Majalah Ekonomi Tahun XVII, No.3 Desember 2007

EVALUATING THE ACCURACY OF STOCK ANALYSTS’RECOMMENDATIONS PUBLISHED IN BISNIS INDONESIA NEWSPAPER†

Arnold Kaudin & Yustin Hendro Pranoto

Academic Staff of the Faculty of Economics Satya Wacana Christian University &Alumnus of the Faculty of Economics Satya Wacana Christian University

ABSTRACT

Many investors have not enough knowledge, skill, and time to learn which stocks arethe candidate to buy or sell. A stock analyst provides recommendation and help investorsto make buy or sell decision. The purpose of this study is to investigate the stock analysts’recommendation accuracy published in Bisnis Indonesia daily newspaper. The dataused in this research is analysts’ recommendation published in Bisnis Indonesia alongthe year 2005. There are 1,196 buy recommendations, 500 sell recommendations, and649 price recommendations; released by six security companies. The technique used tomeasure the accuracy of price prediction is the Chi Square and to measure the accuracyof recommendation to sell or buy is Wilcoxon Match Pair Test. The result shows thatthe stock price and recommendation to sell tend to be inaccurate while recommendationto buy tends to be accurate.Keywords: recommendation accuracy, stock analyst, price recommendation,

recommendation to sell, recommendation to buy

1. INTRODUCTION

The rapid development of the Indonesian financial market in recent years enablesinvestors to choose various investment instruments. One among the alternatives whichhave been growing rapidly and becoming more popular for the last decade is investmentin stock. The Capital Market Supervisory Agency reported that during the period 1995-2005, stock exchanges in Indonesia recorded an average annual increase of 12.76 percentin term of indices (Indonesian Capital Market Master Plan 2005-2009). In the JakartaStock Exchange, alone in 2005, the indices increased as high as 16.20 percent from1,000.87 to 1,162.35. The market capitalization achieved the value of 710,433,652million rupiah, grew for around 259 percent in five year period. The average tradingvolume was 73 million shares per day, around twice as high the average volume fromfive year period before. There were around 207 companies’ shares listed and around

† This paper was presented in The 4th UBAYA International Annual Symposium on Management2007. We’d like to thank the participants for valuable suggestions.

-283-

Majalah Ekonomi Tahun XVII, No.3 Desember 2007

130,000 investors serviced by more than 100 securities companies. The industry hasbeen growing as the number of players in capital market increases from year to year.

As the industry grows and becomes more complex, not all of the investors, bothinstitutional and individual, have enough expertise in analyzing the market. Most ofthem are only interested in accelerating the growth of their funds. They do not haveenough knowledge, skill, and time to learn which stocks are the candidate to buy orsell. Therefore, as the stock market develops, the profession of financial (stock) analystsexists. Many investors find that choosing and managing investments is not their forteand turn to financial analysts‡ (Penman, 2003:12). A stock analyst gives recommendationand help investors to make buy or sell decision as those people are the ones fully equippedwith knowledge and skill. Related to their professions, these people have connection tothe corporate side and thus have access to the most actual and more complete information.

Analyst report contains a description of company’s business, how the analysts expectthe company to perform, earnings estimates, price estimates or price targets for yearahead, and recommendations as to buy, to hold, or sell (Jones, 2002:298). Further,analysts should do more than simply recommend companies expected to grow rapidly.One of the most important responsibilities of an analyst is to forecast earnings pershare for particular companies because of the widely perceived linkage between expectedearnings and stock returns. Most research concerning quality of analysts entailsmeasuring the ability of analysts to determine the expected earnings (See, for instance,Andersson and Hellman [2004], Eames and Glover [2003], and Armstrong [1983]).

In fact, the recommendation and or advice made by analysts are not always accurate.For cases in U.S., according to Bodie et al. (2005:10), only the smallest fraction offirms were assigned sell recommendations and many firms given buy recommendationswere privately called or worse by the analysts. As the analysts must establish goodrelation to the management of the companies he or she in charge of, there is a possibilitythat publishing sell recommendations would be hurt for the companies’ side (themanagement) and after the sell recommendation those analysts would get trouble inaccessing information. There is a perception that sell recommendation is the beginningof distrust. It has been common in U.S. that “neutral” rating is the euphemism for sell.Some well known cases entails recommendations to sell in U.S. are the case of BostonChicken and Bank of America. Jegadeesh and Kim (2004) documents that the frequenciesof sell recommendations in G7 countries between 1993 and 2002 are far less than thatof buy recommendations. Wong (2002) provides evidence that analyst in Australia issuemore recommendations to buy compared to recommendations to sell. Recommendations

‡ In this paper the terms financial analyst and stock analyst are used interchangeably due todifferent terminologies used by different authors.

-284-

Majalah Ekonomi Tahun XVII, No.3 Desember 2007

to sell have stronger impact to prices but weaker impact to trading activity. This impliesthat analysts are conscious of the benefit of issuing positive recommendations and thecost of issuing negative recommendations.

The inaccurate recommendation may also come from inaccurate prediction of the pricemovement. Off course this is a technical matter. Jones (2002:300) adapted Hovanesianarticle in Business Week that illustrated how sell recommendation from large investmentbankers such as Credit Suisse First Boston (CFSB), UBS Warburg, Merrill Lynch,Morgan Stanley Dean Witter, etc may lead to wrong investment decision. For example,among 1,328 stocks covered by CFSB, only eleven stocks were recommended to besold. Ironically, among those eleven, after 52 weeks, only five were really declined andone other was stagnant. Three from four stocks recommended to be sold by UBS Warburghad grown for 64.2 percent, 9.3 percent, and 17.9 percent.

Bodie et al. (2005:10) argue that agency problems may lead to scandals. Conflict ofinterest and distorted incentives played a role in these scandals as the analysts werecommonly compensated not for the accuracy or insightfulness of their analysis, but fortheir role in garnering investment banking business for their firms. Analysts working inbrokerage house receive commission based on trading volume which is affected bytheir recommendations (Palepu et al., 2004:9). In U.S, the conflict of interest betweenthe analysts who play in stock research business and its investment bank materializedin new regulations adopted by NASD and NYSE that sever the ties between investmentbank and research department. Madureira (2004) finds that after the new regulationsanalysts provide less optimistic ratings for the stocks. Further, after the regulations, thebig 10 brokerage houses have been twice more likely to put a stock in a pessimisticrating than the non-big 10. This indicates that before the regulation, the bigger thebrokerage houses, the bigger the possibility of recommendations manipulated.

Another study concerning the conflict of interest was conducted by Agrawal and Chen(2007). They find that the level of analysts’ stock recommendations is positively relatedto the magnitude of the conflict of interest between the investment bank and researchdepartment. The optimistic bias stemming from the conflict was pronounced during thelate 1990s stock market bubble. However, they also find that investor are sophisticatedenough to adjust for the bias. Another finding is when the analysts’ investment bankwas the lead underwriter of an IPO of the recommended stock in the past five years ora secondary IPO in the past two years (mentioned as affiliated analyst), therecommendations are more optimistic than if the analysts’ investment bank had nounderwriting relation to a company’s recommended stock. However, the affiliatedanalysts’ earnings forecasted found to be more pessimistic than the unaffiliated analysts’forecast (Malmendier and Shantikhumar, 2005)

Accuracy of recommendations becomes the eminent point in this study. Investors basedtheir decisions, or at least some part of their decisions, on the analysts’ advice or

-285-

Majalah Ekonomi Tahun XVII, No.3 Desember 2007

recommendations. Thus, it becomes important to evaluate the performance of the stockanalysts. So far, it is very difficult to find research concerning stock analysts’recommendation accuracy for the Indonesian case. There is still a lack of empiricalevidence whether the Indonesian analysts’ recommendations are informative. Theobstacle to make such an evaluation is the data, which is very limited and only distributedto investment bankers’ customers exclusively. Media such as Bisnis Indonesia dailynewspaper, Tabloid Kontan, Investor magazine, e-bursa.com, and some others providepublicly available stock recommendations.

The purpose of this study is to investigate the stock analysts’ recommendation accuracypublished in Bisnis Indonesia, which is considered as the most consistent mediapublishing recommendations from stock analysts; and to arrange the analysts’ rankingbased on the level of prediction accuracy. Before going through a deeper exploration ofsecurity analysts’ behavior and the market’s perception, this preliminary study isconducted to determine the accuracy of stock analysts’ recommendations for theIndonesian case.

2. THEORETICAL BACKGROUND AND PRIOR STUDIES

Rather than do their own security analysis, individual investors may choose to rely onthe recommendations of the professionals (Jones, 2002:298). When a stock analystdecides that a stock will experience a change in price and inform his or her findings toinvestors, the investors will give response to the information. Thus the price will beaffected. If the information widespread, there will be more responses from more investorsand the price will be affected more. The analysts predict business prospect to findmispriced securities. They provide recommendation to buy for the under-priced securitiesand recommendation to sell for the overpriced securities. When an analyst decides thata stock value is under-priced, the response should be actions to buy which result in anincreasing stock price. On the contrary, when an analyst decides that a stock value isoverpriced; the response should be actions to sell which result in a decreasing stockprice. Observations show that stock prices in the New York and Tokyo Stock Exchangeare highly affected by recommendations from analysts (Bawazer, 1991).

Though theoretically all brokerage house customers and portfolio managers who receiveanalysts’ expert advice should achieve investment success, in reality there are severalfactors that make it difficult to consistently outperform the market (Reilly and Norton,2006:510):1. Efficient markets enable the market to review and absorb information and therefore

stock prices generally approximate fair market value. With many market players, itis difficult for investors to find situation where stock may not be fairly valued.

2. Most analysts spend most of their times in a relentless search of one or more contactor one more piece of information. This preoccupation with more information cankeep the analysts’ mind off the final output – that is, their stock recommendations.

-286-

Majalah Ekonomi Tahun XVII, No.3 Desember 2007

3. There are forces pulling on the sell-side analyst. If the investment bankers areassisting a firm in a stock or bond offering, it will be difficult for an analyst to issuea negative evaluation of the company. Besides, the analyst is in frequent contactwith the top officers of the company he or she analyzes. Although there are guidelinesabout receiving gifts and favors, it is sometimes difficult to separate personalfriendship and impersonal corporate relationships.

Analysts based their recommendations on some analytical tools. They analyze currentperformance of the target and its prospect. They estimate the future earnings per shareof the target, determine whether the current market price is over or under valued, andfinally provide recommendations to buy, to hold, to sell (also mentioned as stock ranksby Morgan and Stocken, 2001) or the variety of these basic forms of recommendations.Nevertheless, missed recommendations have been common and create skeptical amonginvestors or other users of analysts’ recommendations. Some argue that it is not alwaysclear how the recommendation follows from the analysis, or indeed whether it is justified(Penman, 2003:12) and the motives of the analysts providing advice may not betransparent (Morgan and Stocken, 2001).

On the other hand Bradshaw (2004) proves that analysts’ recommendations do workfor the investors. He links valuation techniques used by analysts to recommendationsby examining consistency between analysts’ earnings forecast and their stockrecommendations. He used four valuation models to link earnings forecast and stockrecommendations. He found that analysts’ recommendations are more correlated withheuristic valuations model than with present value models, and buy-and-hold investorswould earn higher returns relying on present values models that incorporate analysts’earnings forecasts than on analysts’ recommendations. Bradshaw also explained thatpersonal opinions or biases dominate recommendations based on present value model.Wong (2002) explains that analysts’ recommendations in Australia were observed to beassociated with abnormal returns on the day they were officially released to clients.Stocks issued with sell recommendations continued to experience negative abnormalreturns in the post-recommendation period while returns to buy recommendations werepartly due to price pressure. Hold recommendations were informative as buy ratherthan sell signals. She also finds that abnormal returns were observed in the pre-recommendations period, means that analysts had poor timing ability, reactive in makingrecommendations, recommend privileged clients first, and left behind the traders. Similarto Wong, but in term of volume (U.S case), Puckett (2002) finds that analyst’ initiationsof recommendations to buy are preceded by abnormal institutional volume and abnormalnet buying positions taken by those institutions for up to four days before the initiationsof coverage are made public.

Welch’s findings (2000) probably can provide an explanation about the abnormal returnduring pre-recommendations period. According to Welch, analysts’ recommendations

-287-

Majalah Ekonomi Tahun XVII, No.3 Desember 2007

may also be affected by prior recommendations. He finds that an analyst’srecommendation revision has a positive influence on the next two analysts’ revisions.The influence can be traced to short-lived information and is stronger when short runex-post returns are accurately predicted by the revisions. The influence becomes strongerduring the bullish market.

Baber et al. (2002) study the returns of stock recommended by analysts over the 1996-2000 periods in U.S. They find that during the period 1996-1999 the more highlyrecommended stocks earned greater market-adjusted returns than did those that wereless highly recommended. The opposite happened in year 2000 where the least favorablyrecommended stocks earned an annualized market-adjusted returns of 48.66 percentwhile the stocks most highly recommended fell 31.20 percent. The pattern prevailedduring most months of 2000, regardless of whether the market was falling or rising,and was observed for both tech and non-tech stocks. Further they propose that theresult should add to the debate over the usefulness to investors of analysts’ stockrecommendations. Ho (2005) examines whether investment decisions based on analystratings without studying the full reports is sound. He finds that analysts are unable toforecast stock performance in the short term (1-2 weeks) as during the short period theexcess return are indistinguishable from zero. Analysts’ ratings reflect excess returnbest about seven months later.

Another study by Jegadeesh et al. (2004) provides evidence that stocks which receivehigher recommendations tends to have positive momentum and higher trading volume.On average, stocks favorably recommended by the analysts outperform stocksunfavorably recommended by the analysts but the predictive ability of the level ofanalyst recommendation is not significant. This poor performance is caused by theanalysts’ failure to quickly downgrade stock rejected by some investment signals(valuation multiples, accounting accruals, capital expenditure). They conclude thatanalyst recommendations may be partly driven by incentives that are not entirely relatedto the investment performance of the analysts’ recommendations.

A study by Ribeiro et al. (2004) provides evidence from Portuguese investment banksthat trading strategies based on analysts’ recommendations have a negative performanceregardless of the investment horizon. However, they find a positive significant returnon the day the recommendations are published. They conclude that the positive impactis partially consistent with market efficiency and supports that analysts have forecastingskills.

Stock analysts are highly trained individuals who possess expertise in financial analysisand background of the industry. Theoretically, all brokerage house customers andportfolio managers who receive analysts’ expert advice should achieve investmentsuccess (Reilly and Norton, 2006:510). As analysts have knowledge, skill, andinformation, we believe that the price and buy or sell recommendation given is based

-288-

Majalah Ekonomi Tahun XVII, No.3 Desember 2007

on a deep calculation and analysis. Thus, our hypotheses are that analysts’ pricerecommendation and their recommendation to buy or sell are accurate.

3. SAMPLES AND RESEARCH METHOD

The population of this research is all analysts’ recommendations published in the BisnisIndonesia newspaper along the year 2005. The samples include price recommendationsand clear recommendation to sell or buy the stocks. More specific recommendationssuch as exit, take profit, hold, etc are excluded to avoid bias in interpretation. There are649 price recommendations, 500 sell recommendations, and 1,196 buyrecommendations; released by six security companies: Bhakti Sekuritas, BNI Sekuritas,Mandiri Sekuritas, Ciptadana Sekuritas, Trimegah Sekuritas, and Sarijaya Sekuritas.

The price recommendation is considered to be accurate when the price at the secondday after the recommendation matches with the two-days before recommendation price.We use the two days before and after recommendation date price by consideration thatthe two-day period is a neutral time horizon because:1. The recommendation must be for a short-term horizon since it is published in a

daily newspaper.2. In many occasions, analysts provided more than one recommendation for one stock

in a month.

Nevertheless, the recommendations may also affect the price at the recommendationdate, one day after the recommendation date, five days after the recommendation date,etc. The technique used to measure the accuracy of price prediction is the Chi SquareTest. We assume that price recommendations are considered to be accurate when theproportion of correct price recommendations (in aggregate) achieves the 50 percentlevel.

To determine the effect of the sell or buy recommendations, we compare the two daysbefore recommendation date price to the two days after the recommendation date price.The recommendation to sell is considered to be accurate when the price at the secondday after the recommendation date is lower than two-days before recommendation price.The buy recommendation is considered to be accurate when the price at the second dayafter the recommendation is higher than the two-days before recommendation price. Tomeasure the accuracy of recommendation to sell or buy is Wilcoxon Match Pair Test.

The ranking of security companies which provide the highest level of accuracy fortheir price recommendation, recommendation to sell and recommendation to buy is

After the Kolmogorov-Smirnov Test, we find that the distribution of the data is not normal.Thus we use the non parametric technique of Wilcoxon Match Pair Test to measure the differ-ence between the two days before the recommendation price and the second day after the rec-ommendation price.

-289-

Majalah Ekonomi Tahun XVII, No.3 Desember 2007

determined by comparing the number of accurate recommendations. The securitycompanies included in the ranking process must produce not less than 52recommendations during the year 2005. The number 52 is established by considerationthat there must be at least one recommendation (in average) released for every week.Security companies which released too few recommendations cannot be compared tothose which actively published their recommendations.

4. RESULTS

Table 1 show that Bhakti Sekuritas was the most productive security company publishingstock recommendations in Bisnis Indonesia in 2005. It released 671 recommendations,which comprised of 221 price recommendations, 313 recommendations to buy, and137 recommendations to sell. BNI Sekuritas following in the second place with 574stock recommendations, which comprised of 345 recommendations to buy and 229recommendations to sell. Almost all of recommendations published by Mandiri Sekuritaswere price recommendations while Trimegah seems to prefer releasing recommendationsto buy or to sell. Ciptadana Sekuritas was the least active among the six as it onlyreleased 70 recommendations for the whole year.

Table 1 The Number of Recommendations Released By the Analysts

Source: Bisnis Indonesia in year 2005, processed

The following are the test results and discussions from price recommendation accuracy,recommendation to sell accuracy, recommendation to buy accuracy, evaluation ofaccuracy from each recommendation provider, and ranking of accuracy.

4.1 Test Result of the Price Recommendation Accuracy

Among 649 price recommendations given along 2005, only 37 (or 5.7 percent)recommendations matched the target and the other 612 (or 94.3 percent) were missed.As mentioned before, we assume that price recommendations are considered to beaccurate when the proportion of correct price recommendations achieves the 50 percent

-290-

Majalah Ekonomi Tahun XVII, No.3 Desember 2007

level. The Chi Square Test results in a Chi Square score of 509.438 with significancelevel of 0.00 (Lower than 0.05). This means that the proportion of correctrecommendation is not 50 percent. Thus, in general, the price recommendations are notaccurate.

4.2. Test Result of the Recommendation to Sell Accuracy

The recommendation to sell is considered to be accurate when the price at the secondday after the recommendation date is lower than the price at two-days beforerecommendation date. The following table shows the result of Wilcoxon Match PairTest for the accuracy of recommendation to sell:

Table 2Wilcoxon Match Pair Test for the Accuracy of Recommendations to Sell

Among 500 recommendations to sell published in Bisnis Indonesia along 2005, thereare 239 data of price differentiation which have negative ranks. This means that theprice at the second day after the recommendation date is lower than the price at two-days before recommendation. The other 219 data of price differentiation have positiveranks, which show that the price at the second day after the recommendation date ishigher than the price at two-days before recommendation. The other 42 data shows thatthe price at the second day after the recommendation date matches to the price at two-days before recommendation.

The p-value of the Wilcoxon Match Pair Test is 0.290. As the test was one sided, thevalue, divided by two equals to 0.145, which is higher than 0.05. Therefore, we acceptthat the price at the second day after the recommendation date is higher than or matchesto the price at the two-days before recommendation date, which means that the sellrecommendation is not accurate.

During year 2005 the Indonesian stock market experienced a bullish period. Manyrecommendations to sell became not relevant. Many stocks predicted to looseexperienced increasing in price because of the domination of sentiment to buy duringthe bullish period. Therefore, in general, the recommendation to sell is not accurate.

-291-

a.SELL_AFT < SELL_BEFb. SELL_AFT > SELL_BEFc.SELL BEF = SELL AFT

Source: data processed

Majalah Ekonomi Tahun XVII, No.3 Desember 2007

4.3 Test Result of the Recommendation to Buy Accuracy

The recommendation to buy is considered to be accurate when the price at the secondday after the recommendation date is higher than the price at two-days beforerecommendation date. The result of Wilcoxon Match Pair Test for the accuracy ofrecommendation to buy is shown in the following table:

Table 3Wilcoxon Match Pair Test for the Accuracy of Recommendations to Buy

Among 1,196 recommendations to buy published in Bisnis Indonesia along 2005, thereare 711 data of price differentiation which have positive ranks, means the price at thesecond day after the recommendation date is higher than the price at two-days beforerecommendation. The other 376 data of price differentiation have negative ranks, whichmeans that the price at the second day after the recommendation date is lower than theprice at two-days before the recommendation date and the other 109 data price showsthat the price at the second day after the recommendation date matches to the price attwo-days before the recommendation date.

The p-value of the Wilcoxon Match Pair Test is 0.000 which is lower than 0.05.Therefore, we reject that the price at the second day after the recommendation date islower than or matches to the price at the two-days before recommendation date. As it ishigher, the recommendation to buy is generally accurate.

The bullish period in the stock market had strengthened sentiment to buy and pulled theprice up in general. Thus many recommendations to buy match the targeted price or thetargeted trend. In case the investors read and consider recommendations published inBisnis Indonesia, the result indicates that those investors give immediate response (twodays) to the recommendation.

4.4 The Analysts’ Level of Accuracy

Table 4 shows the level of accuracy of price recommendations per analyst:

-292-

a. BUY_AFT < BUY_BEFb. BUY_AFT > BUY_BEFc. BUY_BEF = BUY_AFT

Source: data processed

Majalah Ekonomi Tahun XVII, No.3 Desember 2007

Table 4The Level of Accuracy of Price Recommendation

Source: data processed

As mentioned before, the security companies included in the ranking process mustproduce not less than 52 recommendations along the year with consideration publishingone price recommendation in average every week. The table shows that Mandiri Sekuritasprovided 356 price recommendations with 8 percent level of accuracy, Bhakti Sekuritasprovided 221 price recommendations with 3 percent level of accuracy, and CiptadanaSekuritas provided 56 price recommendations with 2 percent level of accuracy. Themore price recommendations provided, the higher the level of accuracy, but the level ofaccuracy remains low.

In fact, price recommendations can be separated into price recommendation to sell andprice recommendation to buy as shown in Table 5. Panel 1 shows the pricerecommendation to buy. In general, the level of accuracy from price recommendationto buy obtains 46 percent. Mandiri Sekuritas was the most active producer of pricerecommendations to buy with its 354 recommendations, which achieved 53 percentlevel of accuracy. The second most active producer was Bhakti Sekuritas (121recommendations) whose level of accuracy was only 25 percent, lower than CiptadanaSekuritas which produced only 56 recommendations but achieved 41 percent level ofaccuracy.

Panel 2 shows the price recommendation to sell from each analyst. In general, the levelof accuracy was only 21 percent. Bhakti Sekuritas was the most active (if not the only)producer of price recommendation to sell with 100 recommendations which achieved19 percent level of accuracy. It seemed that the bullish market did not attract manyanalysts to produce price recommendations to sell.

-293-

Majalah Ekonomi Tahun XVII, No.3 Desember 2007

Table 5Price Recommendation to Buy and Price Recommendation to Sell

Source: data processed

The following table shows the accuracy of recommendation to sell:

Table 6The Level of Accuracy for the Recommendation to Sell

Source: data processed

-294-

Majalah Ekonomi Tahun XVII, No.3 Desember 2007

The aggregate level of accuracy for recommendation to sell is 48 percent. The tableshows that the best three recommendations to sell provider were Sarijaya Sekuritaswith 63 percent level of accuracy, followed by Bhakti Sekuritas (58 percent), andTrimegah Sekuritas (43 percent). Again, the low level of accuracy from recommendationto sell might be caused by the bullish market period.

Table 7The Level of Accuracy for the Recommendation to Buy

Source: data processed

Table 7 shows that Sarijaya Sekuritas had the highest level of accuracy for therecommendations to buy. Among 172 recommendations it released, 130 were accurate.The level of accuracy was 76 percent, and this level is considered as the best1. In thesecond place is Trimegah Sekuritas which produced 348 recommendations to buy andachieved 62 percent level of accuracy, followed by Bhakti Sekuritas (313recommendations with 58 percent level of accuracy), and BNI Sekuritas (345recommendations with 51 percent level of accuracy). In aggregate, the level of accuracyfor recommendations to buy is 59 percent.

We consider that the level of accuracy from recommendation to buy has not been areflection of the analysts’ quality since during bullish period the price moves in anuptrend. Thus, any recommendations to buy have a bigger opportunity to match thetarget price.

5. CONCLUSIONS

This study examines the accuracy of stock analysts’ recommendations published inBisnis Indonesia newspaper along year 2005. Our findings are:

1 Mandiri Sekuritas achieved 75 percent level of accuracy. However; it only produced fourrecommendations along the year and therefore not comparable to the ones which actively pro-duced the recommendations.

-295-

Majalah Ekonomi Tahun XVII, No.3 Desember 2007

1. The proportion of accuracy from price recommended by the analysts does not achieve50 percent level. Thus, we conclude that in general the price recommendationpublished in the Bisnis Indonesia newspaper is not accurate.

2. The two-days after recommendation price of stocks recommended to be sold arehigher than the two-days before recommendation price. If the recommendation isaccurate, it should be lower. This means that in general the sell recommendation isnot accurate.

3. For the recommendation to buy, the two-days after recommendation price of stocksrecommended to be sold are significantly higher than the two-days beforerecommendation price. Therefore, we conclude that the recommendation to buy isgenerally accurate.

4. The best rates of accuracy for price recommendation published in Bisnis Indonesianewspaper along 2005 are as follows: Mandiri Sekuritas (8 percent), Bhakti Sekuritas(3 percent), and Ciptadana Sekuritas (2 percent).

5. The best three rates of accuracy for sell recommendations published in BisnisIndonesia are as follows: Sarijaya Sekuritas (63 percent), Bhakti Sekuritas (58percent), and Trimegah Sekuritas (43 percent). The best three rates of accuracy forbuy recommendations are Sarijaya Sekuritas (76 percent), Trimegah Sekuritas (62percent), and Bhakti Sekuritas (58 percent).

6. LIMITATIONS

Some limitations from this study are:1. The recommendations resulted from fundamental and technical analysis cannot be

detected since the information to make distinction cannot be achieved. Besides,recommendations that are published in Bisnis Indonesia daily newspaper are therepresentation of financial institutions (securities companies), not the individualanalyst.

2. Some more specific recommendation types such as buy on weakness, sell on strength,trading buy, trading sell, accumulate buy, speculative buy, hold, take profit, exit areexcluded from samples to avoid ambiguity. The samples includes only clear “buy”or “sell”recommendations.

3. The two days before and after recommendation price that is used as time indicatorto evaluate the accuracy of buy and sell recommendations; and the 50 percent levelof accuracy to evaluate the accuracy of price recommendation can be considered assubjective.

4. The results do not reflect the real quality of the analysts as the observation isconducted during bullish market period.

-296-

Majalah Ekonomi Tahun XVII, No.3 Desember 2007

REFFERENCES

Agrawal, Anup and Mark A. Chen, 2007, Do Analyst Conflict Matter? Evidence fromStock Recommendations, www.bama.ua.edu, accessed on January 2007

Armstrong, J. Scott, 1983, Relative Accuracy of Judgemental and Extrapolative Methodsin Forecasting Annual Earnings, Journal of Forecasting, vol. 2, pp 437-447

Andersson, Patric and Niclas Hellman, 2004, The Impact of Pro Forma Profits on AnalystForecast: Some Experimental Evidence, SSE/EFI Working Paper Series in BusinessAdministration no 2004:5, http://swoba.hhs.se/hatsba/, accessed on January 2007

Baber, Brad, Reuven Lehavy, Maureen McNichols, and Brett Trueman, 2002, Prophetsand Losses: Reassessing the Returns to Analysts’ Stock Recommendations, FinancialAnalysts Journal 59, no 2, pp 88-96

Badan Pengawas Pasar Modal, 2005, Indonesian Capital Market Master Plan, BadanPengawas Pasar Modal Departemen Keuangan Republik Indonesia

Bawazer, Said A., 1991, Sekilas Tentang Pialang Bursa Efek Jakarta, Manajemen danUsahawan Indonesia, no. 12, year XX, December, pp 13

Bodie, Zvi, Alex Kane, and Alan J. Marcus, 2005, Investments, Sixth edition, McGraw-Hill International Edition

Bradshaw, Mark T., 2004, How Do Analyst Use Their Earnings Forecast in GeneratingStock Recommendation?, The Accounting Review, vol. 79 no. 1, January, pp 25-50

Eames, Michael J. and Steven M. Glover, 2003, Earnings Predictability and the Directionof Analysts’ Earnings Forecast Errors, The Accounting Review, vol. 78 no. 3, July,pp 707-724

Ho, Kevin, 2005, Analysts: Oracles?, Working paper, Columbia University

Jegadeesh, Narasimhan, Joonghyuk Kim, Susan D. Krische, and Charles M.C. Lee,2004, Analyzing the Analysts: When Do Recommendations Add Value?, Journal ofFinance 59, pp 1083-1124

Jegadeesh, Narasimhan and Woojin Kim, 2004, Value of Analysts’ Recommendations:International Evidence, Working paper, Emory University

Jones, Charles P., 2002, Investments, eight editions, John Wiley & Sons, Inc.,International Edition

Madureira, Leonardo, 2004, Conflicts of Interest, Regulations and StockRecommendations, Working paper, The Wharton School, University of Pennsylvania

Malmendier, Ulrike and Devin Shantikhumar, 2005, Do Security Analysts Speak inTwo Tongues?, Working paper, www.faculty-gsb.stanford.edu, accessed on January2007

Morgan, John and Phillip Stocken, 2001, An Analysis of Stocks Recommendations,www.faculty.haas.berkeley.edu, accessed on January 2007

-297-

Majalah Ekonomi Tahun XVII, No.3 Desember 2007

Palepu, Khrisna, Paul Healy, and Victor Barnard, 2004, Business Analysis and ValuationUsing Financial Statements, Third edition, Thomson South Western

Penman, Stephen H., 2003, Financial Statement Analysis and Security Valuation, Secondedition, McGraw Hill, International Edition

Puckett, Andy, 2002, Who Trades on Analysts’ Recommendations?, www.fma.org,accessed on January 2007

Ribeiro, João Martins, Rúben Barreto, and Ana Paula Serra, 2004, Analysts’Recommendations: Evidence from a Portuguese Investment Bank, Working paper,www.efmaefm.org, accessed on January 2007

Reilly, Frank K. and Edgar A. Norton, 2006, Investments, Seventh edition, ThomsonSouth Western, International Edition

Welch, Ivo, 2000, Herding among Security Analysts, Journal of Economics Literature,vol. 58, issue 3, pp 369-396

Wong, Kathryn L., 2002, The Impact of Analysts’ Recommendations, Working paper,University of Sidney, SIRCA, and ABN AMRO Australia, www.sirca.org.au,accessed on January 2007

-298-