Dreman compares research analysts to dealers at the casino. The players ask the dealers, who seem to know the game well, what to play; unfortunately, the players are still destined to lose.
Investors rely on these "dealers" for their estimates in formulating their own investment decisions. However, in a paper Dreman himself authored, Dreman showed that analysts are off in their EPS estimates by about 40% per quarter. The dramatic difference is there even after removing companies with low earnings (to avoid large percentage effects just because earnings were small on an absolute basis).
Furthermore, the effect is prevalent even across industries, from cyclical to non-cyclical alike. The trend is also getting worse, with analysts in the most recent decades actually performing worse, despite their seeming informational advantages.
Analysts also tend to be overly optimistic. EPS growth from 1982 to 1996 was about 8% per year, but analyst forecasts taken at the beginning of each year suggested predictions over this period were for 21% per year.
Other studies have confirmed these findings, noting other interesting tidbits in the process. One study found that analyst estimates would be more accurate if they simply blindly assumed a 4% rise in earnings every year for every company.
Why might this be? For one thing, analysts are overconfident in their own research. Many expect their findings to be accurate within 5%, but they are not. Despite the research suggesting analysts are not accurate, each individual analyst seems to believe he is better at predicting than he really is. Furthermore, not a lot of emphasis is placed on being accurate. Analyst pay/bonus structures are often based on how much trading brokerage the analyst brings in. This helps explain why there are so many more buys than sells. (By comparison, few brokerage commissions are brought in for “sell” recommendations.) So, it’s not about being right; instead, it’s about bringing in clients.
This helps explains why one study showed that analysts that work at firms that also have investment banks issue 25% more “buy” recommendations and 46% fewer sell recommendations than their counterparts at firms without investment banks. (Investment banking clients have been known to shun firms that make negative recommendations about their stock.)
Dreman also discusses anecdotal evidence that further suggests analysts are not paid for their accuracy but for their clients. In one example, Donald Trump once requested that an analyst be fired after he issued a sell recommendation on Trump’s company. The analyst was fired, and won a few years’ worth of salary in court as a result, shortly after Trump’s company filed for bankruptcy.