nently display the prospective earnings per share and, more pertinently, the prospective price earnings ratio at the front of the prospectus. The prospective price
earnings multiple is the issue price divided by forecast earnings per share. The news media and stockbrokers’ research reports focus on the prospective price earnings
ratio. They compare the ratio with the ratios of already listed companies that are similar in industry, size, or other dimension, and then estimate a likely market price
for the IPO. Based on the estimated market price which is a function of the forecast profit, analysts and investors decide whether to subscribe to the new issue
or which IPO to subscribe to if there are several new issues occuring at the same time. Firth 1997, 1998 demonstrated empirically that investors rely on profit
forecasts in pricing shares on the first day of trading and that deliberate underpric- ing of IPOs can be signalled via profit forecasts.
The IPO process is similar to the practices in Australia, Britain, Malaysia, New Zealand, and Singapore, but contrasts with the US. In the U.S. the precise offer
price is not fixed in the prospectus but is instead set a day or so before share allocation and after the underwriter has measured the likely demand. American
IPO shares are often sold to specific clients of the underwriter and to specialist investors rather than the general public. The company, the underwriter, and other
sponsors of the issue can communicate privately to these select investors and so the prospectus is of relatively less importance than in Hong Kong, where the prospec-
tus is the only communication between the company and investors. American IPO shares are sometimes initially listed on an OTC market whereas in Hong Kong the
shares are listed on the full exchange board.
4. Methodology
4
.
1
. Models The forecast error for company i for the year of the IPO t is calculated as:
FE
it
= AP
it
− FP
it
FP
it
, 1
where FE is profit forecast error for the company; AP, actual profit for the company; FP, profit forecast as given in the IPO prospectus.
The mean forecast error MFE measures the bias in forecasts. A positive value for MFE implies that on average IPO companies have a pessimistic bias firms
under-forecast while a negative value for MFE represents an optimistic bias firms over-forecast.
Taking the absolute value of the forecast errors FEs gives the absolute forecast error AFE for each IPO. AFE is the major metric used to evaluate forecast
accuracy. The mean of the absolute forecast errors, denoted as MAFE, represents the overall accuracy of IPO profit forecasts. AFEs vary quite significantly across
companies. One reason for these differences will be the inherent difficulty in predicting a specific company’s earnings; this inherent difficulty is not, however,
directly measurable. One proxy for inherent difficulty is the change in annual profits
measured from before the IPO to after it. We argue that the greater the change in profit, the more difficult it will be to forecast the profit.
6
Brown et al. 1987 developed a statistic that measures the superiority of financial analysts in forecasting profits relative to the actual change in profits. Their measure
is adapted here, for the IPO market. The statistic is: SUP = ln[AP
t
− AP
t − 1
AP
t
− FP
t 2
]. 2
ln is the natural logarithm operator. The denominator measures the error in the IPO forecast while the numerator is the change in profit from year t − 1 to year t.
The numerator can also be regarded as the forecast error from a simple time series forecasting process where AP
t − 1
is a random walk model estimate of the profit in year t. By construction, a positive value for SUP means the IPO profit forecast is
more accurate than a forecast based on the random walk model; a negative value implies the IPO forecast is inferior to a random walk forecast. We hypothesize that
the mean of the SUP scores will be greater than zero.
Instead of using AP
t − 1
in Eq. 2, we can use growth in historical profits. This makes more use of information given in the prospectus. The growth model forecast
is given as: GFP = AP
t − 1
AP
t − 1
AP
t − 3 12
, 3
where GFP is growth model forecast profit; AP
t
, actual profit in year t t − 1, t − 3; year t − 1 is the profit for the last year before the IPO and year t − 3 is the
profit for the third year prior to the IPO. The forecast, GFP, takes the profit in AP
t − 1
and multiplies it by one plus the growth rate of profits over the previous 2 years. The modified version of the SUP
measure, named MSUP, is calculated thus: MSUP = ln[AP
t
− GFPAP
t
− FP
t 2
]. 4
We hypothesize that, on average, MSUP will be positive, indicating superior forecasting ability of the IPO management vis-a`-vis the growth model.
Absolute forecast errors AFEs, SUPs, and MSUPs vary across companies and we construct cross-sectional models to help explain the variations. We base our
models on a priori reasoning and on the methodologies and results from previous studies see Section 2. The model for AFE is:
7
AFE = b + b
1
SIZE + b
2
HORIZON + b
3
PROFVAR + b
4
AGE + b
5
LEV + b
6
AUDIT + b
7
UNW + b
8
OWN + b
9
RED + b
10
IND + b
11
ROA, 5
6
A large change in an IPO’s profit may signify a rapid change in the company’s size, strategies, and investment opportunities and-or changes in the business and economic conditions of the production
factors and product markets in which the IPO operates.
7
SUP and MSUP are also modelled as functions of the same independent variables. We illustrate AFE first because other studies modelled absolute forecast errors and we compare our results with those
from previous research.
where, SIZE is log of total assets after the IPO. Size is measured in millions of Hong Kong dollars. HORIZON, length of the forecast period. This is the number
of months between the prospectus date and the next fiscal year end. PROFVAR, standard deviation in profit growth during the 3 years immediately prior to the IPO
date. AGE, the number of years from the date of the company’s incorporation to the IPO date. LEV, total debttotal assets. AUDIT, a dummy variable taking the
value one 1 if the auditor is one of the Big Six firms; otherwise AUDIT is coded zero 0. UNW, a dummy variable taking the value one 1 if the underwriter is
Bear Stearns, Credit Lyonnais, Goldman Sachs, Merrill Lynch, Warburg, Schroders, Smith Barney, Nomura, Peregrine, Sun Hung Kai, HSBC, Jardine
Fleming, and Crosby; otherwise UNW is coded zero 0. Firms coded one 1 are the major underwriters operating in Hong Kong. The selection of these firms was
made after consultation with investment professionals in Hong Kong. OWN, proportion of shares sold in the new issue. RED, a dummy variable taking the
value one 1 if the IPO is a Red Chip or H-share; otherwise RED is coded zero 0. IND, a dummy variable taking the value one 1 if the company is a public utility,
or in the industries of transportation, banking, and finance; otherwise IND is coded zero 0. ROA, return on assets given by net profit divided by total assets.
Regression analysis is used to examine whether investors can at least partially anticipate the forecast errors at the time of listing. The initial returns underpricing
are hypothesized to be a positive function of the signed forecast errors Firth, 1998.
8
In order to test the hypothesis, the following model is constructed: RET = a + b
1
FE + b
2
SIZE + b
3
RED + b
4
PROFVAR + b
5
LEV + b
6
ROA, 6
where, RET is percentage stock return on the IPO on the first day of listing. It is calculated as price at the end of the first trading day-issue offer priceissue
offer price; FE, actual profit minus forecast profit, divided by forecast profit, and multiplied by 100; SIZE, RED, PROFVAR, LEV, and ROA are as described
earlier see Eq. 5 and they are added as control variables; FE is the focus of attention. We hypothesize a positive coefficient on forecast error FE.
4
.
2
. Data Our sample data come from all Hong Kong and Chinese company IPO listings
on the Stock Exchange of Hong Kong during the period 1993 – 1996.
9
These listings include China domiciled H-shares and China-affiliated Red Chip stocks. IPOs that
do not have an explicit forecast of profit or earnings per share for the year ending after the prospectus date are omitted from the sample.
10
The data are hand
8
If profits are greater than forecast, and if investors anticipate this immediately, then initial returns should be positive as the issue price is based on the profit forecast.
9
Foreign companies that list in Hong Kong are excluded from the sample.
10
A few IPOs give a range of profits for the forecast. These companies are excluded from the sample. There are no cases of Hong Kong companies not giving a profit forecast.
collected from IPO prospectuses, annual accounts for the actual profit numbers, and from the Stock Exchange of Hong Kong SEHK for share prices. A total of
162 initial public offerings are identified that meet our data requirements. Approx- imately 76 of companies are local, 14 are H-shares, and 10 are Red Chips.
Most IPOs 55 are classified by the SEHK as being ‘industrials and manufactur- ing’; remaining companies are spread across the utility, finance, property, and
conglomerate sectors. Descriptive statistics of the independent variables used in the cross-sectional regressions are reported in Table 1.
5. Results