*** -0.0059 0.0006 0.0041*** -0.0022* CAR[-1.0023*** 0.0] 0.Table 5 . ** and * indicates significance at the 1%.0] CAR[-4.0025 0.0049*** Underperform/ Sell -0.0118 0.+7] CAR[0.0155* 0.0007 -0.0159*** -0.0013 -0.0047 CAR[-7.+4] CAR[0.0] 0.001 -0. ….0013 0.

0138*** 0. against the alternative H1.0202* 0.0178*** 0. We test the null hypothesis that H0.returns of type B recommendation = 0.0051*** 0.0072*** Recommendation Event window Virgin Mean Median Initiating Mean Median Continuing Mean Median Panel B: CARs estimated using a DGTW-Approach Recommendation Buy vs.0063*** Median ***.0069** 0.0202 0.0135*** Initiating Mean 0.0014 0.0132*** 0. the returns of type A recommendations .0252 0. Pairwise Comparison of Cumulative Abnormal Returns (CARs) Panel A: CARs estimated using a Market Model Approach Buy vs.0055** 0.0053 -0.0142*** 0.0178*** 0.0014 0.0008* 0.0141*** 0.0055 0.0125** 0. respectively.0154 -0.0154 0.   38  .0060*** 0.Table 6 – Pairwise Comparisons of CARs around Recommendation Date Table 6 presents pairwise comparisons of CARs around recommendation date. Sell Buy vs.0054 0. Means and medians are computed based on daily abnormal returns and the CARs are reported for t = -10 to t = 0 and also t = 0 to t = +10 event windows.10] -0.0097* 0.10] -0. Sell Event window CAR[-10.0082 0.0154 -0.0091* 0.0002 0.0084* 0.0036 -0.0050 0.0116*** 0.10] CAR[-10.0110 -0. 5%.0171*** Continuing Mean 0.0140* 0.0225 Virgin Mean -0.returns of type B recommendation ≠ 0.0] CAR[0.0423 0.0052 0.0264* -0.0075*** 0.0382** -0. and 10%.0074 0.0065*** 0.0035 0.0191 -0.0130*** Median 0.0146*** 0.0] CAR[0.0068** 0.0] CAR[0.0187*** 0.0006* 0.0104*** 0.0016 -0. Sell CAR[-10.0124 Median 0.0] CAR[0.0] CAR[0.0049** 0.0058*** 0.10] -0.0063 0.0505 0.0043 0. Sell CAR[-10. Statistical significance on means is calculated using a paired t-test with Newey-West (1987) standard errors.10] CAR[-10.0040*** 0.0047*** Buy vs.0048 0.0017 0. **. that the returns of type A recommendations .0253* 0.0055** 0. Hold Hold vs.0071** 0.0085*** 0.0230*** 0.0157 -0.0101 0.0] CAR[0. * indicates two–sided statistical significance at 1%. Hold Hold vs.10] CAR[-10.0086*** 0. Statistical significance on medians is based on a non-parametric Wilcoxon Signed Rank test.

Returns-based performance evaluation .2233 0.7223 50.1671 33. the model is specified as.1317 0.0737 0.1133 0. where b’s represent the loading coefficient on their corresponding factor.0501 0.6429 0.t where ri.8183 0.7648 0.0791 0.1285 0. lagged two months.3617 0.1355 0.0172 1.9517 114 0.t where PR1YR is a momentum factor. Panel A reports results derived from our five-factor Broker Recommendation model.0050 0.1146 0. Panel A: Five-Factor Model α βSO βHML βSMB βMOM βBMS Adjusted R-squared Number of Months α βSO βHML βSMB βMOM Adjusted R-squared Number of Months 0.t + biHMLt HMLt + biSMB SMBt + biPR1YR PR1YRt + e1.0516 -0.9652 0.t = a i + biSO rm .123 -0.Multi-Factor Models .1612 33.Table 7 .t where SMB is a size factor. Neg Sig.3555 13 0.7179 38 -0.0588 0.3909 0. ri .0051 0.9518 114 0.4934 -0. specified as follows.0977 0.1403 0. formed by the difference in monthly returns between the top and bottom 30 percent of firms. Note: Alpha estimates are reported in decimal form.0526 0. HML is a book-to-market factor. (5) ri .071 -0..t + biHML HMLt + biSMB SMBt + ei . ranked on the basis of book-to-market-ratio.1003 0.0556 0.8781 0. This is defined as book value of equity (BVE) sourced from Aspect divided by MVE.3425 0. Sig.0064 0. All t-statistics and significance levels are calculated using Newey West (1987) standard errors which adjust for heteroskedasticity and autocorrelation in the residuals of panel data. T-tests: H0 coefficient = 0.t = ai + biSO rm.8572 -0.1108 0.t + biHMLt HMLt + biSMB SMBt + biPR1YR PR1YRt + biBMS BMS t + ei .6429 0.4044 13 -0.t = ai + biSO rm .0705 0. specified as follows. Panel C reports the results derived from a three-factor Fama-French (1992) model.3849 0.0135 0.8132 0. based on the difference in returns on a mimicking portfolio which purchases (sells) stocks with a Strong Buy or Buy (Underperform or Sell) recommendation in any month.7185 50. specified as follows.005 0.0058 0.t (8) where BMS is a broker recommendation factor.0928 0.13 -0.0072 0. Pos.t are the monthly pre-expense returns of fund i and the monthly excess return of the Small Ordinaries Index over the monthly 90-day bank bill rate from the RBA and is the unconditional ‘Jensen’s alpha’. The monthly return on the BMS factor is the total return averaged across the number of stocks in the portfolio for that month.t + e i .2375 0.0054 0. (4) ri . formed by the difference in monthly returns between ‘small’ and ‘big’ firms.0052 0. formed by the monthly difference in returns on the top and bottom third of stocks sorted on the basis of their prior six-month returns.6962 0.0609 -0.5129 -0.3617 0. Panel B reports results derived from a four-factor Carhart (1997) model.March 1995 to March 2004 Mean Median Min Max Std.3875 0.9652 5 13 2 5 3 3 0 0 0 1 0 0 Panel B: Four-Factor Model (Carhart 1997) 6 14 2 4 3 0 0 0 1 0   39  .t and rm. The results for a single-factor model are presented in Panel D.7286 0. (6) ri .2704 0.t = ai + biSO rm.6848 0.1365 0.7178 38 0.0189 1.2147 0.Returns-based performance measures – Multi-Factor Models Table 7 presents the results derived from returns-based models. Dev. ranked on the basis of market value of equity (MVE) from the AGSM Share Price and Price Relative (SPPR) database.2133 0.

0209 0.0260 0.9529 0.9371 0.0077 -0.6429 Panel C: Three-Factor Model (Fama French 1992) 0.1982 -0.0301 0.6986 0.1033 0.0595 0.2360 0.9652 6 14 3 3 0 0 0 1 Panel D: Single Factor Model (Jensen's Alpha) α 0.4315 0.1547 Number of Months 50.7098 50.3527 0.5516 1.8450 0.1271 0.6979 0.8478 0.3616 0.0075 -0. based on T-tests 0 0   40  .0044 0.0063 8 βSO 0.0072 -0.2186 0.2081 14 Adjusted R-squared 0.9652 ***.8669 0.0082 0. ** and * indicates significance at the 1%.0069 0. respectively.1525 38 13 114 33.0058 -0.0069 0. 5% and 10% (two-tail) level.6429 38 13 114 33.4037 0.7068 0.6764 0.8465 0.5105 1.Table 7 continued  α βSO βHML βSMB Adjusted R-squared Number of Months 0.0808 0.0025 0.

0355*** 0.50] 0.0031) (0.0] 0.0079) (0.0029) (0.0086) (0.0020 0.0026 (0.0092) CAR[0.0060 (0.0097 (0.0033) CAR[0.0156*** 0. All t-statistics and significance levels are calculated using Newey West (1987) standard errors which adjust for heteroskedasticity and autocorrelation in the residuals of panel data.0058) CAR[0.0097*** 0.0050 (0.60] 0.Table 8 – Result of Transaction-based Measures .0033 (0.0047 (0.0094) (0.0022) (0.0025 0.0024) (0.0051) (0.0089 0.0032) (0.0029 0.0055** -0.0389*** 0.0056) (0.0050) (0.0046) (0.0040) (0.0008 0.0032 0.0068) CAR[-10.0082) CAR[0.0042) (0.0096) (0.40] 0.0045*** 0.30] 0.0] 0.0034) (0.0030 (0.0035 0. ** and * indicates significance at the 1%.0020) (0.0096) ***.0] 0.0088) CAR[-30.0053) (0.0115 0. 5% and 10% (two-tail) level.0080** -0.0091** 0.0082 0.0261*** -0.0031) CAR[0.February 1997 to June 2004 CARs around Transactions by Recommendation Level and Trade Type Buy Pack Sell Pack Event day Buy Sell Buy Sell N 1248 324 1003 225 CAR[-60.0033) (0.0072) CAR[0.0059** 0.0072*** -0.0052) (0.0071) (0.0085 (0.0069) (0.0437*** 0.0031 0.0] 0.0058 (0.0024 0.0108) CAR[-40.0058) (0.0021 (0.0063 (0.CARs around Transactions Table 8 presents cumulative abnormal returns calculated using the last day of the trade package as the reference date.0] 0.0] 0.0046) (0.0069) (0.0042) (0.0196*** 0. respectively.0027 0.0121*** 0.0126** 0.0351*** 0.0017) (0.0261*** 0.004) (0.0044) (0.0074) (0.0061) (0.0012 0. Abnormal returns are computed based on DGTW-adjusted daily alphas and the mean daily abnormal return and cumulative abnormal returns are then computed.0044) (0.0036) (0. based on T-tests   41  .0152*** -0.10] 0.0193** 0.0082) CAR[-20.0106) CAR[-50. Transaction-Based Performance Measures .0054) (0.0296*** 0.0013 (0.0051 -0.0076) (0.0025 0.20] 0.0023) (0.0276*** 0.

624 0.February 1997 to June 2004 Panel A: All trade packages in recommendation subsample Buy Transactions Sell Transactions Recommendation Level Strong Buy/Buy Underperform/Sell Strong Buy/Buy Underperform/Sell N 1.774 0.16 1. The mean transaction weight is total value of the trade package divided by the fund’s Net Asset Value (NAV).42 313. Overweightijt = RealtivePosition( Post ) ijt − WeightitXSO (12) Transaction-Based Performance Measures .03 1.33 344.Prior to Trade (% NAV) Relative position .95 0.87 187 98.32 1. Transactio nWeight ijt = TradePacka geValue ijt NAV jt (9) where N is the number of trade packages. Relative Position (Post) trade is defined as the sum of a fund’s existing holding in a stock and the trade package value as a proportion of NAV. RelativePosition( Post ) ijt = TradePackageValueijt + HoldingValueijt −1 NAVjt (11) Overweight relative to index is weight of the stock underlying trade package i within the Small Ordinaries Index subtracted from Relative Position (Post) computed as per Equation 11.623 0.28 2. This is expressed as.43 1.453 0.37 325. This can be expressed as.87 823 96.98 1.75 0.73   42  .Post Trade (%NAV) Overweight Relative to Index – post-trade position 948 85.55 1. RelativePosition( Prior ) ijt = HoldingValueijt −1 NAVjt (10) Similarly. NAVjt is fund j’s NAV in time t.Table 9 – Results from Transaction-Based Performance Measures Table 9 presents descriptive statistics of the trade-related metrics used in analysing transaction-based performance.03 272 90.82 0.07 249.111 302 854 190 N with existing position in stock % transactions with existing position Mean trade package size $ Mean Transaction weight (% NAV) Relative position . L is the total number of managers. averaged across number of trade packages.34 2. Relative Position (Prior) to trade is defined as a fund’s holding in a stock in the prior month as a proportion of NAV in the current month. Trade Package Valueitj is the dollar value of the trade package i made by fund j at time t.63 1. This is expressed as.

0013 0.0113 0.0049 -0.0] CAR[0.0042 0.0047 0.0020 0.10] CAR[0.0141*** Sell Packs 4 to 6 1686 0.0071 0. of brokers) N CAR[-60.0040 0.0026 0.0090* -0.0064* 0. Similarly.0215*** 0.0062 Buy Packs 4 to 6 2107 0. with a split around broker coverage levels.0044 1 to 3 1596 0.0074 0.0] CAR[-30.30] 0 356 0.0009 -0. denoted as CAR[t.0013 0.0007 0.0049 -0.0034 -0.10] CAR[0.0217 0. respectively.60] -0.0253*** 0.0005 0.February 1997 to June 2004 Panel A .0098 0.0018 -0. Abnormal returns are computed based on DGTWadjusted daily alphas.0000 0.0197*** 0.0042 -0.0117** 0. ** and * indicates significance at the 1%.0081 0.0428** Panel B .0089*** -0.60] Trade type Coverage level (no. The mean daily abnormal return and cumulative abnormal returns between day t and T.0044 7 to 9 1106 -0.0032 0.0] CAR[-30.0024 0.0274*** 0. are computed by summing the mean daily abnormal returns from days t to T.0269*** 0.0019 -0.0024 -0.0064*** 0. defined as number of unique brokers who have issued a recommendation within the same year.0020 >9 106 0.0015 -0.0466** 0. based on T-tests with NeweyWest standard errors   43  . All t-statistics and significance levels are calculated using Newey West (1987) standard errors which adjust for heteroskedasticity and autocorrelation in the residuals of panel data.Transaction-based CARs by Coverage Level for Buy Packs Trade type Coverage level (no.0119 0. N refers to the total number of brokers following the underlying stock of a particular trade package.0264* -0.0088*** 0. of brokers) N CAR[-60.0171*** 0.0163 -0.0038 1 to 3 1139 0.T].Table 10 – Coverage Levels and Transaction-based Performance Panel A of Table 10 presents the CARs from the last day of a buy trade package.0092 0 559 0.0] CAR[0.Transaction-based CARs by Coverage Level for Sell Packs CAR[0.0] CAR[-10.0486*** 0.0023 -0.0099 ***.0030* 0.0126* >9 101 0.0] CAR[-10.0022 7 to 9 1371 0. Transaction-based performance by Coverage Level .0204 0. 5% and 10% (two-tail) level.30] CAR[0.0307* 0.0024 0. Panel B presents the results for sell packs.

5873 -0.0182 0. Dev.1441 33. Small Ordinaries and Microcap Indices.9506 114 0.9101 0.3928 0.1399 -0. three and single-factor models is specified as per the five-factor model.7439 50.0075 0.4725 0.9639 0.0698 -0.6733 0.1854 -0.6019 -0.7944 0..t = a i + biSO rm .8360 0.March 1995 to March 2004 Mean Median Min Max Std.1356 0.3670 0.0896 0.9474 0. Further. The Balanced Index is constructed using the weighted returns on the ASX 100.2192 0.2556 0.0073 0.4278 13 -0.1358 0.1487 33.3482 0.1047 0.1199 0.6381 0. All t-statistics and significance levels are calculated using Newey West (1987) standard errors which adjust for heteroskedasticity and autocorrelation in the residuals of panel data. the Micro-cap index is defined as stocks outside of the ASX 300 but included in the All Ordinaries index.t where rm. Neg Sig.7417 50.2391 0. 70/90 and 10/90. the market factor is based on the monthly excess return of a ‘Balanced Index’ over the monthly 90-day bank bill rate from the RBA.0450 -0.0073 0.1023 0.0606 -0.6429 0.0115 0.0743 0.9652 8 13 3 5 5 2 0 0 0 1 0 0 Panel B: Four-Factor Model (Carhart 1997) 6 14 3 4 5 0 0 0 1 0   44  .3203 0.0032 0.9510 114 0. Across each model.7512 38 -0.Multi-Factor Models .0720 0.0073 0.t is the monthly return of the Balanced Index over the one-month risk free rate proxied by the monthly 90-day bank bill rate from the RBA and is the loading coefficient for the new market factor. Sig. weighted by proportions of 10/90.t + biHMLt HMLt + biSMB SMBt + biPR1YR PR1YRt + biBMS BMS t + ei .0062 0.7164 0.0075 0.0209 1. Similarly.3927 0.0480 0.0678 0. Panel A: Five-Factor Model α βBALANCED βHML βSMB βMOM βBMS Adjusted R-squared Number of Months α βBALANCED βHML βSMB βMOM Adjusted R-squared Number of Months 0.7480 38 0.9652 0.1319 0.0718 0.3952 13 0.1301 0.0494 0.0547 0. Returns-based performance evaluation .Table 11 – Returns-Based Performance Measures– Balanced Index Market Factor Table 11 presents the results derived from returns-based performance evaluation using single and multifactor models.2213 0. The new five-factor model of Panel A is now specified as follows.0192 1.0752 0. the β’s represent the loading coefficient on their corresponding factor. Pos.6429 0. In each model.1133 0. (8) ri . the market factor in the four.5621 0.0094 0.

9652 8 14 3 3 0 0 0 2 Panel D: Single Factor Model (Jensen's Alpha) α 0.0007 0.0228 0. ** and * indicates significance at the 1%.6429 38 13 114 33.6603 0.9652 ***.1382 50.5762 0.9280 0.6429 38 13 114 33. 5% and 10% (two-tail) level.0081 -0.0063 11 βBALANCED 0.0024 0.1032 0.2210 0.0095 -0. respectively.0090 0.6237 1.5622 0.0282 0.7294 0.Table 11 continued α βBALANCED βHML βSMB Adjusted R-squared Number of Months Panel C: Three-Factor Model (Fama French 1992) 0.9303 0.0069 0.7269 0.7427 0.0216 -0.9347 0.9326 0.4384 0.1419 Number of Months 50.0065 -0.7314 0.0822 0.9440 0.0563 0.2338 0.2283 0.3919 0.1803 0.2266 14 Adjusted R-squared 0.5829 1.952 0.0100 0. based on T-tests 0 0   45  .0087 0.

Sign up to vote on this title
UsefulNot useful