# IPO underpricing, firm quality, and analyst forecasts.

We find that IPO underpricing is positively related to post-IPO growth in sales and EBITDA, but is not significantly related to growth in earnings. Our evidence suggests that accrual reversals or earnings management may cause this inconsistency. We interpret the growth rates of sales and EBITDA as measures of firm quality, and conclude that our evidence supports the notion that IPO firms with greater underpricing are of better quality. Our tests on analysts 'earnings forecast errors show that analysts are less positively biased in their earnings forecasts for IPO firms that have greater underpricing.**********

In this paper we examine the relation between initial public offering (IPO) underpricing and post-IPO growth rates of accounting performance variables. Previous research on IPOs suggests that IPO underpricing is related to the quality of the IPO firms (Allen and Faulhaber, 1989, Welch, 1989, Grinblatt and Hwang, 1989, and Rock, 1986). If we use growth rates of accounting performance variables (such as earnings) as measures of firm quality, then this suggestion implies a positive relation between IPO underpricing and growth rates. Our hypothesis is that IPO firms with greater underpricing should have higher post-IPO growth rates of accounting performance variables. However, rather than just concentrate on growth in earnings, we also include growth in EBITDA and growth in sales. Relative to earnings, these latter variables are more difficult for managers to manipulate. Thus, to reduce the possible contamination of our analysis by earnings management strategies, we consider the growth rates of earnings, EBITDA, and sales, and how they relate to IPO underpricing.

Consistent with our hypothesis, we find that underpricing is positively related to growth in sales and EBITDA in the five years following an IPO. These results suggest that the IPOs with greater underpricing have higher quality. We also find that IPO growth rates in earnings are much lower than the growth rates in sales and EBITDA. Further tests show that this inconsistency is likely caused by earnings management at the time of the IPO, which inflates earnings initially and thus reduces earnings growth rates in subsequent years. This result suggests that earnings growth rates may not be accurate measures of IPO firm quality compared to sales and EBITDA growth rates. We find that the firms with greater underpricing experience larger decreases in accruals after the first year. This finding implies that earnings management may also reduce the correlation between IPO underpricing and earnings growth rates.

We examine the accuracy of analysts' earnings forecasts to determine how analysts react to the relation between IPO growth rates and IPO underpricing. If analysts are misled by earnings management, they may predict higher future earnings for the IPOs with greater underpricing. Our test results do not support this conjecture.

The paper is organized as follows. In Section I, we review the relevant prior research and develop our hypotheses. We describe the sample and discuss sample characteristics in Section II. In Section III, we describe our tests conducted on the relation between IPO growth rates and IPO underpricing. We report our examination of analysts' earnings forecast errors in Section IV. Section V presents our conclusions.

I. Prior Research and Hypotheses Development

Numerous studies show that IPOs are underpriced (see, for example, Ibbotson, 1975, Ritter, 1984, Hanley, 1993, and others). One important rationale for the underpricing of IPOs is the "winner's curse" explanation introduced by Rock (1986). Rock argues that rationing will result if IPO demand is unexpectedly strong. Informed investors will attempt to buy shares only when an issue is underpriced. Uninformed investors do not know which issues will be underpriced or overpriced, and so they will be allocated only a fraction of the most desirable new issues, and allocated all of the least desirable new issues. Faced with this adverse selection problem, uninformed investors will usually submit purchase orders only if IPOs are underpriced enough to compensate them for the bias in the allocation of new issues. Beatty and Ritter (1986), Koh and Walter (1989), and others provide empirical evidence consistent with Rock's (1986) model.

One implication of Rock's model is that greater underpricing will result for firms that informed investors identify as having good quality. Several other papers have similar predictions. The signalling models in Allen and Faulhaber (1989), Welch (1989), and Grinblatt and Hwang (1989) suggest that in order to signal their good quality, IPO firms are intentionally underpriced. In these models, the IPO firm itself knows its prospects best. The firms with the most favorable prospects find it optimal to signal their type by underpricing their initial issue, because they can expect to recoup the cost of underpricing in subsequent seasoned issues. Low-quality firms must invest in imitation expenses to appear to be high-quality firms; and with some probability this imitation is discovered between offerings (Welch, 1989). The underpricing of high-quality firms at the initial public offering adds sufficient signaling costs to the imitation expenses of low-quality firms and makes the expected gain from imitation negative. Thus, low-quality firms abandon the imitation strategy and voluntarily reveal their quality. This argument implies that low-quality firms do not underprice their IPOs as much as do high-quality firms, so investors correctly perceive underpricing as a signal of the firm's quality.

If we use growth rates of accounting performance measures as indicators of firm quality, then the models we cite above imply that IPO underpricing should be positively related to these growth rates. Therefore, we hypothesize that IPO underpricing should be positively related to growth rates in sales, EBITDA, and earnings. This hypothesis is consistent with the results in Chung, Li, and Yu (2005) that IPO underpricing is positively related to the fraction of the offer price that is accounted for by the present value of growth opportunities.

II. Sample and Data Description

In this section we describe the sample, the growth rates of accounting performance variables for the IPOs in the sample, and the changes in accruals for the IPOs.

A. Sample Description

Our initial sample comprises IPOs from 1982 to 1998. We do not use IPOs after 1998 because we need five subsequent years of accounting data for our analysis. We collect the IPOs from the Securities Data Corporation (SDC) database and update them using Jay Ritter's corrections (see http://bear.cba.ufl.edu/ritter). To be included in our study, IPOs must satisfy the following criteria:

1) The IPO stock must be listed on the Center for Research in Security Prices (CRSP) database at the time of IPO.

2) The IPO company must have financial data available from the Compustat database for the fiscal year ending after the IPO.

3) The IPO must be for the issue of ordinary common shares and should not be a unit offering, closed-end fund, real estate investment trust (REIT), or an American Depository Receipt (ADR).

4) The IPO must not be a banking firm (firms with SIC codes between 6000 and 6199).

5) The IPO should have an offer price of at least $5. (1)

There are 3,879 IPOs from 1982 to 1998 in our final sample that satisfy these criteria. Table I reports their characteristics. Consistent with other studies, we find that the average initial return (underpricing) is around 12%. Underpricing, sales, EBITDA, and earnings of the IPO companies fluctuate significantly over time. Underpricing varies from 3.98% in 1984 to 22.37% in 1995; this result is consistent with Loughran and Ritter (2004). Although there is no obvious trend in average sales and EBITDA, the earnings of the IPO companies seem to decrease in the 1990s. Additional tests show that the higher average sales and EBITDA in 1985 are caused by two large outlier companies.

B. IPO Growth Rates

Following Chan, Karceski, and Lakonishok (2003), we estimate growth on a per-share basis for three operating performance measures: 1) net sales (Compustat item #12), 2) operating income before depreciation (EBITDA, Compustat item #13), and 3) income before extraordinary items available for common equity (earnings, Compustat item #237). We adjust these measures for stock splits and dividends. To place firms with different dividend policies on an equal footing, we reinvest all cash dividends and any special distributions in the stock. We then calculate the growth rates for the three measures by comparing their values at the end of the second, third, fourth, and fifth fiscal years after the IPO (Years 2, 3, 4, and 5) to the values at the end of the first fiscal year (Year 1) after the IPO. We include only those firms with positive measures in Year 1 in our calculation of growth rates.

Different industries grow at different rates during different stages of the economic cycle. IPO companies tend to be concentrated in high-growth industries, so their growth rates may simply reflect industry-wide growth patterns. To reduce the effect of such patterns on IPO growth rates, we calculate industry-adjusted growth rates for IPO firms. Specifically, we classify the IPO firms and all seasoned firms into 48 industry groups following Fama and French (1997). To qualify as a seasoned firm, a company must be included in CRSP for five years. Then for each IPO firm and growth rate variable, we calculate the growth rates during the same period for all seasoned firms in the same industry and subtract the median of these from the IPO firm's growth rate to obtain the IPO firm's industry-adjusted growth rate. Table II reports both the IPO growth rates and industry-adjusted IPO growth rates.

Table II shows that the sales of the IPO firms grow at high rates from Year 1 to Year 5, with or without adjustment for industry growth rates. The median IPO sales growth rate from Year 1 to Year 2 is 29.3% before industry adjustment and 21.5% after industry adjustment. IPO firms' incremental sales growth in Years 3, 4, and 5 gradually slows, but remains significantly higher than that of seasoned firms in the same industry.

IPO firms' EBITDA numbers grow at relatively lower rates. For example, the median EBITDA growth rate from Year 1 to Year 2 is 17.1% before industry adjustment and 7.9% afterwards. The incremental EBITDA growth in Years 3, 4, and 5 also slows down, but the rates are still significantly higher than their seasoned industry counterparts (with p-values < 0.05 except for Year 5 with a p-value of 0.1).

The earnings growth rates of IPO firms are much lower than those for sales and EBITDA. From Year 1 to Year 2, the median earnings growth rate is 13.4% before industry adjustment and 6.9% afterwards. Incremental earnings growth in Years 3, 4, and 5 is actually negative, which we can see by the declining median total growth numbers reported for these years. Through Years 4 and 5, the median total growth is negative, which indicates that earnings have dropped below Year 1 values. The earnings growth of seasoned industry peers does not show the same pattern. This difference makes industry-adjusted earnings growth rates even lower in Years 3, 4, and 5. Overall, the pattern of growth in earnings of IPO firms is not consistent with the patterns of growth in sales and EBITDA.

C. IPO Accruals

A possible explanation for the inconsistency between earnings growth rates and other growth rates is earnings management. Teoh, Wong, and Rao (1998) find that IPOs have high positive issue-year earnings and abnormal accruals followed by poor long-run earnings and negative abnormal accruals. This evidence implies that IPO companies manipulate their accounting figures to inflate their earnings reported in the first fiscal year ending after the IPO. Since it is difficult to continually increase accruals, they cannot continue this manipulation in the following few years, so earnings growth rates are either significantly reduced or even become negative.

To determine if earnings management causes the inconsistency between our observed earnings growth rates and our other growth rates, we examine the change in accruals for the IPO firms by examining the difference between accruals at the end of Years 2 through 5 relative to the accruals at the end of Year 1. For fiscal years ending in 1987 and earlier, we calculate accruals as follows:

[Accruals.sub.k] = ([CA.sub.k]-[CA.sub.k-1]) - ([CASH.sub.k]-[CASH.sub.k-1]) - ([CL.sub.k]-[CL.sub.k-1]) + ([CLD.sub.k]-[CLD.sub.k-1]) + ([TAXP.sub.k]-[TAXP.sub.k-1]) - [DEP.sub.k] (1)

where [Accruals.sub.k] is total accruals for year k;

[CA.sub.k] is current assets for year k (Compustat item #4);

[CASH.sub.k] is cash for year k (Compustat item #1);

[CL.sub.k] is current liabilities for year k (Compustat item #5);

[CLD.sub.k] is debt included in current liabilities for year k (Compustat item #34);

[TAXP.sub.k] is taxes payable for year k (Compustat item #71); and [DEP.sub.k] is depreciation and amortization for year k (Compustat item #14).

Following Hribar and Collins (2002), for fiscal years ending in 1988 and later, we calculate accruals as follows:

[Accruals.sub.k] = [Earnings.sub.k] - [CFO.sub.k] (2)

where [Accruals.sub.k] is total accruals for year k, [Earnings.sub.k] is income before extraordinary items for year k (Compustat item #237), and [CFO.sub.k] is cash flow from operations for year k and is equal to net cash flow from operating activities for year k (Compustat item #308) minus extraordinary items and discontinued operations for year k (Compustat item #124).

To make accruals comparable between different companies, we scale them by the average of the total assets (Compustat item #6) at the beginning and end of year k. We examine the variable [DELTA][Accruals.sub.it], which is the change in scaled accruals for firm i from Year 1 to Year t following the IPO (t = 2, 3, 4, or 5).

Using similarly constructed changes in scaled accruals variables, we also calculate medians for seasoned firms in the same industry during the same periods. We subtract the relevant industry median accruals change from the IPO firm's accruals change to determine an industry-adjusted accruals change for the IPO firm. Table III reports the summary statistics.

Table III shows that the accruals of IPO firms decline significantly in Years 2, 3, 4, and 5, relative to Year 1. These declines are roughly 2.4%, 6.3%, 8.4%, and 8.6% of average total assets in the four years, respectively. When we adjust for the accruals change of industry peers, we still find declines of similar magnitudes. All the declines are statistically significant based on Wilcoxon tests. This reversal of accruals reduces earnings growth, and we observe that it eventually becomes negative.

This evidence is consistent with the view of Teoh, Wong, and Rao (1998) that IPO firms manipulate their earnings. It is also consistent with explanations that do not rely on the disingenuous behavior of management. For example, IPOs raise capital and this infusion of capital allows many IPO firms to expand their fixed assets. The asset expansion, in turn, creates larger depreciation charges in future years that reduce earnings in those years. Thus, earnings growth rates may appear low relative to sales and EBITDA growth rates even when no accounting manipulation has occurred. Regardless of the cause of the changes in accruals, our evidence suggests that to measure IPO firm quality, the growth rates of reported earnings may not be as useful as the growth rates of sales and EBITDA.

III. IPO Underpricing and Growth Rates

In this section, we examine the relation between IPO growth rates and IPO underpricing.

A. IPO Underpricing and Growth Rates: Univariate Tests

Table IV reports the Spearman correlation coefficients between industry-adjusted IPO growth rates and IPO underpricing.

This table shows that the industry-adjusted sales growth and EBITDA growth of IPO firms are positively related to IPO underpricing. All the correlation coefficients are positive and significant at a level of at least 1%. For sales growth rates, the correlation coefficients range from 0.11 to 0.14. For EBITDA growth rates, the correlation coefficients range from 0.06 to 0.12. These results are consistent with our hypothesis.

Table IV also shows that there are only weak correlations between industry-adjusted earnings growth rates and IPO underpricing. The correlation coefficients between IPO underpricing and earnings growth are positive and significant only for growth from Year 1 to Year 2 with a value of 0.06. Growth from Year 1 to Years 3, 4, or 5 is not correlated with IPO underpricing and the resulting correlation coefficients are not statistically different from zero.

These results provide limited support for our hypothesis. It appears that the relation between IPO underpricing and growth is only consistently significant for sales and EBITDA growth, but not for earnings growth.

Although not reported in Table IV, the correlations between raw growth rates (before industry adjustment) and IPO underpricing are similar to those in Table IV. Sales and EBITDA raw growth rates are positively related to IPO underpricing, and the correlations between earnings raw growth and underpricing are much weaker.

B. IPO Underpricing and Growth Rates: Regression Tests

The univariate tests in Table IV do not control for many factors that affect growth rates, such as firm size and research and development (R&D) expenditures. To provide a more refined test, we estimate regressions using the industry-adjusted actual growth rates in sales, EBITDA, and earnings as dependent variables. We use the regression model:

[Growth.sub.it] = [[beta].sub.0] + [[beta].sub.1[IR.sub.i] + [[beta].sub.2][RTN.sub.i] + [[beta].sub.3][LMV.sub.i] + [[beta].sub.4][SGR.sub.i] + [[beta].sub.5][RDS.sub.i] + [[beta].sub.6][EP.sub.i] + [[beta].sub.7][BM.sub.i] + [[beta].sub.8][DP.sub.i] + [[beta].sub.9][TECH.sub.i] + [[epsilon].sub.it] (3)

where [Growth.sub.it] is the industry-adjusted growth rate of either sales, EBITDA, or earnings (all on a per-share basis) for IPO firm i. We calculate the year t growth rate for the performance measure in Years t = 2, 3, 4, and 5, relative to the performance measure in the first year following the IPO (Year 1). IR is IPO underpricing (i.e., the first-day initial return) and is our variable of interest. RTN is the retention ratio (the proportion of shares retained by the pre-IPO shareholders). We include it as a control variable in this model because Leland and Pyle (1977) suggest that higher pre-IPO shareholder equity retention should signal higher quality for the post-IPO company. We calculate retention as 1-(shares offered)/(shares outstanding after the IPO). LMV is the log of market value at the end of the first trading day after the IPO. We include it as a control variable for size effects. Following Chan, Karceski, and Lakonishok (2003), we include SGR, RDS, and TECH in our regression. SGR represents the sustainable growth rate, which we calculate as the product of return on equity and plowback ratio. We measure return on equity as the firm's earnings before extraordinary items divided by book equity. We measure plowback as one minus the ratio of dividends to income before extraordinary items. We set the value of SGR to zero if an IPO firm has negative income. RDS is the intensity of R&D relative to sales. When a firm has no R&D spending, we set this variable to zero. TECH is a dummy variable with a value of one for a company in a high-tech industry. We define an industry as high tech if it has a threedigit SIC code of 283, 357, 366, 737, or a two-digit SIC code of 38 or 48. We obtain all these accounting data from Compustat for the first fiscal year ending after the IPO.

We include three price multiples in the regression: EP, BM, and DR EP is the earnings-to-price ratio, BM is the firm's book-to-market value of equity, and DP is the ratio of dividends to price. We estimate the multiples on a per-share basis, using the price at the end of the first trading day after the IPO and accounting data in the first fiscal year ending after the IPO. These price multiples should reflect information about growth expectations. If the price multiples reflect all future growth information contained in underpricing, then the coefficient for underpricing will be nonsignificant. The EP ratio can also serve as control for current earnings.

To reduce the influence of growth rate outliers on regression results, we drop observations that show growth rates within the bottom or top 1% of observed growth rates. Table V reports the regression results.

Panel A of Table V provides the regression results when we use industry-adjusted growth rates in sales as the dependent variables. The coefficients of underpricing are positive and significant in all four years. Thus, even after we control for other variables such as price multiples, we see that greater underpricing is associated with a higher sales growth rate. This result is consistent with the univariate test results in Table IV and provides support for our hypothesis.

Panel B of Table V provides the regression results using industry-adjusted growth in EBITDA as the dependent variable. Similar to our findings in Panel A, the coefficients of underpricing are positive and significant for all four years. Thus, IPO firms with greater underpricing tend to have higher growth rates in EBITDA. This evidence supports our hypothesis. Thus, when we measure firm quality by growth rates in either sales or EBITDA, we find that firms with greater underpricing have higher quality.

However, in Panel C of Table V, the regressions of industry-adjusted growth in earnings show that the coefficients of underpricing are nonsignificant in all years. This evidence implies that IPO underpricing is not significantly related to growth in earnings of IPO firms.

We note that most of the independent variables have little explanatory power, and that the regressions' [R.sup.2] numbers are usually smaller than those found for the regressions using sales or EBITDA. This result is consistent with the evidence in Chan, Karceski, and Lakonishok (2003), who show that earnings growth rates are hard to predict. Although not reported in our paper, we also estimate the regressions after dropping the price multiples, and the results are similar to the results in Table V. Thus, we believe that the nonsignificant coefficients for underpricing are not caused by the fact that the price multiples also reflect growth expectations. The results for earnings regressions do not support our hypothesis.

The coefficients of the other independent variables are generally consistent with other studies and our intuition. For example, the coefficients of the price multiples are generally consistent with valuation theory and the coefficients of RDS are generally positive in the sales regressions. The coefficients of retention (RTN) are mostly nonsignificant. One interpretation is that the price multiples have already incorporated the information contained in retention. Consistent with this interpretation, when we re-estimate the sales regressions without the price multiples, we find that the coefficients of retention become positive and significant. Another reason could be the correlation between retention and underpricing. In our sample, IPO retention is positively related to underpricing with a correlation coefficient of 0.07. When we drop underpricing from the sales and EBITDA growth regressions, the coefficients of retention become positive and significant in the majority of cases. Therefore, IPO underpricing and retention may convey similar information about post-IPO growth rates. Leland and Pyle (1977) assume that the market's valuation coincides with the insider's. However, if the insider can sell at an attractive price, then a high-quality firm may have low retention. This possibility may also reduce the significance of retention in the regressions.

One concern about our results is that our sample sizes change considerably across different regressions: the sample sizes in Table V range from 3,481 for Year 2 of Panel A to 1,911 for Year 5 of Panel C. This variation arises because some firms are lost as the growth measure is changed and the horizon of growth is extended. This fact raises the question of whether the differences between earnings results and the other results are due to changes in the underlying sample.

To address this concern, we construct a common subsample that contains only the IPOs with growth rates available for sales, EBITDA, and earnings in Years 2 through 5. The IPO firms in this subsample have survived the whole five years, have positive EBITDA and earnings in the first year, and have data available for sales, EBITDA, and earnings for all five years. There are 1,890 IPOs that satisfy these requirements. We re-estimate all the growth regressions using this common subsample. Although not reported in the table, the results are similar to those in Table V. In regressions of sales and EBITDA growth rates, the coefficients of underpricing are all positive and significant. In regressions of earnings growth rates, these coefficients are all nonsignificant. We find the same pattern: the sales and EBITDA growth results provide support for our hypothesis, whereas the earnings results do not.

Another concern about our results is that we drop many firms used in the sales growth regressions from the EBITDA and earnings growth rate regressions because they have negative EBITDA or earnings. This practice might affect our tests on the relation between IPO underpricing and the growth in EBITDA or earnings. To ensure that our results are robust, we use another method to calculate the growth rates in EBITDA and earnings: we find the changes in EBITDA per share and in earnings per share (EPS) in Years 2, 3, 4, and 5 relative to Year 1 and normalize these changes by sales per share in Year 1. Then we use these normalized changes as the growth rates in the regression specified in Equation (3). Although not reported in the tables, the regression results are similar to those in Table V: the changes in EBITDA are positively related to underpricing, and the changes in earnings are not significantly related to underpricing. Again, the earnings growth results do not add support for our hypothesis, but the EBITDA results do. When we reestimate the regressions with these alternative growth measures using a common subsample, the results are similar. We conclude that the discrepancy between the results using earnings growth and the other growth measures is not caused by any of the potential sample differences inherent in our analysis.

Although not reported in the tables, we find similar results when we estimate the regressions using raw growth rates (not adjusted by industry) as dependent variables. The results are qualitatively similar when we eliminate the top and bottom 5% of growth observations or winsorize at the mean plus or minus three standard deviations. Finally, the results are unaffected by: 1) estimating all the regressions by using annual fraction ranks instead of the actual values of the relevant variables, 2) adding industry and year dummies, and 3) estimating growth rates relative to accounting data in the fiscal year ending prior to the IPO.

C. IPO Underpricing and Accruals Change: Regression Tests

Earnings management reduces the reliability of earnings growth rates as measures of firm quality. Thus, we investigate whether accruals changes, which could imply earnings management, cause the lack of relation between earnings growth and IPO underpricing. Relative to other IPO firms, if an IPO firm with greater underpricing and greater growth in sales and EBITDA also has greater reductions in accruals, then its earnings growth rates will not be correspondingly higher. The outcome is that we will not find a positive relation between IPO underpricing and earnings growth even though growth in sales and EBITDA are positively related to IPO underpricing.

We estimate the following regression model to investigate whether underpricing is related to accrual changes. The regression results will indicate whether accruals changes obscure the relation between earnings growth and underpricing.

[DELTA][[Accruals.sub.it] = [[beta].sub.0] + [[beta].sub.1][IR.sub.i] + [[beta].sub.2][RTN.sub.i] + [[beta].sub.3][LMV.sub.i] + [[beta].sub.4][SGR.sub.i] + [[beta].sub.5][RDS.sub.i] + [[beta].sub.6][EP.sub.i] + [[beta].sub.7][BM.sub.i] + [[beta].sub.8][DP.sub.i] + [[beta].sub.9][TECH.sub.i] + [[epsilon].sub.it]. (4)

In this model, the dependent variable, [DELTA][Accruals.sub.it], is the industry-adjusted accruals change for firm i in year t following the IPO year. We measure each accruals change for years t = 2, 3, 4, or 5 relative to Year 1. We normalize the accruals in each year by the average of the beginning and year-end total assets. The independent variables are the same as those in Equation (3). Table VI reports the regression results.

The results in Table VI support our hypothesis. The coefficients of underpricing are negative in regressions of accruals changes for Years 3 and 4. This evidence indicates that the IPOs with greater underpricing experience a greater decrease in accruals in the two years and explains why these IPO companies have higher growth in sales and EBITDA, but not correspondingly higher growth in earnings. The high accruals in Year 1 are reversed in subsequent years, so the earnings and earnings growth in subsequent years are dampened. As we noted earlier in the paper, the pattern of accruals is consistent with earnings management at the time of the IPO. However, it may also be consistent with other explanations.

The coefficients of other independent variables are generally consistent with our expectations. For example, the coefficients of EP are negative and significant in all four regressions. These negative coefficients are consistent with higher Year 1 earnings being associated with a greater decrease in accruals. Again, earnings management in Year 1 is one possible explanation of this behavior.

In further tests we find that the use of raw changes in accruals (not industry adjusted) as dependent variables produces results that are similar to those in Table VI. When we estimate the regressions for different years using a common subsample (as outlined in the robustness tests for Table V), the results are also similar.

Overall, the results of the accruals change regressions suggest that the inconsistency in the relation between underpricing and the different growth rates is at least partially caused by the reversal of accruals.

IV. IPO Underpricing and Analysts' Earnings Forecast Errors

The tests above show that IPO underpricing is positively related to growth in sales and EBITDA, but not to growth in earnings. One of the possible causes of this discrepancy is earnings management. This possibility leads us to ask how the market reacts to this and whether the market still expects higher earnings growth for the IPOs with greater underpricing.

There are reasons to suspect that earnings management can fool the market. Teoh, Welch, and Wong (1998) find that the IPOs with higher accruals in the first year underperform in the long run. This evidence implies that the market initially does not recognize earnings management and still expects high earnings growth rates for these IPO firms. If so, it is possible that the market may still project higher future earnings for the IPOs with greater underpricing, given the fact that these IPOs do have higher growth in sales and EBITDA. To explore this possibility, we use analyst forecasts as our proxy for market expectations and examine the forecast errors of analyst forecasts for annual earnings. If analysts naively forecast higher future earnings for IPO firms with greater underpricing, then they will have higher positive forecast errors for these firms.

A. Analysts' Earnings Forecast Errors for IPO Firms

We collect EPS forecasts and actual EPS numbers for each IPO firm from the IBES database. These forecasts are made in the year after the IPO for the EPS in the first four fiscal years ending after the IPO (Years 1, 2, 3, and 4). For example, the forecast for Year 4 is a four-year-ahead forecast of EPS made in the year after the IPO. We do not include Year 5 because very few IPO companies have this data available in the year after the IPO. We calculate forecast errors as follows:

[FE.sub.it] = Consensus EPS [Forecast.sub.it]--Actual [EPS.sub.it]/ Consensus EPS [Forecast.sub.it] (5)

where we calculate the Consensus EPS [Forecast.sub.it], as the average of EPS forecasts made by different analysts for firm i for the tth fiscal year ending after the IPO. Positive forecast errors imply analyst over-optimism, and negative forecast errors imply analyst over-pessimism.

We calculate Consensus EPS Forecast as the average of EPS forecasts made by different analysts for the same fiscal year. In calculating the average, we use only the first forecast made by each analyst. We note that because we require all the forecasts to be made in the year after the IPO, the first forecast of one analyst may be made as much as 12 months before the first forecast of another analyst. However, for about 90% of the IPOs, the first forecasts are all made within six months of each other. If any observation has a negative consensus EPS forecast, then we drop it.

To adjust for any possible industry-wide analyst bias, we also calculate industry-adjusted EPS forecast errors. We do so by calculating the forecast errors for all seasoned firms in the same industry during the same period and then subtracting the median forecast error of the industry from the IPO forecast errors. We report the key statistics for IPO forecast errors and industry-adjusted IPO forecast errors in Panel A of Table VII.

We find positive analyst forecast errors for our group of IPO firms. This result is consistent with Rajan and Servaes (1997), who also find that analysts are systematically overly optimistic in their EPS forecasts. For Year 2, the median calculated forecast error is 9.7%. This result implies that the actual EPS of an average IPO firm is about 90% of the forecast. For Years 3 and 4, the median forecast errors are 40.5% and 50.1%, respectively, which suggests that the actual EPS of IPO firms for Years 3 and 4 are only 60% and 50%, respectively, of the EPS forecasts. The median forecast errors for Years 2, 3, and 4 are all significantly different from zero with p-values less than 0.01 (using the Wilcoxon signed rank test).

Since many studies find that analysts tend to be positively biased in their forecasts (see, for example, Brown, Foster, and Noreen, 1985; Butler and Lang, 1991 ; and O'Brien, 1988), we must adjust industry-adjusted forecast errors to remove the influence of this bias in our analysis. After we adjust for the industry forecast errors, we find that the median forecast errors for Years 1 and 2 become negative (Year 2 is significantly different from zero with a p-value of less than 0.01). However, the median industry-adjusted forecast errors for Years 3 and 4 are both positive and significant (with p-values at 0.01 or less). Thus, for our group of IPO firms, analysts' initial forecasts seem to have less positive bias for EPS within the first two years, but a greater positive bias for Years 3 and 4.

B. IPO Underpricing and Analysts' EPS Forecast Errors: Univariate Tests

If analysts are misled by earnings management, they may project higher future earnings for the IPOs with greater underpricing because these IPOs have higher growth in sales and EBITDA. This possibility will lead to higher forecast errors for the IPOs because, as Table V shows, the growth of actual earnings is not related to underpricing. To determine if there is a positive relation between these variables, we first analyze the Spearman correlation coefficients between underpricing and forecast error. We report the results in Panel B of Table VII.

Panel B of Table VII shows that the forecast errors are negatively related to IPO underpricing. Before the industry adjustment, underpricing is negatively related to forecast errors for all four years (with p-values less than 0.01). After the industry adjustment, underpricing is still negatively related to forecast errors for the four years, with all p-values less than 0.1 and p-values less than 0.01 for the first three years. These results suggest that analysts are less positively biased for IPO firms that have greater underpricing. This implication is contrary to what we would expect if analysts are fooled by earnings management. In fact, this implication is consistent with analysts being wary of IPO firms with greater underpricing, and indicates that analysts may untangle the accruals and possible earnings management.

C. IPO Underpricing and Security Analysts' EPS Forecast Errors: Regression Tests

To conduct a more refined test of the relation between forecast errors and underpricing, we estimate the following regression model:

[FE.sub.it] = [[beta].sub.0] + [[beta].sub.1][IR.sub.i] + [[beta].sub.2][RTN.sub.i] + [[beta].sub.3][LMV.sub.i] + [[beta].sub.4][NAN.sub.i] + [[beta].sub.5][AW.sub.i] + [[beta].sub.6][SGR.sub.i] + [[beta].sub.7][RDS.sub.i] + [[beta].sub.8][EP.sub.i] + [[beta].sub.9][BM.sub.i] + [[beta].sub.10][DP.sub.i] + [[beta].sub.11][TECH.sub.i] + [[epsilon].sub.it]. (6)

[FE.sub.it] is the industry-adjusted forecast error. Among the independent variables, NAN is the number of analysts who make EPS forecasts for the IPO stock in the year after the IPO. AW is the average forecast window of the EPS forecasts, where we define the forecast window as the period between the forecasted fiscal year-end and the time when the forecast is made.

To reduce the effects of extreme values of forecast error, we drop the top and bottom 1% of observations. We report the regression results in Table VIII.

In Table VIII, the coefficients of underpricing (IR) are negative and significant at a 5% level in Years 1, 2, and 3. This result implies that for these years, analysts are less positively biased in their EPS predictions for IPO firms with greater underpricing. In Year 4, the coefficient of underpricing is nonsignificant. These results are generally consistent with our univariate analysis results.

We conduct several robustness checks of our regression results. Although not reported in the table, our results are similar when we: 1) estimate the regressions using raw analyst forecast errors (not adjusted by industry) as dependent variables, 2) winsorize the observations at the mean plus or minus three standard deviations, 3) re-estimate the regressions using a common subsample of IPOs that have EPS forecast errors available for all years, and 4) use forecasts that we restrict to a common three-month window following the IPO.

When we ask what the univariate and regression results suggest about analysts' reactions to IPO underpricing, we believe that one answer is obvious: analysts are not fooled by earnings management or accruals into predicting higher EPS for the IPO firms with greater underpricing. If they had been misled, then we would find a positive relation between analyst forecast errors and underpricing, but in fact, we find a negative relation. However, it seems that analysts have only partially incorporated the information into their EPS forecasts. They seem to correctly reduce their positive forecast bias for firms with greater IPO underpricing. Yet, our results do not answer the question of why there is a higher positive EPS forecast bias for firms with lower IPO underpricing.

V. Conclusions

In this paper we examine the relation between IPO underpricing and post-IPO growth rates of accounting performance variables. Other studies suggest that underpricing may be positively related to firm quality. Using growth rates in sales, EBITDA, and earnings as measures of firm quality, we hypothesize that IPO firms with greater underpricing should have higher growth rates.

Our tests find that IPO underpricing is positively related to growth in sales and EBITDA. This evidence is consistent with our hypothesis. However, we find that underpricing is not positively related to growth in earnings. We find evidence that the reversal of accruals (possibly a sign of earnings management) may explain this discrepancy: IPO firms with greater underpricing have greater reductions in accruals, thus causing earnings to be depressed in future years. This pattern of accruals changes obscures the relation between earnings growth and quality and obscures the link to IPO underpricing. Since sales and EBITDA are less influenced by accrual changes or earnings management, we conclude that the positive relation between underpricing and sales and EBITDA growth rates is sufficient to support our hypothesis in spite of the accruals-obscured earnings results. Thus, our results add evidence to the notion that IPO firms with greater underpricing are of higher quality.

We also investigate how analysts react to the relation between IPO firms' growth rates and underpricing. We find that analysts do not make higher EPS forecasts for IPOs with greater underpricing. This result implies that analysts do not naively project sales and EBITDA growth into earnings growth, and that they seem to untangle the effect of accruals on future EPS. What remains unexplained is why analysts still seem to be more positively biased in their EPS forecasts for IPO firms that have lower underpricing. We leave this mystery to further research on analysts' forecasts bias.

References

Allen, F. and G.R. Faulhaber, 1989, "Signalling by Underpricing in the IPO Market," Journal of Financial Economics 23, 303-323.

Beatty, R.P. and J.R. Ritter, 1986, "Investment Banking, Reputation, and the Underpricing of Initial Public Offerings," Journal of Financial Economics 15, 213-232.

Bradley, D.J., J.W. Cooney, S.D. Dolvin, and B.D. Jordan, 2006, "Penny Stock IPOs," Financial Management 35, 5-29.

Brown, P., G. Foster, and E. Noreen, 1985, Security Analyst Multi- Year Earnings Forecasts and the Capital Market, Studies in Accounting Research No.21, Sarasota, American Accounting Association.

Butler, K. and L. Lang, 1991, "Evidence of Systematic Optimism and Pessimism," Journal of Accounting Research 29, 150-156.

Chan, L., J. Karceski, and J. Lakonishok, 2003, "The Level and Persistence of Growth Rates," Journal of Finance 58, 643-684.

Chung, K. H., M. Li, and L. Yu, 2005, "Assets in Place, Growth Opportunities, and IPO Returns," Financial Management 34, 65-88.

Fama, E. F. and K. R. French, 1997, "Industry Costs of Equity," Journal of Financial Economics 43, 153-193.

Grinblatt, M. and C. Y. Hwang, 1989, "Signalling and the Pricing of New Issues," Journal of Finance 44, 393-420.

Hanley, K. W., 1993, "The Underpricing of Initial Public Offerings and the Partial Adjustment Phenomenon," Journal of Financial Economics 34, 231-250.

Hribar, P. and D.W. Collins, 2002, "Errors in Estimating Accruals: Implications for Empirical Research," Journal of Accounting Research 40, 105-134.

Ibbotson, R. G., 1975, "Price Performance of Common Stock New Issues," Journal of Financial Economics 2, 235-272.

Koh, F. and T. Waiter, 1989, "A Direct Test of Rock's Model of the Pricing of Unseasoned Issues," Journal of Financial Economics 23, 251-272.

Leland, H.E. and D.H. Pyle, 1977, "Informational Asymmetrics, Financial Structure, and Financial Intermediaries," Journal of Finance 32, 371-388.

Loughran, T. and J. Ritter, 2004, "Why Has IPO Underpricing Changed over Time?" Financial Management 33, 5-37.

O'Brien, P, 1988, "Analysts' Forecasts as Earnings Expectations," Journal of Accounting and Economics 10, 53-83.

Rajan, R. and H. Servaes, 1997, "Analyst Following of Initial Public Offerings," Journal of Finance 52, 507-529.

Ritter, J.R., 1984, "The Hot Issue Market of 1980," Journal of Business 57, 215-241.

Rock, K., 1986, "Why New Issues Are Underpriced," Journal of Financial Economics 15, 187-212.

Teoh, S.H., I. Welch, and T.J. Wong, 1998, "Earnings Management and the Long-Run Market Performance of Initial Public Offerings," Journal of Finance 53, 1935-1974.

Teoh, S.H., T.J. Wong, and G.R. Rao, 1998, "Are Accruals During Initial Public Offerings Opportunistic?" Review of Accounting Studies 3, 175-208.

Welch, I., 1989, "Seasoned Offerings, Imitation Costs, and the Underpricing of Initial Public Offerings," Journal of Finance 44, 421-449.

We thank Mark Huson, participants at the 2005 Northern Finance Association Conference, and, in particular, Bill Christie (the Editor) and an anonymous referee for helpful comments and suggestions. The authors gratefully acknowledge the contribution of Thompson Financial for providing earnings forecast data, available through I/B/E/S (Institutional Brokers Estimate System)

(1) See Bradley, Cooney, Dolvin, and Jordan (2006) for discussion about the characteristics of lPO stocks with offer price of $5 or less (penny stocks).

Steven X. Zheng and David A. Stangeland *

* Steven X. Zheng is an Assistant Professor of Finance and David A. Stangeland is the CMA Professor of Strategic Financial Management at the University of Manitoba in Winnipeg, Manitoba, Canada.

Table I. Sample Characteristics This table reports descriptive statistics about our sample of IPOs from 1982 to 1998. The relevant data are retrieved from the Compustat, CRSP, and SDC databases. The initial return is calculated as (1st Day Closing Price)/(Offer Price)-1. Market value is calculated as the product of 1st day closing price and shares outstanding after the IPO, deflated to 1982 dollars. Sales, EBITDA and earnings are also deflated to 1982 dollars. Year # of Average Average IPOs Initial Market Return Value (%) ($millions) 1982 55 12.76 65.12 1983 351 11.05 93.36 1984 124 3.98 49.83 1985 147 7.30 73.63 1986 293 6.76 72.25 1987 198 6.84 57.59 1988 71 7.15 66.11 1989 80 9.75 59.63 1990 79 11.69 65.75 1991 187 14.55 76.65 1992 280 10.63 69.62 1993 380 13.77 72.77 1994 292 10.36 52.80 1995 344 22.37 86.84 1996 515 17.43 93.20 1997 334 14.58 82.32 1998 149 19.48 95.60 All 3879 12.91 76.59 Year Average Average Average Sales EBITDA Earnings ($millions) ($millions) ($millions) 1982 27.66 3.92 2.19 1983 56.03 6.85 2.33 1984 52.01 7.88 3.27 1985 110.49 16.25 3.78 1986 74.54 10.07 3.45 1987 64.40 7.51 2.13 1988 75.26 10.97 4.98 1989 56.34 6.56 2.42 1990 60.27 6.61 2.31 1991 51.07 6.27 1.89 1992 56.40 6.77 1.67 1993 50.96 6.22 1.13 1994 42.59 5.10 1.67 1995 42.00 4.46 1.12 1996 46.62 4.39 0.32 1997 53.50 5.13 0.53 1998 45.27 3.91 -1.90 All 54.99 6.52 1.61 Table II. IPO Growth Rates in Sales, EBITDA, and Earnings IPO growth rates are calculated on a per share basis for three operating performance measures: net sales, operating income before depreciation (EBITDA), and income before extraordinary items available for common equity (earnings). These measures are adjusted for stock splits and dividends. The growth rate for Year t following the IPO is calculated as the percentage change in the performance measure relative to its value in the first year after the IPO (Year 1). Only firms with positive values in Year 1 are included in the calculation. All cash dividends and any special distributions are reinvested in the stock. The industry-adjusted growth rates are calculated as the difference between IPO growth rates and median growth rates in the same year for seasoned firms in the same industry. The Wilcoxon p-values are for the two-tailed test about the null hypothesis that the medians are not significantly different from zero. % Growth Relative to the Amount Reported at the End of the First Fiscal Year Year (Year 1) Following the IPO Following the IPO Sales Percentile Wilcoxon 25% 50% 75% p-value 2 8.3 29.3 59.1 0.00 3 16.2 52.1 117.8 0.00 4 20.6 70.4 151.1 0.00 5 19.7 83.9 190.1 0.00 % Growth Relative to the Amount Reported at the End of the First Fiscal Year Year (Year 1) Following the IPO Following the IPO Sales (Industry-Adjusted) Percentile Wilcoxon # of 25% 50% 75% p-value obs 2 0.6 21.5 50.8 0.00 3575 3 0.4 36.9 100.0 0.00 3221 4 -0.9 48.1 124.9 0.00 2865 5 -9.7 53.0 157.6 0.00 2555 EBITDA Year Percentile Wilcoxon Following the IPO 25% 50% 75% p-value 2 -30.5 17.1 57.0 0.00 3 -49.0 28.3 93.8 0.00 4 -59.5 32.5 118.4 0.00 5 -59.3 34.9 149.6 0.00 EBITDA (Industry-Adjusted) Year Percentile Wilcoxon Following # of the IPO 25% 50% 75% p-value obs 2 -38.6 7.9 48.8 0.00 2814 3 -63.8 15.0 77.9 0.00 2561 4 -83.6 12.9 98.3 0.04 2284 5 -86.1 8.1 120.0 0.10 2054 Earnings Year Percentile Wilcoxon Following the IPO 25% 50% 75% p-value 2 -85.6 13.4 55.6 0.00 3 -123.3 1.7 76.8 0.87 4 -156.0 -13.7 96.9 0.23 5 -182.1 -21.2 120.2 0.01 Earnings (Industry-Adjusted) Year Percentile Wilcoxon Following # of the IPO 25% 50% 75% p-value obs 2 -90.9 6.9 51.3 0.00 2648 3 -124.9 -5.5 69.3 0.35 2428 4 -159.6 -22.3 90.8 0.00 2178 5 -191.2 -30.2 108.0 0.00 1959 Table III. Change in Accruals for IPO Firms Accruals are calculated as income before extraordinary items minus cash flow from operations for fiscal years after 1987. For fiscal years before 1987, we calculate accruals from depreciation and amortization and balance sheet items. We scale accruals by the average of the beginning and year-end total assets. [DELTA][Accruals.sub.it] thet is change in scaled accruals for firm i calculated as the change from Year 1 (the first fiscal year end following the IPO) to Year t (t = 2, 3, 4, or 5) following the IPO. To calculate industry-adjusted changes in scaled accruals, we calculate changes in scaled accruals for all seasoned firms in the same industry during the same period and subtract the median of seasoned firms' accruals changes from the IPO firm's accruals changes. The Wilcoxon p-values are for the two-tailed test about the null hypothesis that the medians are not significantly different from zero. [DELTA]Accruals Relative to Year 1 Year Following the IPO Percentile Wilcoxon 25% 50% 75% p-value 2 -12.9 -2.4 5.6 0.00 3 -18.4 -6.3 2.2 0.00 4 -21.7 -8.4 1.7 0.00 5 -22.8 -8.6 1.8 0.00 [DELTA]Accruals (Industry-Adjusted) Relative Year to Year 1 Following the IPO Percentile Wilcoxon 25% 50% 75% p-value # of obs 2 -12.9 -2.2 5.9 0.00 3090 3 -18.5 -5.9 2.9 0.00 2778 4 -21.0 -7.6 2.7 0.00 2474 5 -21.7 -7.3 2.9 0.00 2215 Table IV. Spearman Correlation Between Growth Rates and IPO Underpricing This table reports the Spearman correlation coefficients between IPO underpricing and industry-adjusted growth rates of IPO firms. Growth rates in net sales, operating income before depreciation, and income before extraordinary items available for common equity are calculated on a per share basis for IPO firms and are adjusted for stock splits and dividends. All cash dividends and any special distributions are reinvested in the stock. The growth rate for Year t is calculated as the percentage change in the performance measure in Year t relative to the value reported as of the first fiscal year end after the IPO (Year 1). Only firms with positive values in the first year following the IPO are included in the calculation. The industry-adjusted growth rate is calculated as the difference between the IPO firm's growth rate and the median of growth rates in the same year for seasoned firms in the same industry. Underpricing is calculated as (1st Day Closing Price)/(Offer Price)-1. Spearman correlation coefficients between growth rates and IPO underpricing are reported in the table. The p-values of the coefficients are reported in parentheses. Spearman Correlation with IPO Underpricing Variables Year 2 Year 3 Year 4 Year 5 Industry-adjusted 0.14 0.13 0.13 0.11 Sales Growth (0.00) (0.00) (0.00) (0.00) Industry-adjusted 0.12 0.10 0.07 0.06 EBITDA Growth (0.00) (0.00) (0.00) (0.01) Industry-adjusted 0.06 0.01 0.00 -0.01 Earnings Growth (0.00) (0.63) (0.97) (0.75) Table V. Regressions of IPO Growth Relative to Year 1 This table reports results of the following regressions: [Growth.sub.it] = [[beta].sub.0] + [[beta].sub.1][IR.sub.i] + [[beta].sub.2] [RTN.sub.i] + [[beta].sub.3][LMV.sub.i] + [[beta].sub.4][SGR.sub.i] + [[beta].sub.5][RDS.sub.i] + [[beta].sub.6][EP.sub.i] + [[beta].sub.7][BM.sub.i] + [[beta].sub.8][DP.sub.i] + [[beta].sub.9][TECH.sub.i] + [[epsilon].sub.it]. Growth represents industry-adjusted growth rates of sales, EBITDA, and earnings reported in Years 2, 3, 4 and 5 relative to the amounts reported in Year 1 (the first fiscal year end after the IPO). The growth rates are estimated on a per share basis and are adjusted for stock splits and dividends. All cash dividends and any special distributions are reinvested in the stock. IR is the underpricing variable (i.e., the first day initial return), calculated as (1st day closing price/offer price-]). RTN refers to retention, which is calculated as (Shares outstanding after IPO--Shares offered in IPO)/ (Shares outstanding after IPO). LMV is the log of market value at the end of the first trading day after the IPO. SGR represents the sustainable growth rate, calculated as the product of return on equity and plowback ratio. RDS is the intensity of R&D relative to sales. All the accounting data are collected from Compustat for the first fiscal year ending after the IPO. EP is the earnings-to-price ratio; BM is the firm's book-to-market value of equity, and DP is the ratio of dividends to price. The multiples are estimated on a per-share basis using price at the end of the first trading day following the IPO and accounting data from the first fiscal year ending after the IPO. TECH is a dummy variable with a value of 1 for a company in a high-tech industry. High-tech industries are defined as those with three-digit SIC codes of 283, 357, 366, 737, or two-digit SIC codes of 38, 48. The independent variables are listed in the first row of the table. Observations are dropped if their growth rates are in the bottom or top 1 % of observed growth rates. Numbers in parentheses are t-statistics. Year IR RTN LMV SGR Following the IPO Panel A. Industry-Adjusted Sales Growth Relative to the Amount in Year 1 Following the IPO 2 0.33 *** 0.18 0.02 -0.46 *** (5.40) (1.55) (1.06) (-3.61) 3 0.47 *** -0.06 0.01 -1.24 *** (3.07) (-0.23) (0.39) (-4.21) 4 0.83 *** 0.35 0.02 -1.63 *** (3.81) (0.88) (0.34) (-3.76) 5 1.16 *** 0.08 -0.05 -1.77 *** (4.06) (0.16) (-0.78) (-3.08) Panel B. Industry-Adjusted EBITDA Growth Relative to the Amount in Year 1 Following the IPO 2 .053 *** 0.23 0.20 *** 0.35 (3.43) (1.10) (6.85) (1.43) 3 0.68 *** 0.20 0.20 *** -0.17 (3.23) (0.66) (5.08) (-0.51) 4 0.82 *** 0.36 0.15 *** -0.35 (2.78) (0.91) (2.74) (-0.75) 5 2.00 *** 0.77 0.14 * -0.75 (5.42) (1.46) (1.93) (-1.26) Panel C. Industry-Adjusted Earnings Growth Relative to the Amount in Year 1 Following the IPO 2 0.43 0.72 0.46 *** 1.20 * (1.11) (1.22) (5.65) (1.82) 3 -0.51 0.55 0.44 *** -0.17 (-1.03) (0.76) (4.58) (-0.20) 4 -0.55 1.34 0.41 *** 2.28 ** (-0.91) (1.61) (3.72) (2.30) 5 0.54 1.53 0.20 1.57 (0.81) (1.57) (1.55) (1.41) Year RDS EP BM DP Following the IPO Panel A. Industry-Adjusted Sales Growth Relative to the Amount in Year 1 Following the IPO 2 0.01 *** -1.31 *** -0.01 1.61 *** (7.17) (-6.41) (-0.22) (3.76) 3 0.06 *** -2.49 *** -0.23 0.29 (8.19) (-5.16) (-1.43) (0.30) 4 0.06 *** -2.63 *** -0.36 0.42 (7.62) (-3.83) (-1.58) (0.31) 5 0.0 6*** -2.65 *** -0.56 * -0.37 (4.67) (-2.96) (-1.82) (-0.21) Panel B. Industry-Adjusted EBITDA Growth Relative to the Amount in Year 1 Following the IPO 2 -1.23 *** 0.27 0.34 ** 3.03 *** (3.24) (0.41) (2.50) (4.04) 3 0.13 -1.84 ** 0.44 ** 2.78 *** (0.20) (-2.02) (2.38) (2.78) 4 -2.36 *** -2.46 ** 0.56 ** 2.18 (-2.81) (-2.02) (2.26) (1.63) 5 -2.95 *** -3.47 ** 0.58 * 1.10 (-2.66) (-2.22) (1.77) (0.65) Panel C. Industry-Adjusted Earnings Growth Relative to the Amount in Year 1 Following the IPO 2 -5.64 *** 10.93 *** -0.05 2.23 (-5.00) (4.41) (-0.12) (1.16) 3 -5.35 *** 15.19 *** -1.04 ** -4.32 * (-3.84) (5.27) (-2.23) (-0.00) 4 -4.42 *** 1.51 1.10 ** 4.20 (-2.75) (0.45) (2.04) (1.51) 5 -3.54 * 3.71 0.77 2.66 (-1.91) (0.96) (1.22) (0.86) Year TECH # of obs Adj. Following [R.sup.2] the IPO Panel A. Industry-Adjusted Sales Growth Relative to the Amount in Year 1 Following the IPO 2 -0.12 *** 3481 0.063 (4.31) 3 -0.20 *** 3133 0.061 (-3.05) 4 -0.21 ** 2788 0.055 (2.23) 5 -0.26 ** 2486 0.035 (-2.08) Panel B. Industry-Adjusted EBITDA Growth Relative to the Amount in Year 1 Following the IPO 2 -0.06 2744 0.039 (-0.95) 3 -0.12 2495 0.026 (-1.35) 4 -0.05 2225 0.017 (-0.43) 5 -0.24 1999 0.032 (-1.55) Panel C. Industry-Adjusted Earnings Growth Relative to the Amount in Year 1 Following the IPO 2 -1.38 *** 2259 0.056 (-3.68) 3 0.14 2370 0.033 (0.69) 4 -0.16 2127 0.021 (0.67) 5 -0.46 * 1911 0.013 (1.71) *** Significant at the 0.01 level. ** Significant at the 0.05 level. * Significant at the 0.10 level. Table VI. Regressions of Accruals Changes Relative to Year 1 This table reports the results of the following regressions: [DELTA][Accruals.sub.it] = [[beta].sub.0] + [[beta].sub.1][IR.sub.i] + [[beta].sub.2][RTN.sub.i] + [[beta].sub.3][LMV.sub.i] + [[beta].sub.4][SGR.sub.i] + [[beta].sub.5][RDS.sub.i] + [[beta].sub.6]EP + [[beta].sub.7]BM + [[beta].sub.8]DP + [[beta].sub.9][TECH.sub.i] + [[epsilon].sub.it]. [DELTA][Accruals.sub.it] is the industry-adjusted change in scaled accruals for firm i calculated as the percent change from Year 1 (the first fiscal year end following the IPO) to Year t (t = 2, 3, 4, or 5) following the IPO. Accruals are calculated as income before extraordinary items minus cash flow from operations for fiscal years after 1987. For fiscal years before 1987, we calculate accruals from depreciation and amortization and balance sheet items. We scale accruals by the average of the beginning and year-end total assets. IR is 1st day initial return (underpricing), calculated as (V day closing price/offer price-1). RTN refers to retention, which is calculated as (Shares outstanding after IPO--Shares offered in IPO)/(Shares outstanding after IPO). LMV is the log of market value at the end of the first trading day after the IPO. SGR represents the sustainable growth rate, calculated as the product of return on equity and plowback ratio. RDS is the intensity of R&D relative to sales. All these accounting data are collected from Compustat for the first fiscal year after the IPO. EP is the earnings-to-price ratio; BM is the firm's book-to-market value of equity; and DP is the ratio of dividends to price. The multiples are estimated on a per-share basis using price at the end of the first trading day following the IPO and accounting data from the first fiscal year ending after the IPO. TECH is a dummy variable with a value of 1 for a stock in the high-tech industries. High-tech industries are defined as those with three-digit SIC codes of 283, 357, 366, 737, or two-digit SIC codes of 38, 48. The independent variables are listed in the first row. Numbers in parentheses are t-statistics. Year IR RTN LMV SGR RDS Following the IPO 2 0.00 0.04 0.01 *** 0.10 ** 0.00 (0.21) (1.13) (2.74) (2.48) (0.34) 3 -0.06 *** -0.05 0.03 *** 0.01 0.00 (-2.63) (-1.09) (5.16) (0.29) (0.57) 4 -0.08 *** -0.02 0.03 *** 0.03 0.00 ** (-3.28) (-0.39) (4.73) (0.70) (2.02) 5 -0.04 -0.05 0.02 *** 0.01 0.00 (-1.65) (-0.97) (3.74) (0.21) (0.67) Year EP BM DP TECH Following the IPO 2 -0.28 *** 0.08 *** 0.50 *** 0.00 (-4.21) (3.26) (3.69) (-0.39) 3 -0.40 *** 0.03 0.37 ** -0.01 (-5.37) (1.24) (2.51) (-0.66) 4 -0.45 *** 0.06 ** 0.40 *** -0.02 ** (-6.05) (2.28) (2.77) (-2.43) 5 -0.54 *** 0.06 ** 0.28 * -0.03 ** (-6.62) (1.99) (1.76) (-2.22) Year # of obs Adj. [R.sup.2] Following the IPO 2 3070 0.009 3 2758 0.022 4 2456 0.029 5 2199 0.029 *** Significant at the 0.01 level. ** Significant at the 0.05 level. * Significant at the 0.10 level. Table VII. Analyst Forecast Errors for IPO Firms Panel A of this table reports analysts' forecast errors related to initial EPS forecasts made within the first year following the IPO for the first four fiscal years following the IPO (Years 1, 2, 3 and 4). Forecast error is calculated as (Consensus EPS Forecast-Actual EPS)/ (Consensus EPS Forecast). To adjust for industry-wide EPS forecast bias, we calculate the forecast errors for all seasoned firms in the same industry and subtract the median forecast error of the industry from the IPO forecast error. The Wilcoxon p-values are for the two-tailed test about the hypothesis that the medians are not significantly different from zero. Panel B reports the correlation coefficients between underpricing and analyst forecast errors. Underpricing is calculated as (1st Day Closing Price)/(Offer Price)-1. The p-values of the correlation coefficients are reported in parentheses. Panel A. EPS Forecast Errors for IPO Firms Year Forecast Error (%) Following the IPO 25% Median 75% Wilcoxon p value 1 -5.7 1.5 14.9 0.76 2 -6.1 9.7 51.8 0.00 3 1.6 40.5 118.7 0.00 4 9.1 50.1 116.3 0.00 Industry-Adjusted Year Forecast Error (%) Following the IPO 25% Median 75% Wilcoxon # of Obs. p value 1 -9.4 -1.3 13.6 0.13 1208 2 -22.2 -4.2 36.7 0.00 2340 3 -21.3 19.4 98.8 0.00 2156 4 -41.8 5.2 71.3 0.01 641 Panel B. Spearman Correlation Coefficients Forecast Error Variables Year 1 Year 2 Year 3 Year 4 Underpricing -0.17 -0.13 -0.07 -0.11 (0.00) (0.00) (0.00) (0.00) Industry-Adjusted Forecast Error Variables Year 1 Year 2 Year 3 Year 4 Underpricing -0.14 -0.09 -0.06 -0.07 (0.00) (0.00) (0.00) (0.08) Table VIII. Regressions of Analysts' EPS Forecast Errors for IPO Firms This table reports the results of the following regressions: [FE.sub.it] = [[beta].sub.0] + [[beta].sub.1][IR.sub.i] + [[beta].sub.2][RTN.sub.i] + [[beta].sub.3][LMV.sub.i] + [[beta].sub.4][NAN.sub.i] + [[beta].sub.5][AW.sub.i] + [[beta].sub.6][SGR.sub.i] + [[beta].sub.7][RDS.sub.i] + [[beta].sub.8][EP.sub.i] + [[beta].sub.9][BM.sub.i] + [[beta].sub.10][DP.sub.i] + [[beta].sub.11][TECH.sub.i] + [[epsilon].sub.it]. [FE.sub.it] represents industry-adjusted IBES EPS forecast errors. Forecast error is calculated as (Consensus EPS [Forecast.sub.it]-- Actual [EPS.sub.it])/(Consensus EPS [Forecast.sub.it]) where Consensus EPS Forecastle is calculated as the average of EPS forecasts made by different analysts for firm i for the tth fiscal year ending after the IPO. IR is 1st day initial return (underpricing), calculated as (1st day closing price/offer price-1). RTN refers to retention, which is calculated as (Shares outstanding after IPO--Shares offered in IPO)/ (Shares outstanding after IPO). LMV is the log of market value at the end of the first trading day after the IPO. NAN is the number of analysts making EPS forecasts for year t in the first year following the IPO. AW is the average window (measured in months) between the date when the EPS forecast is made and the forecasted fiscal year end. SGR represents the sustainable growth rate, calculated as the product of return on equity and plowback ratio. RDS is the intensity of R&D relative to sales. All these accounting data are collected from Compustat for the first fiscal year after the IPO. EP is the earnings-to-price ratio; BM is the firm's book-to-market value of equity; and DP is the ratio of dividends to price. The multiples are estimated on a per-share basis using price at the end of the first trading day following the IPO and accounting data from the first fiscal year ending after the IPO. TECH is a dummy variable with a value of 1 for a stock in the high-tech industries. High-tech industries are defined as those with three-digit SIC codes of 283, 357, 366, 737, or two-digit SIC codes of 38, 48. The independent variables are listed in the first row. Observations are dropped if they have forecast errors in the top or bottom 1% of observed forecast errors. Numbers in parentheses are t-statistics. Year IR RTN LMV NAN Following the IPO 1 -0.32 *** -0.01 -0.04 ** -0.01 (-3.84) (-0.05) (-1.98) (-1.28) 2 -0.29 ** -0.24 -0.07 ** -0.05 *** (-2.37) (-1.34) (-2.17) (-5.10) 3 -0.35 ** -0.64 *** -0.05 -0.07 *** (-2.32) (-2.75) (-1.22) (-4.67) 4 -0.21 -0.15 -0.28 *** 0.02 (-0.57) (-0.25) (-2.84) (-2.84) Year AW SGR RDS EP Following the IPO 1 0.03 *** 0.00 -0.05 -4.07 *** (5.39) (0.02) (-0.93) (-14.24) 2 0.03 *** -0.39 ** 0.01 -6.07 *** (7.63) (-1.98) (0.23) (-12.05) 3 0.02 ** -1.10 *** 0.04 *** -3.40 *** (2.50) (-5.03) (4.84) (-8.30) 4 0.03 ** -0.90 0.04 -5.63 *** (2.16) (-1.18) (1.71) (-3.37) Year BM DP TECH # of obs Adj. Following [R.sup.2] the IPO 1 0.11 0.12 -0.07 ** 1180 0.182 (1.49) (0.33) -2.27 2 -0.05 -0.71 -0.06 2283 0.153 (-0.44) (-1.19) (-1.27) 3 -0.47 *** -2.18 *** 0.10 2102 0.134 (-3.46) (-2.75) (1.83) 4 -0.61 * 0.81 0.04 625 0.117 (-1.67) (0.35) (0.27) *** Significant at the 0.01 level. ** Significant at the 0.05 level. * Significant at the 0.10 level.

Printer friendly Cite/link Email Feedback | |

Title Annotation: | initial public offering |
---|---|

Author: | Zheng, Steven X.; Stangeland, David A. |

Publication: | Financial Management |

Geographic Code: | 1USA |

Date: | Jun 22, 2007 |

Words: | 11372 |

Previous Article: | Takeover activity as a response to time-varying changes in investment opportunity sets: evidence from takeover sequences. |

Next Article: | Stock price response to calls of convertible bonds: still a puzzle? |

Topics: |