Returns to Australian investment newsletter recommendations.

Abstract:

We investigate the recommendations made by four high profile Australian investment newsletters during 2001. Five issues are investigated: (1) The clarity of newsletter recommendations; (2) the return that might have been earned by following newsletter recommendations; (3) the liquidity of shares recommended; (4) the relative success of recommendations on small, medium and large stocks; and (5) the extent newsletters make recommendations consistent with momentum and contrarian investing strategies. We find that newsletters contain a varying mix of clear and less clear recommendations. The clear recommendations we evaluated earned exceptionally high return, relative to control samples matched on size and holding period and relative to the ASX300 index. The exceptional performance was driven by recommendations on stocks ranked outside the top 150, by market capitalization, and was consistent with a momentum based selection strategy.

Keywords:

INVESTMENT NEWSLETTERS; MARKET EFFICIENCY; SHARE RETURNS.

1. Introduction

Investment newsletters are a thriving cottage industry in Australia. There are presently about 20 newsletters, with annual subscription rates ranging from $199 for Bioshares to $990 for marketmad.com. The best known, The Rivkin Report, is said to generate revenue of about $13 million a year from around 17,000 subscribers.

Unsurprisingly, the newsletters claim subscribers can earn superior returns following their advice, but as Dr Michael Dunn, ASIC's director of consumer communications, has noted, this boast has not been subject to rigorous independent evaluation in Australia. Comprehensive U.S.-based studies have concluded that investment newsletters recommendations do not yield abnormal returns. (1) However, if the degree of information search undertaken per security is proportional to market capitalisation, as suggested by Atiase (1985), then there is reason to expect the Australian market to be less efficient than the U.S. market and thus offer greater scope for newsletters to identify mispriced securities. Seventy five percent of the firms listed on the Australian stock exchange (ASX) are smaller than the median-sized firm in the New York stock exchange (NYSE) decile comprising the smallest firms by market capitalisation. Indeed, the median firm from the largest ASX size decile would belong in the fifth NYSE size decile.

We assess returns to recommendations made by four prominent Australian investment newsletters between January 2001 and January 2002. The newsletters are The Rivkin Report, The Intelligent Investor, The GTS Report (since renamed Growthstocks and, subsequently, discontinued), and Huntley's Your Money Weekly. We address four issues:

1. The clarity of the newsletters' recommendations;

2. The potential returns that may be earned by systematically acting on them;

3. The effect of illiquidity on the returns an investor could earn;

4. The relative success of recommendations on small, medium and large stocks; and

5. The extent newsletters employ momentum and contrarian trading strategies.

Our inquiry into the clarity of the newsletters' recommendations is prompted by the informal style of writing they use in making the recommendations. An advantage of this style is that it may serve to make the recommendations more accessible to retail (i.e. individual, non-professional) investors who might be daunted by precise, technical language. The disadvantage is that informal language can serve (perhaps inadvertently) to mask ambiguous recommendations. (2) Ambiguous recommendations lower the information content of newsletters and so their prevalence is an important empirical question.

Our second research issue, the returns that may be earned by acting on identifiable clear recommendations, is the core question we address. We introduce an innovation in investment newsletter research by tracking returns to portfolios of recommendations in event time rather than calendar-time investment portfolios. To illustrate; we identify 43 recommendations to buy stocks from The Rivkin Report during 2001. These were made at different times during the year and had varied recommended holding periods. We define the portfolio return as the simple average of the returns from acting on the 43 recommendations. Our portfolio return therefore does not therefore relate to any particular calendar time period of length of time, that is, it is the average return to a portfolio of stock recommendations separated in time that have different holding periods. We believe our approach improves on methods that measure abnormal performance in calendar time by estimating regressions of returns on factors that proxy for risk. The calendar time method is prone to error because of the need to simultaneously estimate factor coefficients and abnormal returns (Mettrick 1999; Kothari & Warner 2001). Further, the implicit portfolio rebalancing that is required each time a new recommendation is made does not accurately reflect investor experience.

In another innovation in investment newsletter research, we assess significance by comparing the return to our experimental portfolio against the distribution of returns to control portfolios comprising firms matched on size-decile with the recommended stocks but otherwise randomly selected. Kothari and Warner (1997) point out that non-parametric tests of this type reduce likely misspecification of test statistics.

One way by which investment newsletters may appear to facilitate superior performance is by making recommendations on illiquid stocks. Price recorded on illiquid stocks may appear to indicate substantial gains but they are illusory because they cannot be exploited. The degree of illiquidity on the ASX may be appreciated from our finding that for the ten per cent smallest firms, just $37,000 worth of stock is traded over five days, on average. To test the robustness of our results, we review returns to newsletter recommendations after eliminating those made for relatively illiquid stocks. We also review the relative success of recommendations on small, medium and large stocks as another proxy for liquidity and also to determine if the newsletters perform better on the smaller stocks in which there is arguably greater information asymmetry.

Finally, Jegadeesh and Titman (1993, 2001) find evidence in the U.S. that selecting stocks on the basis of past momentum can yield abnormal returns even after controlling for risk. Demir, Muthuswamy and Walter (2003) confirm the existence of a momentum effects in Australian stocks. To assess the extent that momentum effects play in returns to newsletter recommendations, we partition our samples on momentum characteristics. We also separately review returns to recommendations that are consistent with a contrarian selection policy.

Our main findings may be summarized in five points:

1. The four newsletters provided a substantial number of clear buy recommendations during the sample period, January 2001 to December 2001, that could have been followed by investors. We identified 593 clear recommendations across all four newsletters;

2. The stocks recommended by the newsletters delivered significant out-performance. A portfolio comprising the full sample of 593 recommended stocks earned 12.3% and outperformed all 1000 of its control portfolios. (3) An equivalent set of trades in the ASX300 index matched on the recommended stocks' holding periods would have earned 1.6%;

3. The returns that an investor could have earned would not have been reduced by inability to trade because of lack of liquidity, unless the investor confined themselves to stocks in the largest size decile. Recommendations on the largest (and most liquid) stocks did not yield exceptional returns;

4. The investment newsletters demonstrated greater aptitude at picking small stocks than medium stocks and large stocks; and

5. The performance of the newsletter portfolios is driven by a momentum effect. Recommendations that were consistent with a momentum strategy delivered significantly greater returns than recommendations consistent with a contrarian strategy. However, the newsletters in general made an approximately equal number of momentum and contrarian recommendations.

2. Description of Sample and Research Method

We review recommendations made by four Australian investment newsletters: The Rivkin Report, The Intelligent Investor, The GTS Report, and Huntley's Your Money Weekly. All recommendations were in newsletters issued between January 2001 and January 2002 and were for ordinary shares on ASX-listed companies. (4) The number of newsletters we review and the sample period from which we collect recommendations are dictated by the limits imposed by our labour- and time-intensive research design. However, we believe our sample is representative of Australian investment newsletters. McDonald (2003) reports there are about 20 newsletters in Australia. The four we review have been listed among the best-known (Infochoice 2003) and, as may be seen from the descriptions below, represent a broad range of stock selection approaches and include the oldest investment newsletter in Australia, Huntley's Your Money Weekly.

2.1 The Rivkin Report

The Rivkin Report has the biggest subscriber base in Australia and is easily the highest-profile Australian investment newsletter (Main 2003). (5) It was started in 1997. The Rivkin Report seeks to identify opportunities brought about by events that affect individual stocks, such as take-overs, rather than forecast market movements (Dobbie 2003). Its ostensible strategy is to focus on blue chip stocks that are out of favour, and on stocks subject to take-over bids. (6) Subscriptions cost $799 per year for approximately 45 issues.

2.2 The Intelligent Investor

The Intelligent Investor is published fortnightly and focuses on recommending larger stocks. The stated strategy of The Intelligent Investor is to consider the financial fundamentals, management and competitive position of firms with the aim of identifying stocks that will outperform in the long term. Subscriptions cost $395 per year.

2.3 The GTS Report

The GTS Report is the only one of our newsletters to have ceased publication since 2001. GTS stands for Growth, Technology, and Speculative. The GTS Report was published by the same company as The Intelligent Investor, and had the same editor. It provided analysis and recommendations on small and medium sized firms. The name of the report was later changed to Growthstocks and the service closed on 1st August 2003. The newsletter employed the same value investing strategy as The Intelligent Investor, but focused exclusively on smaller stocks that were considered to have significant growth potential. These included technical stocks, biotechnology stocks, small mining stocks, small industrial stocks and new issues. Subscriptions cost $550 per year.

2.4 Huntley's Your Money Weekly (i.e. Huntley's)

Huntley's focuses on blue chip and second-line stocks, where a second-line stock is riskier than a blue chip but has reliable earnings. Huntley's employs a value investing strategy in that it focuses primarily on companies' earnings and the sustainability of earnings. It advocates a long-term investment strategy, suggesting that investors can avoid most market volatility by investing in companies for 3-5 years at a time. (7) The newsletter was started in 1973 and is published weekly. Subscriptions cost $537 per year.

2.5 Interpreting the Newsletters' Recommendations

In essence, all the newsletters are collections of recommendations, along with some analysis and rationale for each recommendation. To ensure that the recommendations selected for evaluation give the investment newsletters the best chance of showing success and to enhance the replicability of our analysis, we model our selection procedure on a hypothetical investor who is conservative when reading the newsletters and acting on recommendations. Our operational definitions of conservatism entails acting only on clear 'buy' recommendations and, asymmetrically, on all 'sell' recommendations, regardless of clarity. This asymmetry in acting on buy and sell recommendations is prompted by our observation that after issuing even vague sell recommendations, the newsletters disavowed responsibility for stocks that subsequently performed poorly. We thus classify as sell recommendations those that counselled readers to: 'hold/take part profits', 'hold/lighten', take part profits', 'lighten holdings', 'traders lighten', 'stop losses' and 'sell'. This rule, along with the principle of acting only on clearly indicated buy recommendations, ensures that our experimental sample portfolio does not include stocks that the newsletter editors have indicated any bearish sentiment or are unsure.

If the recommendation is 'buy', 'speculative buy' or 'long-term buy', it is included in our sample. We include recommendations where the instruction is 'accumulate'. We do so because newsletters generally treat accumulate recommendations as if a buy recommendation had been issued. (8) However, we acknowledge that a conservative investor might choose to ignore accumulate recommendations and only follow the stronger buy recommendations and so in a separate test we remove accumulate recommendations from the sample. Where the recommendation is 'hold', we ignore it because this recommendation contains no instruction to initiate action. The newsletters frequently repeat recommendations. If the hypothetical investor already held the recommended stock, then the subsequent recommendation was followed only if the new price limit was more than 10% above or below the previous price limit.

Newsletter subscribers do not generally short-sell and, as will be shown, their recommended stocks are not usually among the small number that may be short-sold. For these reasons, we assume that sell recommendations are relevant only if preceded by a clear buy recommendation on the same stock.

2.6 Price Limits

To allow assessment of return, we assign a price limit to all buy recommendations. In all cases, we assume that recommended stock was bought on the first trading day after publication of the newsletter that the closing price was equal to or below the price limit. If the stock did not close at of below the price limit on any of the five trading days following the recommendation, the recommendation is deemed impossible to implement and we remove it from the sample. The five-day period includes the day on which the recommendation was issued. In setting the price limit, we follow each newsletter's guidelines, described next.

The Rivkin Report generally specifies the price limit. If the rationale for the recommendation is that the stock is a takeover target, we apply the stated price limit strictly. If the stock is not a 'takeover play,' we set the price limit 1% above the price cited in the recommendation.

The GTS Report does not specify price limits but publishes the most recent price of the stock and maintains the policy that investors should buy the stock if the price is within about 10% of the published price. We follow this advice for buy recommendations. Where the recommendation is to accumulate, we set the price limit 1% above the price cited in the recommendation.

The Intelligent Investor does not publish price limits or guidelines, but does publish the most recent closing price. We set the price limit 2% higher than the published price where the recommendation was to buy, and 1% higher when the recommendation was to accumulate.

Huntley's Your Money Weekly always publishes price limits but with varying instructions. Where the recommendation is to buy below $x or accumulate below $x, we use $x as the price limit. Where the recommendation is to buy near $x, we set the price limit 2% above $x. Where the recommendation is to accumulate near $x, we set the price limit 1% above $x.

For sell recommendations, we ignore price limits because we assume that the newsletter no longer endorses the stock. Newsletters no longer take responsibility for poor-performing stocks after sell recommendations have been issued, regardless of whether the price limit was hit. (9) We assume investors sold at the closing price on the day of the sell recommendation. Note that this assumption biases our estimated return to newsletter recommendations upwards, on balance, since it is unlikely that all stocks are liquid enough to be sold at the closing price on the day of the sell recommendation.

2.7 Holding Period

The holding period for each recommended stock is defined as the date it is purchased to the date a sell recommendation is issued. However, sell recommendations are issued for just a small minority of the stocks in our experimental sample. We keep track of all subsequent comment related to an initial buy recommendation for up to 12 months. If a sell recommendation is not issued in the interim, we sell at the end of the 12-month period. This procedure arguably biases downwards the performance of the Huntley's newsletter that selects stocks on the basis of a long-term value-investing approach. On the other hand, very few sell recommendations are ever issued and so the appropriate holding period is unknown. Investment newsletters tend to state their performance on an annualised basis which suggests a 12-month holding period is a reasonable default.

2.8 Return Computation and Performance Measures

The return to each recommendation is computed by dividing share price at the end of the holding period, adjusted for dividends, stock splits and rights issues, by the share price at the start of the holding period. The return to a portfolio is the simple average of the returns to a group of recommendations. As a robustness check, we also use the median return to a portfolio of recommendations as a performance measure.

2.9 Significance Tests

To assess significance, we compare the performance of our experimental portfolio of recommendations against that of control portfolios comprising firms matched on size-decile but otherwise randomly selected. Each firm in a control portfolio is held for the same calendar period as the recommended firm in the experimental portfolio with which it is matched. We generate 1000 control portfolios and the distribution of the 1000 control portfolio returns is used to assess the significance of the experimental sample portfolio. (10) For instance, we deem the return to the experimental portfolio significantly positive at the 5% level if it is greater than the return to at least 950 (at least 95%) of its 1000 control portfolios. The procedure for constructing the 1000 control portfolios is as follows:

For each recommendation in the experimental portfolio:

(1.) Calculate the size (market capitalization) of all fully paid ordinary shares listed on the ASX at the start of the month in which the recommendation is made and identify the size decile of the recommended stock;

(2.) Selecta stock from the same size decile as the recommended stock, drawing randomly and sampling with replacement, and allocate it to the first of the 1000 control portfolios. The control stock is bought and sold at closing prices on the same days as the recommended stock; and

(3.) Repeat step 2 for each of the 1000 control portfolios.

Our research method assumes that the expected return to a stock is the average return to the stocks in its size decile. For a newsletter to display positive performance, it must identify stocks that will outperform other stocks of a similar size. This method controls for the well-documented size effect (Banz 1981; Durack, Durand & Maller 2003). We do not control for book-market ratio partly on the basis of Halliwell, Heaney and Sawicki (1999) who find the book to market factor is not significantly related to return. Faff (2001) finds the book to market factor is significant, but associated with a negative risk premium. We also do not control for beta on the basis of Durack, Durand and Maller's (2003) finding that beta has a very limited role in explaining returns to Australian stocks.

As a supplementary test, we compare the performance of each experimental portfolio to a broad-based value-weighted market index. We use the S&P ASX 300 Accumulation index, which represents about 91% of the Australian market and measures total investor return, including dividends. To enable comparisons between our experimental portfolios and the index, we construct an index portfolio for each experimental portfolio. For each experimental stock, we calculate the return to the index over the term of the recommendation. The average of these returns is the return to the index portfolio.

3. Results

To put our results in context, we begin with descriptive statistics about the Australian market and the recommendations. The performance of recommended stocks is then described.

3.1 Relative Size of ASX Firms, Their Liquidity & Market Performance during Sample Period

Table 1 shows the median size of the firms listed on the ASX and the NYSE in each size decile (by market capitalization). U.S. firms are substantially larger than Australian firms. The median firm from the largest ASX size decile would belong in the fifth NYSE size decile. Seventy five percent of the firms listed on the ASX are smaller than the median-sized firm in the smallest NYSE size decile. If the degree of information search undertaken per security is proportional to market capitalisation, as suggested by Atiase (1985), then there is reason to expect the Australian market to be less efficient than the U.S. market and conclusions from U.S.-based studies on newsletter recommendations may not apply in Australia.

A review of table 1 also indicates that the change in size across the ASX deciles is not monotonic. The top 150 firms account for around 90% of the total market capitalization of all firms; the firms ranked 151 through to 300 by market capitalization account for 6% of total market capitalization; while the firms ranked 301 and downward make up just 4% of total market capitalization. We label the top 150 firms the 'large' firms, the firms ranked from 151 through to 300 the 'medium' firms while all the rest are 'small' firms. There are around 1,400 firms listed on ASX, so the category of small firms includes by far the largest number of firms. Figure 1 shows the number of recommendations made by each newsletter on large, medium and small stocks. It shows that small firms account for a disproportionately high number of recommendations relative to their share of total market capitalization value. In short, there is evidence that investment newsletters target smaller stocks about which less is known. As expected, The GTS Report recommends a greater proportion of small stocks than the other three newsletters. Table 2 indicates the liquidity of stocks in each decile. Dollar volume is used as a proxy for liquidity because the value of stock that is traded provides an indication of the ease with which an investor would be able to trade. For example, table 2 shows that for a typical stock in size decile one (smallest stocks), about $37,000 of stock is traded over a five-day period, if a large number of investors wished to buy such a stock at a particular price, it is likely that their orders would not be filled. The charts show that recommendations on stocks in the largest two size deciles are likely to be relatively easy for subscribers to follow compared to recommendations on smaller stocks. Low liquidity is likely to be the main reason that the newsletters recommend few stocks in the smallest six size deciles.

[FIGURE 1 OMITTED]

All of the recommendations in our sample were issued between January 2001 and January 2002. As discussed earlier, the maximum period any recommendation was held is 12-months so all holding periods begin and end somewhere between January 2001 and January 2003. From the beginning of January 2001 through to the end of January 2003, the S&P ASX 300 Accumulation Index, a broad index of Australian stocks, declined by 2.4%. The period during which we assess newsletter recommendations cannot be characterized as a 'bull market.'

3.2 Coincidence of Recommendations

The extent buy recommendations by the newsletters coincide is one indicator of their diversity. For instance, The Intelligent Investor and Huntley's both claim to follow value-investing strategies so a potential subscriber might be interested in often they recommend the same stock. Table 3 reveals the degree of similarity in terms of recommendations among the four newsletters by reporting the number of instances newsletters recommended the same stock within 30 days of each other. The table also tallies the incidence of recommendations on the same stock occurring more than 30 days apart but within 90 days, and more than 90 days apart but within 180 days. The results show that the recommendations of each newsletter coincide infrequently. The Intelligent Investor and Huntley's recommend the same stock within 30 days of each other on 24 occasions during the sample period. This represents just 16% of all clearly identifiable recommendations made by The Intelligent Investor and 11% of all recommendations made by Huntley 's.

3.3 Returns to Buy Recommendations

We begin by reviewing the performance of the full sample of 593 newsletter recommendations (the Full Sample experimental portfolio), and of each newsletter.

Table 4 shows impressive performance by the newsletter recommendations in our sample. The average raw return to the portfolio of 593 recommendations was 12.3%, which is greater than the return to all of the 1,000 size-and-holding-period-matched control portfolios. The median control portfolio return was 6.1% and the portfolio of holding period matched trades in the S&P 300 Accumulation Index (the index portfolio) earned 1.5%. Sixty one percent of trades delivered a positive raw return. All portfolios of recommendations beat the median return to their respective 1,000 size-and-holding period matched control portfolios.

The superior performance of the Full Sample experimental portfolio against all of its 1,000 control portfolios allows us to confidently conclude that the stocks recommended by the newsletters in our sample delivered positive risk-adjusted returns, where the risk of recommended stocks is assumed to be equal to the risk of stocks in their size decile.

In what may appear a curious result, the Full Sample portfolio of 593 recommendations outperforms all its 1,000 control portfolios but the same is not true of any of the sub-portfolios comprising recommended stocks from each newsletter. Table 4 shows the mean return to the 177 recommendations from The GTS Report was higher than the mean return to all but one of its 1,000 size-and-holding period matched control portfolios while the mean returns to the 224 Huntley's and 43 Rivkin Report recommendations beat, respectively, 75 and 81 of their 1,000 control portfolios. The 149 recommendations from Intelligent Investor performed least impressively, earning a mean return of 7.9% that was bettered by 339 control portfolios, however it should be recalled that The Intelligent Investor ostensibly concentrates on identifying undervalued larger stocks and it may require more than one year (our maximum holding period) for the newsletter's stock selection ability to be evaluated.

The explanation for the Full Sample portfolio showing better relative performance than any of its constituent portfolios lies in the positive asymmetry in firms' returns, first documented in ASX stocks by Beedles (1986). A minority of the 1,000 control portfolios will include, by chance, a disproportionately high number of firms with an extreme positive return. It is these few control portfolios that outperform the portfolios comprising the recommended stocks from the respective newsletters. The impact of a few extreme positive returns on the simple average is diluted the more stocks in a portfolio and so across-the-board positive abnormal performance in an experimental portfolio will be more evident in comparisons with control portfolios the higher the number of stocks comprising the portfolios.

Given the mean return is substantially affected by outliers, the median return arguably provides a more reliable indicator of the overall performance of the constituent firms in each portfolio. Table 5 reports the newsletters' performance on a median return basis. The findings are consistent with those in table 6. The relative performance of both The Rivkin Report and Intelligent Investor is reduced when median returns are reviewed but The Rivkin Report still manages to outperform 869 of its 1,000 control portfolios whilst The Intelligent Investor outperforms just 508.

In comparing the average returns across newsletters it should be borne in mind that they relate to different holding periods. For instance, The Rivkin Report has a higher proportion of sell recommendations made within one year of its buy recommendations than Huntley's. This is reflected in the lower average holding period for Rivkin Report recommendations at 170 days while the average holding period for Huntley's recommendations is 326 days. To facilitate appropriate comparisons across portfolios, the last row in table 4 reports annualised returns. Note that although expressing returns on an annualised basis increases the mean return to all newsletters, the least increase is achieved by Huntley's which goes from 11.3% to 12.0% whilst The Rivkin Report has its return increased from 10.2% to 23.2%, the highest annualised return. The annualised returns may indicate that The Rivkin Report performed best however other considerations temper this conclusion. For instance, The Rivkin Report provided far fewer recommendations than Huntley's in 2001 and also has an events-based stock selection strategy rather than a fundamental analysis approach, for example, stocks subject to a takeover bid are selected on the basis that the first bid is often not the last. Given that the events on which selection is based have their impact reflected in stock prices in a short period after their occurrence, evaluation of an events-based stock selection approach will require a narrower window relative to a fundamental analysis approach.

3.4 Newsletter Performance Excluding 'Accumulate' Recommendations

Given that the newsletters have superior stock selection ability it might be expected that the recommendations about which the newsletters are most confident, that is, those flagged with a clear 'buy' recommendation as opposed to an 'accumulate' recommendation, would exhibit better performance. Table 6 summarises the return to portfolios comprising just clear 'buy' recommendations. The Rivkin Report made no identifiable accumulate recommendations so its return is unaffected, however the 'buy' recommendations of the other three newsletters performed worse than their portfolios comprising both buy and accumulate recommendations. The experimental portfolios of buy recommendations delivered lower returns and were outperformed by a greater number of control portfolios. Huntley's Your Money Weekly exhibited the greatest deterioration in performance. The 224 recommendations in its 'buy and accumulate' portfolio had a mean return higher than all but 75 of its matched 1,000 control portfolios but the mean return to its 78 buy recommendations was outperformed by a majority, 570, of its 1,000 control portfolios. A possible explanation for these results is that accumulate recommendations are made subsequent to a period of good performance and as such are effectively momentum-based recommendations. The likelihood of the newsletter employing momentum-based stock selection methods is discussed in section 3.6.

3.5 The Liquidity of Stocks Recommended by Newsletters

We investigate the extent the newsletters' success is a function of their recommending relatively illiquid stocks by reviewing their returns after excluding stocks that had less than $100,000 worth of trades in the five trading days following each recommendation (including the day of recommendation). We refer to the stocks excluding after applying this criterion as the 'very illiquid' stocks. We also review returns after excluding stocks that did not have at least $500,000 worth of trades in the five trading days following each recommendation. Stocks excluded after applying the $500,000 minimum trading volume threshold are referred to as the 'illiquid' stocks.

Panels B and C of table 7 report the portfolio returns when recommendations on very illiquid stocks and recommendations on illiquid stocks are excluded. From Panel B, we see that less than 10% of the original sample of 593 recommended stocks are dropped when the $100,000 minimum trading volume threshold is applied. None of the 43 buy recommendations from The Rivkin Report are for 'very illiquid' stocks so its mean return remains unchanged. (12) For the other newsletters, excluding the recommendations they make for very illiquid stocks decreases their mean return. This outcome is arguably most surprising for The GTS Report which accounts for 38 of the 51 very illiquid stocks (i.e. 21% of its total recommendations). The GTS Report focused on smaller companies that could be labelled growth, technology and speculative stocks so our surprise is not about the high number of very illiquid stocks in its recommendation but that it showed relatively poor stock selection ability among the smallest firms.

Imposing an arguably more economically realistic minimum trading volume threshold of $500,000 over five trading days eliminates 29% of the original 593 recommendations. Recommendations from The GTS Report make up 65% of the 174 that are excluded and its performance is also the most dramatically affected, falling from 17.2% when all recommendations are reviewed to 9.2% when illiquid and very illiquid stocks are removed. The other newsletters' mean returns are relatively unaffected by imposing the liquidity constraints. This observation also applies to the aggregate portfolio which had its mean return reduced by just 2.2% from 12.3% to 10.1% and still outperformed all but 24 of its 1,000 control portfolios.

Our minimum trading volume thresholds to establish the recommended firms' liquidity is somewhat arbitrary. Market capitalisation is arguably a more accurate indicator of liquidity and also of the informational efficiency of firms' share prices. Table 8 reports mean returns to the four newsletters' recommendations when large, medium and small firms are reviewed separately. Recall that large firms are defined as those ranked in the top 150 ASX firms by market capitalisation, medium firms are those ranked between 150 and 300 and small firms are those ranked from 301 onwards. Recall also that large stocks account for around 90% of the total market capitalization of all firms; medium stocks account for 6% while small stocks make up just 4% of total market capitalization.

Table 8 shows that, in general, the recommendations on smaller stocks deliver higher raw returns than those on medium and large stocks. The exceptions are recommendations on small stocks by The Intelligent Investor, which performed poorly, and recommendations on large stocks by The GTS Report, which performed very well. The GTS Report portfolio of trades in large stocks contains only five trades, and the performance of the portfolio is driven by a single outstanding trade that delivered a return of 128%. It is not clear why Intelligent Investor recommendations on small stocks performed worse than the other newsletters.

The performance of the size-matched control portfolios reveals whether the newsletters are better able to identify opportunities in small stocks than in medium or large stocks. The performance of the Full Sample experimental portfolios indicates that this is the case. The portfolio of trades in large stocks was outperformed by 566 control portfolios, while the portfolio of trades in medium stocks was outperformed by 14, and the portfolio of trades in small stocks was outperformed by only three of the 1000 control portfolios. Each portfolio is compared to a separate set of size-matched control portfolios so the superior returns to trades in small stocks cannot be explained by a size effect.

The individual newsletters deliver mixed results on the issue of size-classification, however, some of the individual newsletter portfolios are very small, containing as few as five stock recommendations. In general, with the notable exception of Intelligent Investor recommendations on small stocks, the newsletters appear to possess greater aptitude for picking medium stocks than large stocks, and even greater aptitude for picking small stocks.

3.6 Returns to Recommendations Consistent with Momentum and Contrarian Selection Strategies

The superior performance of the returns to the newsletter recommendations could be due to their following, perhaps inadvertently, either a contrarian or momentum stock selection strategy. Table 9 reports the returns to portfolios of recommendations that are consistent with a momentum strategy (momentum recommendations) and a contrarian strategy (contrarian recommendations). We define a contrarian (momentum) recommendation as one where the recommended stock underperformed (outperformed) its size decile by more than 5 percentage points over the six months prior to the recommendation.

Table 9 shows that 224 (i.e. about 38%) of all 593 recommendations we identify are consistent with a contrarian investing style and 267 (i.e. about 45%) are momentum recommendations. The about even selection of contrarian and momentum recommendations suggests that the newsletters are not deliberately mimicking either approach, although it should be noted The GTS Report's 125 momentum recommendations substantially outnumber its 36 contrarian recommendations.

The results show a striking difference in the performance of contrarian and momentum recommendations. In all cases, portfolios of momentum recommendations strongly outperform portfolios of contrarian recommendations. Apart from The Rivkin Report, the mean return to all the newsletters' contrarian recommendations was lower than the median return to their respective 1,000 control portfolios. One element of The Rivkin Report's ostensible strategy is to focus on recommending out of favour blue chips. The evidence suggests that while The Rivkin Report earns a higher return, on average, to its contrarian recommendations than other newsletters, its performance is not outstanding being bettered by the mean return to each of 308 control portfolios.

All of the newsletters made successful momentum recommendations. The portfolios of momentum recommendations of The GTS Report, Huntley's, and the Full Sample portfolio outperformed all of their control portfolios. Our evidence is consistent with momentum based stock selections driving the remarkable superior performance of investment newsletter recommended stocks.

4. Concluding Comments

The survival of a thriving market for investment newsletters is an anomaly in the context of the efficient market hypothesis. One potential explanation is that such newsletters have, for their largely layperson subscribers, 'entertainment' value that is independent of any financial reward they may facilitate. The informal style of the newsletters is consistent with this explanation. Nevertheless, we document that newsletters do provide specific recommendations that are capable of being acted upon and evaluated. We find that, considered as a group, newsletter recommended stocks generate returns that may be confidently said to comprehensively outperform a random stock selection strategy, after controlling for the robust association of firm size with return and holding period.

Our results indicate the apparent returns one may expect from following newsletter recommendations vary in line with their ostensible stock selection strategy. A 'value' based investment strategy tends to result in larger, more liquid stocks being selected with lower return. Growth or 'glamour' based investing stock selection strategies result in higher apparent returns but liquidity constraints lower the effective return one is likely to receive. Nevertheless, even after controlling for liquidity, the investment newsletters deliver significant out-performance.

Notwithstanding the impressive performance of their recommended stocks, it is a moot point whether the newsletters' ostensible strategies explain the results. The newsletters' recommendations are not obviously biased towards selections consistent with either a momentum or contrarian strategy yet it is the momentum selections among the mid-size stocks that drive their superior performance. The newsletters show no aptitude for picking winners among small and large firms, where large firms comprise the top 150 firms by market capitalisation and small firms comprise the several hundred companies that are ranked 301 and above in market capitalisation.

(Date of receipt of final transcript: March, 2004. Accepted by Doug Foster & Garry Twite, Area Editors.)

Table 1

Size of ASX and NYSE Listed Firms

Market value of equity of the median firm in each size decile of all
firms listed on the ASX and the NYSE, as at June 2001. Size of NYSE
firms is shown in Australian dollars and U.S. dollars. The exchange
rate used is USD$0.51 = AUD$1, which was current on 30th June 2001.
The size of the median firm in each NYSE size decile was obtained
from the ME breakpoints file on Kenneth French's website. (11) All
firms on the ASX were ranked by market value of equity on 29th June
2001, obtained from (the monthly sppr-price file) and sorted into
size deciles. Decile 1 comprises the 10% smallest firms, by market
capitalization.

               Median ASX firm    Median NYSE firm    Median NYSE firm
Size Decile    $AUD (millions)    $AUD (millions)     $USD (millions)

     1               1.5                 108                 55
     2               3.9                 373                190
     3               6.6                 702                358
     4              10.4               1,157                590
     5              16.3               1,819                928
     6              26.4               2,669              1,361
     7              47.1               4,185              2,135
     8             105                 7,267              3,706
     9             302                15,006              7,653
    10           1,635                56,872             29,004

Table 2

Average Dollar Value of Shares Traded in Firms in Each ASX Size
Decile Over Five Trading Days

Decile 1 comprises ten percent smallest firms, by market
capitalization. The values were estimated by taking, for each stock in
each size decile, the average dollar volume of trades on the last
trading day of each of eight months in 2001 (from the monthly
sppr-price file) and multiplying by 5/8. The values for the final
trading days of March, June, September and December were not used
because trading volumes on these days is expected to be higher than a
typical trading day because of portfolio rebalancing by fund managers
at the end of each quarter.

Decile      $ Trades      % of Total Trades

  1           38,527            0.05%
  2           37,096            0.05%
  3           89,041            0.13%
  4          119,612            0.17%
  5          206,244            0.29%
  6          266,687            0.38%
  7          720,989            1.03%
  8          917,487            1.31%
  9        2,886,463            4.11%
  10      64,919,796           92.48%

Table 3

Coincidence of Recommendations

Number of times each newsletter recommended a stock within 30 days of
another newsletter, within 90 days (but not within 30 days), and
within 180 days (but not within 90 days). The numbers in bold along the
diagonal are the number of identifiable, clear buy recommendations made
by each newsletter.

              Rivken    GTS       Intelligent    Huntley's Your
              Report    Report    Investor       Money Weekly HUN

Rivken Report
  30 days       43
  90 days
  180 days
GTS Report
  30 days       1        177
  90 days
  180 days
Intelligent Investor
  30 days       2         1           149
  90 days       4         4
  180 days      2
Huntley's Your Money Weekly
  30 days       7         8            24               224
  90 days       3         7            14
  180 days      7         2             9

Table 4

Average Returns to Buy Recommendations, in Aggregate (full
sample) and by Newsletter

Portfolio returns are calculated as the simple average of the returns
to buy recommendations held until a sell recommendation was issued or
12-months from date of recommendation, whichever came sooner. All
recommendations were made in the issues of the newsletters made in
2001. 'Exp. Return' is the return to the experimental portfolio; 'n' is
the number of trades in the portfolio; 'Median control' is the median
of the 1000 control portfolio returns; and 'Control higher' is the
number of control portfolios, out of 1000, with a higher return than
the experimental portfolio. 'S&P 300' is the average return to a
portfolio of trades in the S&P ASX 300 Accumulation Index, matched to
the experimental portfolio on calendar holding period. 'Proportion +ve'
is the proportion of trades in the experimental sample that delivered
positive raw returns. Average Term is the average holing period of the
trades in the portfolio. Annualised Return is calculated by compounding
the Portfolio Return by (365/Average Term).

                       Full                        Int.
                      Sample   Rivkin    GTS     Investor   Huntley's

Exp. Return            12.3%    10.2%    17.2%     7.9%      11.3%
n                     593       43      177      149        224
Median Control          6.1%     2.9%     4.9%     6.3%       7.2%
Control Higher          0       81        1      339         75
S&P 300                 1.5%     2.1%     0.0%     2.1%       2.2%
Proportion +ve         61%      77%      56%      57%        63%
Average Term (days)   310      170      286      326        346
Annualised Return      14.7%    23.2%    22.5%     8.9%      12.0%

Table 5

Median Returns to Buy Recommendations, in Aggregate (full
sample) and by Newsletter

Panel A, portfolio reports the simple average of the returns to the
trades in the portfolio. Panel B reports the median return to the
trades in the portfolio. Exp. Return is the return to the experimental
portfolio; n is the number of trades in the portfolio; Median control
is the median of the 1,000 control portfolio median returns; and
Control higher is the number of control portfolios, out of 1000, with
a higher return than the experimental portfolio.

                 Full Sample   Rivkin     GTS    Int Inv   Huntley's

                  Panel A: Portfolio Return: Average

Exp. Return         12.3%      10.2%     17.2%      7.9%       11.3%
n                  593         43       177       149         224
Median Control       6.1%       2.9%      4.9%      6.3%        7.2%
Control Higher       0         81         1       339          75

                  Panel B: Portfolio Return: Median

Exp. Return          7.8%       5.3%      7.8%      5.7%        9.9%
n                  593         43       177       149         224
Median Control       4.1%       2.8%      0.4%      5.5%        6.5%
Control Higher       0        131         0       492          18

Table 6

Performance of Buy Recommendations Only

In panel A, the sample of recommendations includes BUY and ACCUMULATE
recommendations. In panel B, the sample of recommendations includes
only BUY recommendations. Portfolio returns are calculated as the
simple average of the returns to the trades in the portfolio. Exp.
Return is the return to the experimental portfolio; n is the number
of trades in the portfolio; Median control is the median of the 1000
control portfolio returns; and Control higher is the number of
control portfolios, out of 1000, with a higher return than the
experimental portfolio.

                 Full Sample   Rivkin    GTS     Int Inv   Huntley's

         Panel A: All Recommendations (BUY and ACCUMULATE)

Exp. Return         12.3%      10.2%    17.2%     7.9%       11.3%
n                  593         43      177      149         224
Median Control       6.1%       2.9%     4.9%     6.3%        7.2%
Control Higher       0         81        1      339          75

                 Panel B: BUY Recommendations Only

Exp. Return         10.4%      10.2%    16.2%     5.4%        4.6%
n                  360         43      142       97          78
Median Control       5.1%       2.7%     4.8%     5.3%        5.5%
Control Higher      27        108       11      486         570

Table 7

Performance of Recommendations on Liquid Stocks

Panel A reports the results of the full sample of recommendations.
Panel B reports returns to portfolios of recommendations that satisfy
the minimum liquidity requirement of $100,000 volume over the five
trading days following the recommendation. Panel C includes the returns
to portfolios of recommendations on stocks that had at least $500,000
volume over the five trading days after the recommendation. Portfolio
returns are the simple average of the returns to recommendations. Exp.
Return is the return to the experimental portfolio; n is the number of
recommendations; Median control is the return to the median portfolio;
and Control higher is the number of control portfolios, out of 1000,
with a higher return than the experimental portfolio.

                     Full
                    Sample   Rivkin     GTS    Int Inv   Huntley's

                          Panel A. All Recommendations

Exp. Return          12.3%    10.2%    17.2%     7.9%       11.3%
n                   593       43      177      149         224
Median Control        6.1%     2.9%     4.9%     6.3%        7.2%
Control Higher        0       81        1      339          75

               Panel B. Excluding 'Very Illiquid' Recommendations

Exp. Return          13.1%    10.2%    18.4%    10.4%       11.5%
n                   542       43      139      140         220
Median Control        6.1%     3.1%     4.6%     6.4%        6.9%
Control Higher        0       97        2      129          64

     Panel C. Excluding 'Very Illiquid' and 'Illiquid' Recommendations

Exp. Return          10.1%    10.7%     9.2%    10.0%       10.3%
n                   419       41       64      119         195
Median Control        6.0%     2.2%     4.2%     6.3%        6.7%
Control Higher       24       52      190      172         107

Table 8

Performance of Recommendations on Large, Medium and Small
Stocks

The table shows the performance of the aggregate newsletter and each
individual newsletter. Panel A presents the results of the full sample
of recommendations. Panel B presents the results of portfolios of
recommendations on stocks ranked in the top 150 on the ASX by market
value of equity. Panel C presents the results of portfolios of
recommendations on stocks ranked between 151 and 300 on the ASX by
market value of equity. Panel D presents the results of portfolios of
recommendations on stocks ranked below 300 on the ASX by market value
of equity. Portfolio returns are calculated as the simple average of
the returns to the trades in the portfolio. Exp. Return is the return
to the experimental portfolio; n is the number of trades in the
portfolio; Median control is the median of the 1000 control portfolio
returns; and Control higher is the number of control portfolios, out of
1000, with a higher return than the experimental portfolio.

                   Full
                  Sample   Rivkin    GTS     Int Inv   Huntley's

                         Panel A: All Recommendations

Exp. Return        12.3%    10.2%    17.2%     7.9%       11.3%
n                 593       43      177      149         224
Median Control      6.1%     2.9%     4.9%     6.3%        7.2%
Control Higher      0       81        1      339          75

                  Panel B: Recommendations on Large Stocks

Exp. Return         7.4%     5.8%    26.6%     7.3%        6.9%
n                 211       21        5       70         115
Median Control      7.7%     3.9%    10.1%     8.1%        8.3%
Control Higher    566      324      145      584         638

                 Panel C: Recommendations on Medium Stocks

Exp. Return        11.6%    12.9%     4.8%    17.3%       11.2%
n                 161        8       35       45          73
Median Control      2.8%    -1.8%     1.6%     2.5%        3.8%
Control Higher     14      106      334       25         114

                        Panel D: Recommendations on Small Stocks

Exp. Return        17.1%    15.2%    20.0%    -3.4%       25.7%
n                 221       14      137       34          36
Median Control      6.6%     2.8%     5.5%     6.1%        9.8%
Control Higher      3      187        2      854          46

Table 9

Performance of Momentum and Contrarian Recommendations

The table shows the performance of the aggregate newsletter and each
individual newsletter. Panel A presents the results of the full sample
of recommendations. Panel B presents the results of portfolios of
contrarian recommendations. Panel C presents the results of portfolios
of momentum recommendations. A contrarian (momentum) recommendation is
one where the recommended stock underperformed (outperformed) its size
decile by more than 5 percentage points over the six months prior to
the recommendation. Portfolio returns are calculated as the simple
average of the returns to the trades in the portfolio. Exp. Return is
the return to the experimental portfolio; n is the number of trades in
the portfolio; Median control is the median of the 1000 control
portfolio returns; and Control higher is the number of control
portfolios, out of 1000, with a higher return than the experimental
portfolio.

                     Full
                    Sample   Rivkin    GTS     Inv Int   Huntley's

                         Panel A: All Recommendations

Exp. Return          12.3%    10.2%    17.2%      7.9%     11.3%
n                   593       43      177       149       224
Median Control        6.1%     2.9%     4.9%      6.3%      7.2%
Control Higher        0       81        1       339        75

                         Panel B: Contrarian Recommendations

Exp. Return           0.3%     7.7%    -1.9%      1.3%     -1.0%
n                   224       20       36        61       107
Median Control        5.8%     4.5%     2.3%      5.9%      6.5%
Control Higher      984      308      685       791       982

                         Panel C: Momentum Recommendations

Exp. Return          22.1%    19.1%    23.4%     13.6%     28.0%
n                   267       14       125       60        68
Median Control        5.6%     1.5%     4.8%      6.5%      6.6%
Control Higher        0       57        0       117         0

References

Atiase, R. 1985, 'Predisclosure information, firm capitalization and security price behavior around earnings announcements,' Journal of Accounting Research, vol. 23, no. 1, pp. 21-37.

Banz, R. 1981, 'The relationship between return and market value of common stocks', Journal of Financial Economics, vol. 9, no. 1, pp. 3-18.

Beedles, W. H. 1986, 'Asymmetry in Australian equity returns', Australian Journal of Management, vol. 11, pp. 1-12.

Brown, P. & da Silva Rosa, R. 1998, 'Research method and the long-run performance of acquiring firms', Australian Journal of Management, vol. 23, no. 1, pp. 23-39.

Demir, I. Muthuswamy, J. & Walter, T. 2003, 'Momentum returns in Australian equities: The influences of size, risk, liquidity and return computation', Working Paper, School of Banking and Finance, University of New South Wales.

Dobbie, M. 2003, 'Finding the straight tip', Shares Magazine, November 2003.

Durack, N., Durand, R.B. & Maller, R. 2004, 'A Best Choice Among Asset Pricing Models? The Conditional CAPM in Australia', Accounting and Finance, vol. 44, pp. 139-62.

Faff, R. 2001, 'An examination of the Fama and French three-factor model using commercially available factors', Australian Journal of Management, vol. 26, no. 1, pp. 1-16.

Halliwell, J., Heaney, R. & Sawicki, J. 1999, 'Size and book to market effects in Australian share markets: A time series analysis', Accounting Research Journal, vol. 12, no. 2, pp. 122-137.

Huntley, I. 2001, Huntley's Sharemarket Education Course Module 1, Huntley's Investment Information Pty Ltd, Sydney 

InfoChoice, 2003, Research Publications, (Online 6th Aug 2003), Available from: <http://www.infochoice.com.au/investment/onlinebroking/compare/research /default.asp>.

Jaffe, J. & Mahoney, J. 1999, 'The performance of investment newsletters', Journal of Financial Economics, vol. 53, no. 2, pp. 289-307.

Jegadeesh, N. & Titman, S. 1993, 'Returns to buying winners and selling losers: Implications for stock market efficiency', The Journal of Finance, vol. 48, no. 1, pp. 65-92.

Jegadeesh, N. & Titman, S. 2001, 'Profitability of momentum strategies: An evaluation of alternative explanations', The Journal of Finance, vol. 56, no. 2, pp. 699-720.

Kothari, S. & Warner, J. 1997, 'Measuring long-horizon security price performance', Journal of Financial Economics, vol. 43, no. 3, pp. 301-39.

Kothari, S. & Warner, J. 2001, 'Evaluating mutual fund performance', The Journal of Finance, vol. 56, no. 5, pp. 1985-2010.

Main, A. 2003, Australian Financial Review, 7th May 2003, p. 32.

McDonald, A. 2003, Newsletters Snap up a Retail Niche, The Sydney Morning Herald, 21 June, <http://www.smh.com.au/articles/2003/06/20/1055828488814.html> (Online 6th Aug 2003).

Metrick, A. 1999, 'Performance evaluation with transactions data: The stock selection of investment newsletters', The Journal of Finance, vol. 54, no. 5, pp. 1743-75.

Rivkin, R. 2001, The User's Guide to the Rivkin Report, Tarfaya Nominee Pty Ltd, Sydney.