Centre for Economic Policy Research

Investment Banking: Conflicts of Interest in Underwriting and Research 23
Michaely and Womack (1999) examined ‘buy’ recommendations of lead under-
writer and other analysts after the SEC’s 25 day post-IPO ‘quiet period’ for 391
IPOs in 1990 and 1991. They found that in the month after the quiet period, lead
underwriters’ analysts made 50% more buy recommendations than other firms’
analysts for the same securities, suggesting some conflict of interest. One striking
feature was that stock prices of firms recommended by lead underwriting banks
declined during the quiet period, while other banks’ picks rose. The market
appears to recognize this difference in the quality of information, and the excess
return at the recommendation date is 2.7% for underwriters’ analysts and 4.4% for
other analysts. Considering a two year holding period from the IPO date, the 
performance of other analysts’ recommended issues was 50% better than the 
performance of underwriters’ recommendations. Finally, the same investment
banks made better recommendations on IPOs when they were not the lead
underwriter, implying that it was not a difference in analysts’ ability but an under-
writer bias.
While Michaely and Womack’s findings are consistent with the presence of
conflicts of interest, one cannot rule out two alternative explanations. It is 
possible that underwriters’ analysts exhibit cognitive bias (Kahneman and
Lovallo, 1993), where they have very strong prior beliefs that the firms they
underwrite are better and additional research will not alter this view. Non-
underwriters’ analysts do not have strong prior beliefs and allow their judgement
to evolve. There may also be some selection bias if underwriters are chosen by
issuers because they hold favourable views of the firm and interpret new 
information differently than other analysts.
14
These alternative interpretations
may explain some of the considerable heterogeneity in the industry. One 
interpretation is that conflicts of interest may have dominated Merrill Lynch,
Salomon Smith Barney and Credit Suisse First Boston’s research departments,
while cognitive bias may have been the leading force at Morgan Stanley.
It is popularly believed that investors’ attention to earnings performance
increased in recent years. This increased sensitivity of the market to earnings 
forecasts seems to have influenced management behaviour with companies under
heavy pressure to prevent earnings from falling short of targets – including 
analysts’ forecasts. Managers have some discretion in reporting the timing and
magnitude of revenue and expenses and can manipulate earnings through 
accruals and other devices (Chan et al., 2003). Missing earnings targets is 
regarded as extremely bad news. Managers may thus have an incentive to ensure
that analysts keep down their forecasts – permitting analysts to exceed their fore-
casts and thereby gain a boost to the firms’ stock prices. There is some empirical
evidence for this increased focus of the markets on companies’ earnings. Francis
et al. (2002) and Landsman and Maydew (2002) find that the magnitude of
abnormal returns and abnormal volume increased around earnings announce-
ments from the 1980s to the 1990s. In addition, the Francis et al. (2002a) study
also produced evidence that market reaction to analysts’ earnings announcements
has increased. 
Again, anecdotal evidence in the popular press implies that analysts manipu-
late their forecasts. These beliefs find some support in recent academic research.
Chan et al. (2003) examined whether analysts bias their opinions in favour of a
company by adjusting earnings estimates to help managers match or exceed
expectations. They found that for the period 1984-2001 there was a pronounced
shift in the cross-sectional distribution of earnings surprises for the United States.
The share of non-negative surprises rose from 49% in the late 1980s to 76% in
1999-2001. Furthermore, there is evidence that the higher incidence of non-
negative surprises arose from analysts’ strategic adjustments. When earnings fell

---

24 Conflicts of Interest in the Financial Services Industry
short of the consensus three months before the announcements, analysts revised
the estimates downward by enough to yield a non-negative surprise upon
announcement. This pattern was more pronounced for growth firms compared to
value firms, which Chan et al. (2003) attribute to analysts’ disposition to deliver a
positive surprise for firms with relatively high valuations. In addition, more firms
initially meet or surpass expectations for consecutive quarters, than would be
expected statistically, suggesting manipulation. In the late 1990s, growth firms
with four consecutive quarters of non-negative surprises occurred 35% more often
than predicted. Non-negative surprises became increasingly predictable based on
whether it was a growth or value firm and the sign on past surprises. 
It should be noted, however, that managers have significant incentives to
ensure they limit the ‘surprise’ on an earnings announcement especially if it is
negative. The price effects of a negative surprise are positively correlated with the
price-earnings ratio, and in the late 1990s these multiples were very high by any
standard. Any shock was met with a sharp price reaction and some class action
suits against the management. The result was that managers started informing the
market giving guidance through various sources including favoured analysts. 
Compared to the United States, foreign markets typically do not show this
increased disposition to positive earnings surprises. Overall, they display a 
median negative surprise and a stable distribution of surprises over time. While
European markets certainly experienced a boom in technology and telecommu-
nication industries, only in the United Kingdom was there a pattern similar to the
US one. The proportion of positive surprises rose from 45% in the late 1980s to
59% in 1998-99. For the European Continent, over half of the surprises were 
negative. The relatively depressed Japanese market of the 1990s did not have a
wave of IPO activity and there is no trend in earnings surprises. Chan et al. (2003)
argue that the incentives for firms and managers to control earnings surprises are
weaker in foreign equity markets. Only in the United Kingdom, which shares 
similar investment bank features with the United States, does it appear that 
analysts were managing earnings surprises. Elsewhere, conflicts of interest due to
investment banking business were seemingly less severe because IPO activity was
lower and competition among investment banks weaker, while compensation for
analysts was lower.