Original Historical Analysis – Update: Out-of-Sample Test
ORIGINAL HISTORICAL ANALYSIS
In this entry, we calibrate Herb Greenberg’s stock commentary on MarketWatch. In general, Herb Greenberg provides negative/skeptical comments on specific stocks. His sources are often anonymous and may already have shorted stocks he mentions. In his own words, his “…column does not make recommendations; it points out risks or situations that are otherwise overlooked. It’s then up to readers to decide whether to sell the stock (if they own it), avoid it (if they’re considering owning it), ignore what I’ve written or, if they’re so inclined, make a negative bet on a company.” For this analysis, we catalog 318 distinct negative mentions involving 99 different stocks over the period 4/12/04 (when he joined MarketWatch) through 6/15/05. Here’s what we find…
The chart below shows the aggregate performance of the stocks negatively mentioned by Herb Greenberg over seven intervals:
- From 11 days before publication to one day before publication;
- From 6 days before publication to one day before publication;
- From 1 day before publication to one day after publication;
- From the day of publication to two days after publication;
- From the day of publication to five days after publication;
- From the day of publication to 21 days after publication; and,
- From the day of publication to 63 days after publication;
If he mentions a stock more than once, we include its performance more than once, sometimes with overlapping intervals. For all stocks and all holding intervals, we use daily closing prices for entry and exit (so the “day of publication” means after the story has initial impact). We do not include dividends that may be received (paid out) while stocks are held (shorted). Nor do we include transaction costs.We exclude one outlier stock ( the one-time mention of IPIX on 4/13/04) that had a gain of 835% in a very short period and therefore would distort results. For comparison, the chart also shows the average performance of the S&P 500 index over the same length intervals during 4/12/04 through 6/15/05.

The following table provides additional detail:

The chart and table show that:
- On average, targeted stocks underperform the S&P 500 and produce negative returns over all intervals examined.
- Average underperformance is most dramatic before and well after publication.
- The standard deviations in the price behavior of targeted stocks over all intervals are extremely high compared to corresponding deviations for the S&P 500. This high variability means that investing in these stocks is very risky, at least around times when Mr. Greenberg writes about them.
The pre-publication price behavior of targeted stocks suggests the possibility of frontrunning of Herb Greenberg’s public articles. Relying on public sources for stock-selection information risks frontrunning (prior buying or selling) by those who had the information before it became public. News sources may act on their private information before tipping the media. Before making public pronouncements about specific stocks, analysts give their clients opportunities to buy or sell. Mutual fund managers buy before making their stock selections known. An alternative explanation here is that Mr. Greenberg and his sources are reacting to other news that has already affected stock price.
In his 10/20/04 column, Herb Greenberg states: “Short-sellers are often sources for reporters, which itself is often the topic of controversy. Not that it should be, because to think reporters are running stock screens and randomly poring over SEC filings to spot potential “needle-in-a-haystack” stories is ludicrous. More often than not, stories are sparked by news or tips, and the tips come in all shapes and sizes: from employees, investors, brokers, brokerage analysts and short-sellers. I make no excuses for talking to short-sellers. What, reporters should only talk to longs? The idea is to get ahead of the news, and short-sellers are among the best sources because their research is often the most impeccable, which is required if you’re going against the grain.” Those ahead of Mr. Greenberg’s news may be the short-selling tipsters. In fact, the very act of their short-selling may be the news.
In summary, Herb Greenberg successfully identifies stocks that, on average, underperform the market. However, the variability of the performance of these stocks makes trading on “Greenberg mentions” extremely risky. Also, the average price behavior of these stocks suggests the possibiliity of frontrunning by his tipsters.
Note that the above analysis addresses only Herb Greenberg’s public column, not his paid “RealityCheck” offering, which could be a stimulus for frontrunning of public articles. Also, Mr. Greenberg occasionally mentions a stock positively, but not often enough to create a meaningful sample for performance testing.
UPDATE: OUT-OF-SAMPLE TEST
In the above analyis we calibrated Herb Greenberg’s stock commentary on MarketWatch over the period 4/12/04 (when he joined MarketWatch) through 6/15/05. We concluded that Mr. Greenberg successfully identifies stocks that, on average, underperform the market. However, the variability of the performance of these stocks makes trading on “Greenberg mentions” extremely risky. Also, the aggregate price behavior of these stocks suggests frontrunning by his tipsters. In this entry, we look at Mr. Greenberg’s commentary from 6/15/05 through 1/27/06 to see whether these conclusions still hold. We find that:
The chart below compares the aggregate performance of two “red-flag portfolios” constructed from Herb Greenberg’s MarketWatch columns appearing during 4/12/04-6/15/05 and during 6/16/05-1/27/06. The latter portfolio involves 33 different stocks mentioned negatively a total of 63 times. The chart shows returns for these two portfolios over seven intervals:
- From 11 days before publication to one day before publication;
- From 6 days before publication to one day before publication;
- From 1 day before publication to one day after publication;
- From the day of publication to two days after publication;
- From the day of publication to five days after publication;
- From the day of publication to 21 days after publication; and,
- From the day of publication to 63 days after publication;
In constructing these portfolios, if Mr. Greenberg mentions a stock in more than one column, we include its performance more than once, sometimes with overlapping intervals. In other words, a trader seeking to capture these abnormal returns would have to increase the (short) position in a stock each time Mr. Greenberg mentions it negatively in a new column. For all stocks and all holding intervals, we use daily closing prices for entry and exit, so the “day of publication” usually captures the initial impact of a column. We do not include dividends that may be paid out while stocks are shorted. Nor do we include trading fees or stock borrowing costs. For comparison, the chart also shows the average performance of the S&P 500 index during 4/12/04 through 6/15/05. The average performance of the S&P 500 index during 6/16/05 through 1/27/06 is similar.

The chart shows that both red-flag portfolios, on average, underperform the S&P 500 index from just before publication to well after publication. The newer sample shows slightly greater underperformance immediately after publication but much less underperformance over the intermediate term. The pre-publication price behavior of targeted stocks in the older sample suggests frontrunning of Herb Greenberg’s public articles. However, most of this effect has reversed for the newer sample, with targeted stocks outperforming on average during the two weeks before appearance of negative articles about them. A possible explanation is that tipsters are now using the information channel presented by Mr. Greenberg more for containing prices on existing short positions than for establishing new short positions. Over half the mentions in the newer sample involve stocks also appearing in the older sample. Also, the newer sample is fairly small, so confidence in its statistics is relatively low.
The standard deviations of returns for the red-flag portfolios over all intervals are extremely large. The next chart illustrates this variability, showing the average returns for each portfolio with error bars of one standard deviation above and below the average. This high variability means that investing in these stocks is very risky, at least around times when Mr. Greenberg writes about them.

In summary, Herb Greenberg successfully identify stocks that, on average, underperform the market. However, the high variability of returns for these stocks continues to make trading on “Greenberg mentions” extremely risky.
Note that the above analysis addresses only Herb Greenberg’s public column, not his paid “RealityCheck” offering that existed during most of the period considered above. Also, Mr. Greenberg occasionally mentions a stock positively, but not often enough to create a meaningful sample for performance testing.
MarketWatch has (discontinued Mr. Greenberg’s paid “RealityCheck” offering) as of mid-February 2006. See also Mr. Greenberg’s recent denial (in connection with his receipt of a subpoena) that he intentionally participates in front-running by his sources.
For reviews of the outputs of other market experts, browse Guru Grades.
Frequently Asked Questions
What is a red-flag portfolio of negatively mentioned stocks?
A red-flag portfolio is a group of stocks that a commentator has criticized or flagged as risky, tracked to see how they perform around publication. In the Herb Greenberg analysis, portfolios were built from his negative MarketWatch mentions: 318 mentions of 99 stocks (4/12/04-6/15/05) and 63 mentions of 33 stocks (6/16/05-1/27/06). A stock mentioned in multiple columns was counted each time.
Do stocks criticized by financial columnists like Herb Greenberg underperform the market?
Stocks negatively mentioned by Herb Greenberg underperformed the S&P 500 on average in both samples, from just before publication to well after it. The original sample (4/12/04-6/15/05) showed the most dramatic underperformance before and well after publication, while the newer sample (6/16/05-1/27/06) showed slightly greater underperformance right after publication but much less over the intermediate term. The returns were highly variable, so trading on the mentions was judged extremely risky.
How do you test whether a stock commentator's negative picks beat the market?
The analysis measured the aggregate performance of the mentioned stocks over seven intervals, from 11 days before publication to 63 days after, using daily closing prices for entry and exit, and compared them with the S&P 500 over the same interval lengths. Repeat mentions of a stock were counted again, sometimes with overlapping intervals, and one outlier (IPIX, a gain of 835%) was excluded. Dividends, transaction costs, and stock borrowing costs were not included.
Does price weakness before a negative stock article mean frontrunning or is it just reaction to earlier news?
Pre-publication price drops in Herb Greenberg's original sample suggested possible frontrunning by tipsters such as short-sellers who knew the information before it became public. An alternative explanation is that he and his sources were reacting to other news that had already moved the stock price. In the newer sample, most of the pre-publication effect reversed, with targeted stocks outperforming on average in the two weeks before negative articles.
What are the risks of trading on negative stock mentions from a financial columnist?
Trading on Herb Greenberg's negative mentions was judged extremely risky because the standard deviations of the targeted stocks' returns were very high compared with the S&P 500's. Relying on public sources also risks frontrunning by those who had the information earlier. The analysis also excluded dividends, transaction costs, and borrowing costs, covered only his public column rather than his paid RealityCheck service, and had too few positive mentions to test.