Investars 2005 : Methodology

Intended Usage
Investars allows investors to investigate the performance record of the research provider offering them investment advice on a stock, for example IBM. It also shows them what other firms are making recommendations on IBM and what the performance of their advice has been. By ordering the performance records of those firms from best to worst, Investars assists the investor in identifying which research providers can offer them the best advice on whether to buy, hold or sell IBM at this moment in time.

The Performance Measurement's System's Objective
Investars goal is to produce an accurate measurement of the performance of research providers. To do this it measures the success or failure of each of recommended buy, hold and sell recommendations in the providers stock coverage universe.

Background
Research organizations that analyze common stocks may provide several types of information to their institutional and retail clients. These items can include a commentary on a company's current financial progress, a review of its most recent earnings announcement and an estimate of a company's future earnings. The one single item that summarizes all of this information is a recommendation to buy, hold or sell a stock. Investars seeks to measure the success of these recommendations and then compare providers by their performance.

Methods
Research providers maintain lists of buy, hold and sell recommendations on the stocks in their coverage universe. These recommendations are constantly updated as the analysts opinions change. Since a list is not the same as a recommended portfolio, methods to measure them must be different.

Investars uses a straightforward set of rules to compare lists of recommendations. It's proprietary approach rewards a buy recommendation when the stock rises and penalizes it when the stock falls. If it is a strong buy verses a buy, then the strong buy is awarded when it outperforms the buy. Sell recommendations are also valued for their accuracy. A sell recommendation is rewarded if the stock falls and penalized if the stock rises. A strong sell recommendation is rewarded if the stock falls more than a sell recommendation. Hold recommendations are by their nature neutral so get no rewards or penalties.

An individual stock's score reflects the success or failure of every buy, hold or sell recommendation made for the stock. These scores are called ROSS returns (Rate of Success System). Scores are calculated for each stock in the research provider's coverage universe. To measure a provider's performance effectiveness in an industry or an economic sector, the appropriate individual stock scores are simply averaged. When all of the stocks a provider covers are averaged, the result is an overall provider score reflecting that research firm's performance. This provider score can then be compared to the scores of all other providers to rank them in order of their past success in making stock recommendations.

ROSS (Rate of Success System) Methodology
The Investars "Rate of Success System" (ROSS) quantifies a firm's performance by creating hypothetical portfolios based on their stock ratings. The system calculates returns for every firm and ranks them based on their returns for every stock ("By Stock") and sector ("By Industry") as well as for their overall portfolio of ratings ("By Portfolio"). ROSS also takes into account the strength of the ratings in creating the hypothetical portfolios and all historical ratings for the specified time period. Ratings are updated throughout the day and the returns are calculated once a day after the market closes. ROSS uses historical and current stock ratings to calculate hypothetical historical investment returns for more than 130 firms and more than 7000 companies.

The ROSS System hypothetically purchases or sells shares in a stock on the day that the rating is issued. The amount of shares purchased depends on the strength of the rating. For example, if a firm's initial rating for the period is a buy on Cisco Systems (CSCO) then the system purchases 300,000 shares in CSCO. If the firm upgrades Cisco at a later date from a "buy" to a "strong buy" then the system increases the number of shares in the hypothetical portfolio by 50% on the day of the upgrade. Similarly, if the firm downgrades the stock then the system decreases the number of shares by 33.3%. If a firm reiterates a rating then the position in that stock is left unchanged. At the end of the calculated period the system closes the position by selling all shares for a long position or by buying shares for a sell position at the last available price for the stock.

The hypothetical portfolios that ROSS creates depends on the rating:
If a firm issues a bullish rating on a stock (outperform, buy or strong buy) then the system purchases shares in the hypothetical portfolio creating a positive position in the stock. The positive position depends on the strength of the rating. For example, the higher the rating, the more shares that are purchased in the hypothetical portfolio.

If a firm issues a neutral rating on a stock (perform) then the system creates no position in the stock.

If a firm issues a bearish rating on a stock (underperform, sell or strong sell) then the system sells shares in the hypothetical portfolio creating a negative position in the stock. The negative position depends on the strength of the rating. For example, the lower the rating, the more shares that are sold in the hypothetical portfolio.

Since firms use different terminology for different ratings, Investars has created a standard scale that can be applied to all ratings for any firm. Shares are purchased or sold according to the following scale:

Investars Equivalent Rating	Shares Purchased (Sold) based on Initial Rating
Strong Buy	450,000
Buy	300,000
Outperform	200,000
Perform	0
Underperform	(200,000)
Sell	(300,000)
Strong Sell	(450,000)

ROSS calculates the "By Portfolio" and "By Industry" returns by using the weighted average return for all trade pairs for all stocks in the firm's portfolio and for all stocks covered in the industry.

Penny stocks are only included in the rankings if they have historically or currently trade above $1.00. The returns for stocks that fall below $1.00 are calculated based on a minimum price of $1.00. Likewise, the returns for penny stocks are not calculated until they go higher than $1.00.

Investars allows investors to evaluate performance using either the ROSS Synthetic Return or the ROSS Standard Return.

ROSS Synthetic Return
In our default algorithm (ROSS Synthetic Return) we consider a bearish rating (underperform, sell, strong sell) as the inverse of a bullish rating (outperform, buy, strong buy) to eliminate any mathematical bias towards long positions in a bullish rating and reflect an analyst's timing. For example, if a firm issues a sell at $10 and the stock falls to $2, the firm would have the same return as if they had issued a buy at $2 and the stock rose to $10.

ROSS Standard Return
Investors also have the option of evaluating performance based on our ROSS Standard Return which more closely mimics actual dollar trading returns. The Standard Return applies the standard method of calculating returns on short positions and limits the upside of a bearish rating to 100% just as the upside on a short position would be limited to 100%.

IMPORTANT: ALL RETURNS ON INVESTARS are designed for RELATIVE rankings of analysts only and SHOULD NOT be compared to actual portfolio or index returns. Investars DOES NOT RECOMMEND non-professional investors taking short positions based on bearish analyst ratings.

ROSS Algorithm Example:
In order to calculate returns the system coverts the sequences of the hypothetical trades into “Trade Pairs”. A Trade Pair essentially is a single buy/sell or sell/buy trade whereby shares in each Trade Pair have their own return. A Trade Pair is calculated based on the quantity of shares involved, the purchase price and the sell price. To create a Trade Pair the system takes two successive hypothetical trades of opposite sign. For example, let's say we had a "buy" rating for CSCO followed by an "outperform". The price at the time of the "buy" rating was $16.21 and the price at the time of the "outperform" was $20.02. The buy rating would tell the system to invest in 300 shares of CSCO, the "outperform" rating would reduce the position by 1/3 selling 100 shares of CSCO. That would produce a Trade Pair with 100 shares of quantity, $16.21 purchase price and $20.02 sell price. The Trade Pair would have a return of (sell price / purchase price – 1).

The pairs are created using FIFO ("first in first out") principle. That means that if there are several hypothetical positive trades followed by a negative trade – the system generates a pair using the first positive and the first negative trade. The number of shares included in a Trade Pair is the minimum of the absolute values of the trades involved. After forming a Trade Pair the trades are adjusted by the quantity of shares that went to build the Trade Pair.

1. Needham initiates coverage on CSCO with a "strong buy" on 9/24/2001 a hypothetical trade "buy 450 shares @ $12.56" is created

2. Needham downgrades CSCO from "strong buy" to "buy" on 10/15/2001 a hypothetical trade "sell 150 shares @ $16.21" is created

3. Needham downgrades CSCO from " buy" to "perform" on 11/16/2001 a hypothetical trade "sell 300 shares @ $20.02" is created

4. Needham upgrades CSCO from "perform" to " buy" on 4/10/2002 a hypothetical trade "buy 300 shares @ $15.07" is created

5. End of the calculated period (Aug 22, 2002) price = $15.10

The pairs created for this sequence of hypothetical trades are:
150 shares, buy price $12.56, sell price $16.21
300 shares, buy price $12.56, sell price $20.02
300 shares, buy price $15.07, sell price $15.10

If there are ratings preceding the currently period (e.g. we are calculating returns for the one year period 1/31/2002 – 1/31/2003) we create a substitution rating (carry-over) using the value of the last rating before the period, the price at the first date of the period and assign the first day of the period to the date of the rating.

Each trade pair has its own return. The trade pair returns are combined to calculate the aggregate returns like Stock Return, Portfolio Return and Industry Return.

In our default algorithm (ROSS Synthetic Return) we consider a bearish rating (underperform, sell, strong sell) as the inverse of a bullish rating (outperform, buy, strong buy) to eliminate any mathematical bias towards long positions in a bullish rating and reflect an analyst's timing. For example, if a firm issues a sell at $10 and the stock falls to $2, the firm would have the same return as if they had issued a buy at $2 and the stock rose to $10.

ROSS Synthetic Return (long trade pair) = sell price / buy price – 1. A short trade pair return is calculated as the inverse of a long trade pair to reflect an analyst's market timing: buy price / sell price - 1.

Investors also have the option of evaluating performance based on our ROSS Standard Return which more closely mimics actual dollar trading returns. The Standard Return applies the standard method of calculating returns on short positions and limits the upside of a bearish rating to 100% just as the upside on a short position would be limited to 100%.

To calculate the standard return of a short trade pair we take the price at which the stock was short sold as the investment price. The difference between the prices when the stock was short sold and when the stock was purchased is the money earned in the trade pair.

ROSS Standard Return (short trade pair) = short sell price – buy price / short sell price = 1 – buy price / short sell price.

The two types of the trade pair returns above are aggregated to calculate, stock, industry and portfolio returns. These returns are calculated by combining the corresponding returns of all trade pairs of the one stock of a research firm, returns of all trade pairs of all stocks in the firm's portfolio and of all stocks covered by the firm in the industry.There are five types of trade pair returns aggregations currently calculated by the system: Average Return, Cumulative Return, Standard Cumulative Return, Long Positions Return and Short Positions Return.

1. ROSS Synthetic Return is a weighted average of the Trade Pairs synthetic returns. Each Trade Pair’s return is weighted by the number of shares involved in the Trade Pair. Let's say that Rn is the return of the trade pair #n, Nn is the number of shares in this trade pair. Then the Average return would be:

2. ROSS Standard Return is calculated just like the ROSS Synthetic Return, but instead of the synthetic trade pair returns we use the standard ones.

3. Long Positions Return is calculated just like the Standard Average return, but using positive (long) Trade Pairs only.

4. Short Positions Return is calculated just like the Standard Average return, but using negative (short) Trade Pairs only.

Example:

Trade pairs formed with the ratings of Merrill Lynch for EOTPQ for the 3 year period.