Underwriter reputation, earnings management and the long-run performance of initial public offerings
We thank Chief Editor Robert Faff, Professor Ian Zimmer, and one anonymous reviewer for their challenging and constructive comments on earlier versions of this article. We also greatly appreciate the helpful comments provided by an anonymous reviewer as well as the support of the participants at the 2008 Financial Management Association Conference.
Abstract
This study contributes to the extant literature on the nature of earnings management surrounding initial public offerings (IPOs) by investigating the role of underwriter reputation. We argue that prestigious underwriters will protect their reputation by carefully monitoring and certifying financial information on IPO firms, thereby limiting any potential earnings manipulation. As a result, those IPO firms that are associated with more prestigious underwriters are likely to exhibit substantially less-aggressive earnings management. Conversely, we find the existence of a negative relationship between earnings management and the post-offer performance of an IPO firm’s stocks only for those firms associated with less-prestigious underwriters.
1. Introduction
It is already well documented within the extant literature that a characteristic of ‘initial public offerings’ (IPOs) is the high degree of information asymmetry between issuers and investors (Leland and Pyle, 1977; Carter and Manaster, 1990). Given that investors have limited access to financial information on the issuing firm prior to the IPO, they are heavily reliant upon the prospectus, despite the fact that this provides only limited information (Rao, 1993); as firms go public essentially so that they can gain access to the capital markets, presenting the issuing firms in the best possible light raises the likelihood of a successful offering. Thus, managers have considerable incentives to exercise their discretion, which can lead to the inflation of reported earnings.
Hughes (1986) and Titman and Trueman (1986) provide evidence of a positive relationship between the information provided in financial statements and the offer price of IPOs, while Friedlan (1994) and Teoh et al. (1998a, b) point to a significant increase in discretionary accruals prior to IPO announcements. In those cases where earnings are found to have been over-inflated prior to the announcement of IPOs, this is likely to have detrimental effects on the post-IPO performance of the firms involved through the reversal of these accruals. Consistent with this argument, Teoh et al. (1998a, b) provide evidence of a negative correlation between discretionary accruals and long-run stock performance.
Although numerous studies report patterns of abnormal accruals surrounding IPO issues (Friedlan, 1994; Teoh et al., 1998a, b), the explanations provided for such patterns are rather inconclusive. Beneish (2001) argues that although equity offerings create situations in which managers have strong incentives to manipulate their earnings, any conclusion of intentional earnings manipulation based on changes in discretionary accruals at the time of security issuance may be premature.
Ball and Shivakumar (2008) further question the hypothesis of artificial earnings inflation being used as a means of influencing IPO prices, as documented in the prior studies. They argue that as opposed to intentional earnings manipulation, the increase in discretionary accruals surrounding IPOs could actually be attributable to the firms’ use of their IPO proceeds simply to make appropriate changes to their working capital; thus, as many cash-constrained private firms are likely to exhibit under-investment in both receivables and inventory prior to their listing, any significant increase in discretionary accruals may simply represent the relief of such resource constraints.
Furthermore, managers should actually have far less incentive to manipulate their earnings around the listing date, essentially because this is likely to lead to an increase in the litigation and regulation risks for the managers engaging in such action. Thus, Ball and Shivakumar (2008) argue that much of the prior evidence of significant earnings management surrounding around IPO events could be misleading.
Although quite convincing, the argument of Ball and Shivakumar (2008) fails to explain the significant negative relationship between earning management and post-issue stock performance documented in the extant literature on both IPOs (Teoh et al., 1998b) and ‘seasoned equity offerings’ (SEOs) (Teoh et al., 1998c). If an increase in discretionary accruals is mainly attributable to the alleviation of financial constraints or signalling information, then we would not expect to see earnings management having strong, negative impacts on the subsequent performance of the stocks following such offerings.
We therefore set out in this study to determine the actual nature of earnings management by investigating the effects of underwriter reputation. As a result of the limited information available on IPO firms, it is often difficult for investors to distinguish between changes in abnormal accruals attributable to managers exercising their discretion, and those that are due to exogenous influences from the operational decisions taken by firms (Kaplan, 1985; McNichols and Wilson, 1988; Beneish, 1998).
The services of underwriters are engaged by issuing firms to purchase, market and distribute their securities to investors. It is expected that the due diligence investigations carried out by such underwriters will reflect the actual financial condition of the IPO firms, thereby assisting investors in their assessment of the value of the issuing firm. We therefore argue that underwriter reputation plays an important role in the nature of earnings management surrounding IPOs.
Firstly, when the capital markets are unable to make any clear distinction between the changing nature of a firm’s accruals, prestigious underwriters can provide some added credibility. Investors in the IPO markets are faced with a serious problem of asymmetric information, while the underwriters marketing a firm’s equity have strong incentives to show that the firm’s issued security is worthy of investment, regardless of whether or not they have expended sufficient resources to investigate the firm. Some underwriters may produce a careful, detailed evaluation of the firm’s projects, while also demonstrating good quality accounting disclosure practices, whereas others may compromise their due diligence responsibilities for the sake of attracting business.
Chemmanur and Fulghieri (1994) argue that the more prestigious underwriters will invariably attempt to reduce the probability of marketing ‘lemons’, since such business could ultimately damage the reputation of the company. Such prestigious underwriters will therefore tend to set much stricter standards in their evaluation of firms, and although this may prove costly in the short-run, their primary aim is clearly the pursuit of greater long-term benefits. Thus, in the face of noisy discretionary accruals, underwriter reputation may well be found to play an important role in certifying the actual quality of changes in accruals.
Secondly, underwriters also have a role to play in restricting the extent of earnings manipulation. Prestigious underwriters will closely monitor the quality of the financial information which is being provided by engaging the services of reputable auditors and evaluating the viability of business models to reduce the agency costs between issuers and investors; as a result, they can provide reliable certification of their clients (Jo et al., 2007). Since the act of underwriting involves repeat business with a finite number of competitors, investors can readily engage in the ex-post evaluation of the quality of the underwriters’ services. It would clearly be very costly and difficult for underwriters to market future issues if the investors involved had already been misled by the same underwriters in their prior IPO investments.
Furthermore, the more reputable underwriters must protect the enhanced ‘reputation capital’ that they already possess; they therefore have much stronger incentives to provide quality monitoring in order to enhance the transparency of earnings and prevent any aggressive earnings management. In contrast, low-quality underwriters may compromise their responsibilities towards information monitoring in order to acquire more underwriting business.
Hansen and Torregrosa (1992) show that underwriter monitoring improves a firm’s performance by reducing its agency costs, while Jo et al. (2007) point to a strong negative relationship between earnings management and underwriter reputation. In the present study, we examine the association between underwriter reputation and the nature of earnings management. Specifically, we argue that an increase in accruals within IPO firms that are underwritten by high-quality underwriters is more likely to represent changes in the firms operations and overall performance. Conversely, we suggest that an increase in accruals for IPO firms associated with low-quality underwriters may be the result of opportunistic earnings inflation. Our hypothesis is that discretionary accruals may show significant increases for both high- and low-quality underwriters prior to IPOs, but the nature of such increases may differ markedly.
When changes in accruals are due to intentional earnings inflation, the subsequent reversal in earnings is likely to result in a significant negative relationship between earnings management and post-IPO stock performance. In contrast, when changes in accruals are due to changes in business operations and working capital, we would not expect to find any significant relationship. We therefore hypothesize that the negative association between earnings management and subsequent stock performance will be significant only for IPOs underwritten by low-quality underwriters, and not for those underwritten by high-quality underwriters. We collect a sample of the US IPO firms covering the period from 1989 to 2003 to test our hypotheses.
In this study, we adopt discretionary current accruals as our measure of earnings management, while providing controls for the changes in discretionary accruals attributable to changes in performance. Our analysis also uses performance-adjusted discretionary current accruals as proposed in Kothari et al. (2005). Our results show that significant increases in discretionary current accruals are experienced by firms with both high- and low-quality underwriters during their IPO years; however, when using performance-adjusted discretionary current accruals, the discretionary accruals for the high-quality underwriter group become statistically insignificant, while those for the low-quality group remain significant.
We also find that underwriter reputation has a significantly negative correlation with earnings management, a result which is in line with the evidence on SEOs provided by Jo et al. (2007). Consistent with the prior studies, our results indicate that earnings management has a negative correlation with post-IPO stock performance for the full sample of IPO firms. However, when the interaction effect between earnings management and underwriter reputation is included within the regression model, the effect of earnings management becomes statistically insignificant, whereas the interaction effect is highly significant. These results highlight the importance of considering the effect of underwriter reputation when assessing the relationship between earnings management and post-IPO performance.
Our paper contributes to the literature on earnings management surrounding major corporate events. Although Teoh et al. (1998a–c) show that earnings management has a negative correlation with post-issue performance, Shivakumar (2000) and Fan (2007) could find no strong evidence of such a relationship. The findings of the present study could provide one possible explanation for such inconclusive evidence in the prior studies, thereby indicating the importance of considering underwriter quality in addressing the relationship between earnings management and long-run performance.
Our study is also closely related to the literature on earnings management. McNichols and Wilson (1988) and Beneish (1998) both argue that abnormal accruals could be caused by managers intentionally overstating their reporting figures or by performance changes in firms going public. The present study suggests that these abnormal accruals may have a strong correlation with the choice of underwriter. Managers with an incentive to manipulate earnings could engage low-quality underwriters, essentially because such underwriters are likely to provide a less rigorous monitoring function. Conversely, since high-quality underwriters provide stricter monitoring and inspection of a firm’s information, the abnormal accruals of firms with high-quality underwriters could simply be the result of changes in performance.
The remainder of this paper is organized as follows. The data and sample selection are presented in Section 2, followed in Section 3 by an introduction to our methodology and presentation of the empirical results. Section 4 presents the robustness tests on our results, with concluding remarks subsequently being presented in Section 5.
2. Sample
The initial full sample of the US common stock offerings covering the period from 1989 to 2003 is obtained from the Global New Issues database of the Securities Data Company (SDC). We exclude all SEOs, unit offerings and ADRs, with firms in the financial and utilities industries also being excluded from the sample since they are subject to stricter government regulations and unique disclosure requirements (Teoh et al., 1998c; Shivakumar, 2000; Teoh and Wong, 2002; Fan, 2007; Lee and Masulis, 2008).
To qualify for inclusion in the sample, companies must have made an offer price in excess of US$1.00 with market capitalization of at least US$20 million (in December 1997 purchasing power, following Teoh et al., 1998b); this selection procedure resulted in a sample of 3899 offers, of which 241 were unavailable on the annual COMPUSTAT database and/or the Center for Research in Security Prices (CRSP) file. Furthermore, for inclusion in the sample, the offer must have been made by a company: (i) whose industry has more than 10 firms with the same two-digit SIC code to facilitate the estimation of the expected accruals; and (ii) with sufficient accounting data for the estimation of abnormal accruals amongst the sample companies and for cross-sectional analyses. These restrictions eliminated a further 1002 offers, leaving a final sample of 2053 offers.
The sample selection process is presented in Panel A of Table 1, with Panels B and C of Table 1 presenting the respective sample distribution, by year and industry. Panel B shows that the IPOs were concentrated mainly in the years 1994–1997 (accounting for 50.27 per cent of the sample firms). Panel C reveals that IPOs in the manufacturing industries (two-digit SIC code 20–39) accounted for approximately 44.57 per cent of the total sample, while those in the service industries (two-digit SIC code 70–89) accounted for approximately 33.80 per cent.
| Selection Items | Sub-totals | Total | ||||||
|---|---|---|---|---|---|---|---|---|
| Panel A: sample selection process | ||||||||
| Total no. of IPOs (January 1989–December 2003)a | 6821 | |||||||
| Less unit offerings and ADRs | 1505 | |||||||
| Less firms in the utilities or financial industries | 1133 | |||||||
| Less firms whose offer price <$1 or capitalization <$20 millionb | 284 | |||||||
| Sample IPOs available | 3899 | |||||||
| Less IPOs not available within the annual compustat and CRSP databases | 241 | |||||||
| Less firms missing accounting data for computation of abnormal accruals | 1002 | |||||||
| Less firms missing data included in the regression analyses | 603 | |||||||
| Final sample | 2053 | |||||||
| Panel B: sample distribution by offer year | ||||||||
| Year | 1989 | 1990 | 1991 | 1992 | 1993 | 1994 | 1995 | 1996 |
| Total no. | 53 | 61 | 135 | 124 | 169 | 218 | 241 | 359 |
| % | 2.58 | 2.97 | 6.58 | 6.04 | 8.23 | 10.62 | 11.74 | 17.49 |
| Year | 1997 | 1998 | 1999 | 2000 | 2001 | 2002 | 2003 | Total |
| Total no. | 214 | 121 | 135 | 146 | 44 | 19 | 14 | 2053 |
| % | 10.42 | 5.89 | 6.58 | 7.11 | 2.14 | 0.93 | 0.68 | 100.00 |
| Panel C: sample distribution by industry | ||||||||
| SIC codes | Industry | No. of observations | % | |||||
| 1∼9 | Agriculture, forestry and fishing | 3 | 0.15 | |||||
| 10∼14 | Mining | 50 | 2.44 | |||||
| 15∼17 | Construction | 23 | 1.12 | |||||
| 20∼39 | Manufacturing | 915 | 44.57 | |||||
| 40∼48 | Transportation and communications | 104 | 5.07 | |||||
| 50∼51 | Wholesale trade | 77 | 3.75 | |||||
| 52∼59 | Retail trade | 181 | 8.82 | |||||
| 70∼89 | Services | 694 | 33.80 | |||||
| 91∼99 | Public administration | 6 | 0.29 | |||||
| Totals | 2053 | 100.00 | ||||||
- aObtained from the SDC database. bBased upon 1997 purchasing power.
The summary statistics of our sample firms are presented in Table 2, which shows that the IPO firms issued an average of 3.83 million shares, and raised average proceeds of US$55.14 million, with about 46 per cent of the IPOs in our sample involving venture capital. The mean (median) level of under-pricing in the offer price is found to be 21 per cent (10 per cent). The underwriter ranking used in this study is the Loughran and Ritter (2004) measure of underwriter reputation; these underwriter prestige measures are ranked on a 0–9 scale, and are based on the pecking order seen in ‘tombstone’ advertisements.1
| Variables | Mean | SD | Q1 | Median | Q3 |
|---|---|---|---|---|---|
| Proceeds (US$ millions) | 55.14 | 105.42 | 20.00 | 33.60 | 56.10 |
| Offer price (US$) | 12.51 | 4.45 | 9.00 | 12.00 | 15.00 |
| Shares offered (millions) | 3.83 | 5.28 | 2.00 | 2.70 | 4.00 |
| Venture backed | 0.46 | 0.50 | 0.00 | 0.00 | 1.00 |
| Underprice | 0.21 | 0.37 | 0.02 | 0.10 | 0.26 |
| Underwriter | 7.50 | 1.90 | 7.00 | 8.00 | 9.00 |
| Age (years) | 15.30 | 18.84 | 5.00 | 9.00 | 17.00 |
| BV (US$ millions) | 81 | 261 | 19 | 37 | 70 |
| MV (US$ millions) | 351 | 1064 | 69 | 130 | 314 |
| B/M | 0.36 | 0.38 | 0.15 | 0.27 | 0.45 |
| TA (US$ millions) | 183 | 741 | 32 | 60 | 126 |
- The sample comprises of a total of 2053 IPOs which took place between 1989 and 2003 and which met the data requirements described in section 2. The proceeds, offer price, shares offered and venture backed variables are obtained from the SDC database; Underprice is defined as the ratio of the offer price to the closing price minus the offer price on the post-issue date; Underwriter refers to the ranking of the firm’s underwriter using the ‘reputation rankings for IPO underwriters’ obtained from the Jay Ritter website; Age is the number of years between the founding year or incorporation date (whichever is earlier) and the IPO year; BV is the book value of common equity at the end of fiscal year 0; MV is the market value of common equity at the end of fiscal year 0; B/M is the book to market ratio of common equity at the end of fiscal year 0; and TA is the book value of total assets at the end of fiscal year 0.
For our empirical analysis, if there is more than one leading underwriter, then we use the rank of the ‘bookrunner’ or the highest joint-ranking ‘bookrunners’. The mean (median) age for our sample of IPO firms is 15.31 (9.00), while the mean (median) underwriter reputation ranking is 7.50 (8.00).2 The sample firms had a mean (median) common equity book value of US$81 million (US$37 million) in the IPO year, and a mean (median) book-to-market (B/M) ratio of 0.36 (0.27).3
3. Methodology and empirical results
Following the timing convention of Teoh et al. (1998b), we define year 0 as the fiscal year in which the IPO occurred. Although Teoh et al. (1998b) found evidence of earnings management for equity-issuing firms in both year 0, securities prices may react negatively with the reversal of managed accruals in the post-offering period. We therefore examine post-offering returns over the 756-trading-day period starting from the day immediately after the earnings announcement in year 0. If announcement date of financial report is unavailable from the COMPUSTAT file immediately, then the post-offering returns are calculated starting from the first trading day of the fourth month after the end of fiscal year 0.
3.1. Earnings management
The accounting of earnings involves consideration of both operating cash flows and accruals; however, given that managers have considerable discretion in their accounting of accruals, such accruals are susceptible to earnings management. An apparent increase in earnings can be accomplished by the early recognition of revenues, the delayed recognition of expenses, or a combination of both; however, not all accrual items are equally susceptible to managerial manipulation.
Short-term accruals, which are accounting adjustments to short-term assets and liabilities (such as bad debt allowances for accounts receivable and warranty liabilities) are easier to manipulate. We therefore adopt discretionary current accruals (DCA) and performance-adjusted discretionary current accruals (ADCA) variables to measure the extent of earnings management.
3.1.1. Discretionary current accruals
We follow the methodology of Teoh et al. (1998b, c) to measure the expected current accruals from a modified Jones (1991) model. As in Teoh et al. (1998b, c), total current accruals are broken down into their two constituent components, non-discretionary current accruals and discretionary current accruals. The former are the asset-scaled proxies for unmanaged accruals, while the latter are the asset-scaled proxies for earnings that are managed at the discretion of managers.
3.1.2. Performance-adjusted discretionary current accruals
Kothari et al. (2005) argue that the models of discretionary accruals may be specified in a severely incorrect manner when being applied to samples of firms demonstrating abnormal performance. We also employ the Kothari et al. (2005) ADCA approach in the present study, subtracting the discretionary accruals of a set of industry and return-on-assets (ROA) matched control firms prior to the IPO; thus each IPO issuer is matched with a non-IPO firm from the same industry based upon the nearest ROA for the previous fiscal year and the same SIC code.4 The ADCA for an IPO issuer is therefore calculated as the DCA of the IPO issuer minus the DCA of the non-IPO matched firm.
Details on the time-series distribution of DCA for the full sample, from year 0 to year 3, are presented in Table 3, showing an average increase of 4.8 per cent in DCA in the IPO year, followed by a decline of 1.1 per cent from the baseline assets from year 2 to year 3. The time series pattern for ADCA also exhibits a similar peak in the issue year, with a discernible steady decline thereafter. If, as claimed by Ball and Shivakumar (2008), any increase in DCA is thought to be completely attributable to changes in business operations and working capital, then we would not expect to observe any clear reversing trend in accruals in subsequent years.
| Time horizons | No. of observations | Mean | t-testc | Medianc |
|---|---|---|---|---|
| DCA b | ||||
| Initial offer year | 2053 | 0.048 | 9.46*** | 0.0210*** |
| Initial offer year +1 | 1983 | 0.024 | 7.57*** | 0.0110*** |
| Initial offer year +2 | 1793 | 0.010 | 3.37*** | 0.0050*** |
| Initial offer year +3 | 1600 | −0.001 | −0.32 | −0.0003 |
| Common sample | ||||
| Initial offer year | 1600 | 0.047 | 8.18*** | 0.023*** |
| Initial offer year +1 | 1600 | 0.025 | 7.16*** | 0.014*** |
| Initial offer year +2 | 1600 | 0.011 | 3.61*** | 0.005*** |
| Initial offer year +3 | 1600 | −0.001 | −0.32 | −0.0003 |
| ADCA b | ||||
| Initial offer year | 2053 | 0.031 | 4.95*** | 0.020*** |
| Initial offer year +1 | 1983 | 0.021 | 4.96*** | 0.018*** |
| Initial offer year +2 | 1793 | 0.005 | 1.13 | 0.009 |
| Initial offer year +3 | 1600 | −0.010 | −2.26** | 0.003** |
| Common sample | ||||
| Initial offer year | 1600 | 0.028 | 3.92*** | 0.020*** |
| Initial offer year +1 | 1600 | 0.020 | 4.11*** | 0.018*** |
| Initial offer year +2 | 1600 | 0.007 | 1.51 | 0.009** |
| Initial offer year +3 | 1600 | −0.010 | −2.26** | 0.003 |
- aThe sample comprises of a total of 2053 IPOs which took place between 1989 and 2003. Discretionary current accruals (DCA) and performance-adjusted discretionary current accruals (ADCA) are shown for firms offering initial equity, from the initial offer year to 1, 2 and 3-year periods after the offer. bDCA are extracted from current accruals using a two-digit SIC code cross-sectional, modified Jones (1991) model; ADCA are constructed by matching each treatment firm with a control firm based on industry and return on assets in period t − 1. cThe p-value refers to the t-statistic and the Wilcoxon z-statistic is adopted to test the median value. ***indicates significance at the 1 per cent level; and **indicates significance at the 5 per cent level.
The evidence in Table 3 suggests that, for at least some proportion of the sample firms, the increase in DCA is due to opportunistic earnings management, with managers advancing accruals to report higher operational performance in the issuing period, and a gradual reversal in DCA in subsequent years. Since the sample composition changes each year as a result of data availability, Table 3 also presents the results from a common sample of firms for which accruals data is available from year 0 to year 3 to test whether the findings are affected by the changing sample size. We find that the common sample results are the same as those for the full sample.5
3.2. The underwriter reputation and earnings management relationship
To examine whether the more prestigious underwriters succeed in suppressing potential earnings management by IPO firms, we divide the firms into four sub-groups based upon the underwriter reputation ranking; the G1 group comprises of firms whose underwriters have an index of 9, those with an index of 8 form the G2 group, those ranked 6∼8 form the G3 group, and those ranked <6 form the G4 group. As shown in Panel A of Table 4, the mean DCA in year 0 for all sub-groups is positive and significant. Furthermore, with a decline in underwriter reputation, the results reveal a strong pattern of a monotonic increase in DCA. The mean DCA is 2.2 per cent for the G1 group and 13.8 per cent for the G4 group, with this difference between the G1 and G4 groups (−11.6 per cent) being statistically significant at the 1 per cent level.6
| Variables | Year 0 | Year 0∼Year 1 | Year 0∼Year 2 | Year 0∼Year 3 |
|---|---|---|---|---|
| Panel A: DCA | ||||
| G1: UR = 9.0 | ||||
| No. of observations | 776 | 748 | 667 | 591 |
| Mean | 0.022 | −0.006 | −0.013 | −0.013 |
| t-test | 3.03*** | −0.69 | −1.64 | −1.47 |
| G2: UR = 8.0 | ||||
| No. of observations | 613 | 592 | 537 | 486 |
| Mean | 0.024 | −0.012 | −0.011 | −0.018 |
| t-test | 2.51** | −1.13 | −0.94 | −1.57 |
| G3: 6.0 ≤ UR < 8.0 | ||||
| No. of observations | 365 | 358 | 321 | 290 |
| Mean | 0.071 | −0.021 | −0.057 | −0.061 |
| t-test | 6.13*** | −1.57 | −3.95*** | −4.17*** |
| G4: UR < 6.0 | ||||
| No. of observations | 299 | 285 | 258 | 233 |
| Mean | 0.138 | −0.087 | −0.116 | −0.139 |
| t-test | 8.85*** | −4.38*** | −5.73*** | −6.96*** |
| Difference | ||||
| Mean | −0.116 | 0.082 | 0.102 | 0.126 |
| t-test | −6.75*** | 3.77*** | 4.69*** | 5.81*** |
| Panel B: ADCA | ||||
| G1: UR = 9.0 | ||||
| No. of observations | 776 | 748 | 667 | 591 |
| Mean | 0.010 | 0.008 | −0.012 | −0.019 |
| t-test | 1.06 | 0.73 | −1.07 | −1.54 |
| G2: UR = 8.0 | ||||
| No. of observations | 613 | 592 | 537 | 486 |
| Mean | 0.000 | 0.001 | 0.004 | −0.011 |
| t-test | 0.00 | 0.09 | 0.48 | −0.65 |
| G3: 6.0 ≤ UR < 8.0 | ||||
| No. of observations | 365 | 358 | 321 | 290 |
| Mean | 0.057 | −0.014 | −0.058 | −0.053 |
| t-test | 3.91*** | −0.87 | −3.16*** | −2.92*** |
| G4: UR < 6.0 | ||||
| No. of observations | 299 | 285 | 258 | 233 |
| Mean | 0.120 | −0.064 | −0.088 | −0.125 |
| t-test | 6.52*** | −2.72*** | −3.76*** | −5.09*** |
| Difference | ||||
| Mean | −0.111 | 0.072 | 0.076 | 0.106 |
| t-test | −5.36*** | 2.77*** | 2.89*** | 3.86*** |
- DCA are extracted from current accruals using a two-digit SIC code cross-sectional, modified Jones (1991) model; ADCA are constructed by matching each treatment firm with a control firm based on industry and return on assets in period t − 1. UR is the ranking of the IPO firm’s underwriter. IPOs are divided into (i) UR = 9.0; (ii) UR = 8.0; (iii) 6.0 ≤ UR < 8.0; and (iv) UR < 6.0. The p-value refers to the t-statistic. ***indicates significance at the 1 per cent level; **indicates significance at the 5 per cent level; and *indicates significance at the 10 per cent level.
Our results further indicate that while there are declines in accruals for all four groups of underwriter rankings in the years following the IPO, the post-IPO changes in accruals are statistically and insignificantly different from 0 for the G1 and G2 groups, but statistically significant for the G3 and G4 groups. From a comparison of the changes in accruals between the G1 and G4 groups, it is clear that G4 firms experience significantly greater declines in accruals than G1 firms.
We next provide controls for changes in discretionary accruals induced by performance. While Table 4 presents similar evidence for ADCA and DCA, we find that the ADCA for the G1 and G2 groups is insignificantly different from 0. A comparison of the results for DCA (Panel A) with those for ADCA (Panel B) suggests that for those groups with the prestigious underwriter, the increase in discretionary accruals may be attributable to changes in the firm’s performance, as argued by Ball and Shivakumar (2008). Thus, the evidence from Table 4 provides support for our hypothesis of a strong correlation between underwriter reputation and the extent of earnings management in IPO firms.
()Kim and Park (2005) demonstrate an association between earnings management and IPO under-pricing; we therefore include the variable, Underprice, measured as the ratio between the offer price, and the closing price minus the offer price on the post-issue date. Given that the leverage ratio of a firm could also affect accruals if managers were to attempt to avoid debt contract restrictions (Morsfield and Tan, 2006), we also include the variable, Leverage, measured within the model as the ratio of long-term debt to total assets.
AccSales is defined as the total accruals per dollar of sales in the issue year (Fan, 2007), and Big6 indicates the reputation of the auditor, which is equal to 1 if the firm’s auditor is one of the ‘big six’ auditing firms (Fan, 2007; Jo et al., 2007). Both AccSales and Big6 measure the flexibility of managers with regard to earnings manipulation. CapExp is the capital expenditure in year 0 scaled by the total assets at the issue year; Log(1 + Age) is the natural logarithm of 1 plus the age of the firm; and HighTech is a dummy variable which is equal to 1 if the IPO firm is in the high-tech industry, otherwise 0. CapExp, Log(1 + Age) and HighTech measure growth opportunities.
Morsfield and Tan (2006) demonstrate that venture-backed IPOs have less earnings management; we therefore include the variable, Venture Backed, which is equal to 1 if the IPO firm is supported by venture capital, otherwise 0. Kothari et al. (2005) also show that operating performance is associated with the magnitude of discretionary accruals; thus, we include two performance-related variables in the regression models. OCFA is cash flow from operating activities, scaled by assets at the start of the IPO year, and ΔROA is the change in ROA from year 0 to year 1, calculated as income before extraordinary items, scaled by total assets at the start of year 0. We use year dummies to control for the year effect.
The cross-sectional regression results on the relationship between underwriter reputation and the level of earnings management are presented in Table 5, measured by DCA (Model 1) and ADCA (Model 2). Model 1 tests the effects of underwriter reputation on DCA while controlling for other factors, from which we find that the coefficient of underwriter reputation is significantly positive. Model 2 presents the results of ADCA, with the findings being very similar to those shown in Model 1.
| Variablesc | Model 1 (DCA)b | Model 2 (ADCA)b | Model 3 (DCA)b | Model 4 (ADCA)b | ||||
|---|---|---|---|---|---|---|---|---|
| Coeff. | t-Statisticd | Coeff. | t-Statisticd | Coeff. | t-Statisticd | Coeff. | t-Statisticd | |
| Intercept | 0.117 | 2.96*** | 0.126 | 2.53** | 0.135 | 3.55*** | 0.147 | 3.12*** |
| URD | 0.050 | 4.57*** | 0.053 | 3.76*** | 0.041 | 3.86*** | 0.045 | 3.37*** |
| Underprice | −0.002 | −0.16 | −0.028 | −1.51 | −0.019 | −1.28 | −0.032 | −1.74* |
| Leverage | 0.003 | 0.09 | 0.021 | 0.52 | −0.022 | −0.73 | −0.003 | 0.08 |
| AccSales | 0.001 | 3.73*** | 0.001 | 2.97*** | 0.003 | 5.56*** | 0.001 | 3.31*** |
| Big6 | −0.027 | −1.28 | −0.053 | −1.99** | −0.037 | −1.83* | −0.056 | −2.23** |
| CapExp | −0.106 | −2.08** | −0.087 | −1.35 | −0.107 | −2.16** | −0.058 | −0.95 |
| Log(1 + Age) | 0.004 | 0.66 | 0.004 | 0.61 | 0.004 | 0.70 | 0.002 | 0.37 |
| HighTech | −0.026 | −2.25** | −0.035 | −2.43** | −0.024 | −2.24** | −0.039 | −2.85*** |
| Venture backed | −0.041 | −3.75*** | −0.022 | −1.62 | −0.043 | −4.12*** | −0.032 | −2.50** |
| OCFA | −0.149 | −14.67*** | −0.161 | −12.50*** | −0.186 | −17.73*** | −0.214 | −16.16*** |
| ΔROA | −0.007 | −3.94*** | −0.009 | −3.79*** | −0.033 | −6.71*** | −0.013 | −3.30*** |
| Year dummy | Yes | Yes | Yes | Yes | ||||
| No. of observations | 2053 | 2053 | 2032 | 2032 | ||||
| Adjusted R2 | 0.134 | 0.098 | 0.174 | 0.146 | ||||
- aThe sample comprises of 2053 IPOs which took place during the period from 1989 to 2003, on which the following regression is conducted:
-
- bThe dependent variables are discretionary current accruals (DCA) or performance-adjusted discretionary current accrual (ADCA) in the offering year. DCA are extracted from current accruals using a two-digit SIC code cross-sectional, modified Jones (1991) model; ADCA are constructed by matching each treatment firm with a control firm based on industry and return on assets in period t − 1. cURD is a dummy variable which is equal to 1 for a ‘lower’ underwriter reputation, otherwise 0 (firms are placed into the ‘lower’ category if the reputation of the underwriter is below 8); Underprice is defined as the difference between the offer price and the closing price on the first day of trading; Leverage is the ratio of long-term debt to total assets in the IPO year; AccSales is total accruals per dollar of sales in the issuing year; Big6 indicates the auditor reputation, which is equal to 1 if the auditor is one of the ‘big six’ auditors; CapExp is the capital expenditure in year 0 scaled by the total assets in the issuing year; Log(1 + Age) is the natural logarithm of 1 plus the age of the IPO firm; HighTech is equal to 1 if the IPO firm is in the high tech industry, otherwise 0; Venture Backed is equal to 1 if the IPO firm is backed by venture capital; OCFA is cash flow from operating activities scaled by assets at the beginning of the IPO year; ROA is income before extraordinary items scaled by total assets at the beginning of year 0; ΔROA measures the ROA in year 1 minus the ROA in year 0. d***indicates significance at the 1 per cent level; **indicates significance at the 5 per cent level; and *indicates significance at the 10 per cent level.
We calculate the Cook D-influence statistic to exclude the most influential 1 per cent of the observations. The results are reported in Table 5 (Models 3 and 4), from which we can see that the URD coefficients still have significantly positive correlations with DCA and ADCA. The results in Table 5 again provide support for the hypothesis that the practice of earnings manipulation by IPO firms is reduced by more prestigious underwriters.
The results for the control variables show that total accruals per dollar of sales (AccSales), ‘Big-6’ IPO auditing firms (Big6), the high-tech industry (HighTech), venture-backed IPO firm (Venture Backed), cash flow from operating activities (OCFA) and the change in return on assets (ΔROA) are important factors in explaining accruals. The significantly positive coefficient of AccSales is consistent with Fan (2007) and Jo et al. (2007), who argue that IPO firms with higher flexibility with regard to the potential manipulation of earnings incur more DCA or ADCA. The negative coefficients of HighTech and CapExp suggest that firms with higher investment opportunities adopt more conservative reporting strategies since growth firms will be less inclined to manipulate earnings to convey valuable information to shareholders (Jo et al., 2007).
The negative effects of the Big6 and Venture Backed variables suggest that the monitoring effect from auditors and venture capitalists is effective in reducing earnings management (Morsfield and Tan, 2006). Both the OCFA and ΔROA variables are found to have significantly negative associations with accruals, a result which is consistent with those reported by Kothari et al. (2005), who proposed that operating performance was associated with the magnitude of discretionary accruals.
3.3. Long-run stock performance following initial public offerings
If underwriter reputation is indeed an important factor in the monitoring of financial information and in mitigating opportunistic earnings management, then we argue that for those IPOs that were underwritten by high-quality underwriters, any increase in DCA would mainly reflect changes in their working capital, and that there should be no significant relationship between the change in accruals and the post-IPO performance of the firm’s stocks. In contrast, for those IPO firms that are associated with low-quality underwriters, the increase in accruals is more likely to be attributable to opportunistic earnings management; thus, any increase in discretionary accruals is expected to have a negative association with stock performance in the years following the IPOs. In order to test this hypothesis, we examine the buy-and-hold abnormal returns (BHARs) for the years following the IPO year.
3.3.1. Buy-and-hold abnormal returns (BHARs)
Our estimate of BHARs follows the matched-sample approach, with out selection of the matching firms being based on their industry, book-to-market ratio and market value of equity. Each IPO sample is matched with a non-issuing firm, the characteristics of which provide the closest match to those of the sample firm. If the original matched firm drops out before the IPO firm, then the next best match to the original match is substituted for the remainder of the period in order to avoid any survivorship bias within the matched sample. If the sample firm drops out, both the sample firm and the matching firm are assigned zero returns for the remainder period.
We compute the BHARs from the first day after the announcement of the firm’s annual financial statement in the issuing year. This computation continues over the period of 756 trading days or until the sample firm is delisted, whichever comes first. The BHAR is the buy-and-hold returns of sample firm minus the buy-and-hold returns of matching firm.
We divide our sample IPO firms into two groups of relatively equal size based upon DCA and ADCA in Table 6, with Panel A showing that the mean BHAR for firms with high earnings management is −31.27 per cent, significantly lower than the −8.69 per cent for firms with low earnings management.8 Panel A also reveals similar results when earnings management is measured using ADCA.9 Consistent with Teoh et al. (1998b), the results show a negative correlation between earnings management and the post-IPO stock performance.
| Variablesb | No. of observations | % | t-testc | Difference | |
|---|---|---|---|---|---|
| % | t-testc | ||||
| Panel A: BHAR based on earnings management (EM) measures | |||||
| DCA | |||||
| High EM (DCA ≥ 0.021) | 1026 | −31.27 | −4.18*** | −22.58 | −2.41** |
| Low EM (DCA < 0.021) | 1027 | −8.69 | −1.54 | ||
| ADCA | |||||
| High EM (ADCA ≥ 0.020) | 1026 | −28.75 | −3.82*** | −17.53 | −1.87* |
| Low EM (ADCA < 0.020) | 1027 | −11.22 | −2.00** | ||
| Panel B: BHAR based on underwriter reputation (UR) and earnings management (EM) measures | |||||
| DCA | |||||
| High-quality UR (UR ≥ 8.0) | |||||
| High EM (DCA ≥ 0.021) | 629 | −20.38 | −2.70*** | −8.90 | −0.89 |
| Low EM (DCA < 0.021) | 760 | −11.48 | −1.75* | ||
| Low-quality UR (UR < 8.0) | |||||
| High EM (DCA ≥ 0.021) | 397 | −48.53 | −3.20*** | −47.77 | −2.54** |
| Low EM (DCA < 0.021) | 267 | −0.76 | −0.07 | ||
| ADCA | |||||
| High-quality UR (UR ≥ 8.0) | |||||
| High EM (ADCA ≥ 0.020) | 646 | −14.91 | −2.02** | −1.11 | −0.11 |
| Low EM (ADCA < 0.020) | 743 | −16.02 | −2.40** | ||
| Low-quality UR (UR < 8.0) | |||||
| High EM (ADCA ≥ 0.020) | 380 | −52.26 | −3.28*** | −53.62 | −2.83*** |
| Low EM (ADCA < 0.020) | 284 | 1.36 | 0.13 | ||
- aThe sample comprises of 2053 IPOs which took place during the period from 1989 to 2003. Panel A shows the buy-and-hold abnormal returns (BHAR) by issuing DCA or ADCA groups following the announcement date of the first post-issue financial statement (these are the three-year buy-and-hold returns for the sample IPO firms relative to those of industry size and book-to-market (B/M) matched firms). The firms are classified into the ‘High’ earnings management group if the DCA (ADCA) are greater than the median DCA (ADCA), and the ‘Low’ earnings management group if DCA (ADCA) is smaller than the median DCA (ADCA). Panel B shows the BHAR based upon underwriter reputation (UR) and earnings management (EM) measures. Firms are placed into the low (high) quality underwriter category if the underwriter reputation is <8 (≥8). bDiscretionary current accruals (DCA) are extracted from current accruals using a two-digit SIC code cross-sectional, modified Jones (1991) model; Performance-adjusted discretionary current accruals (ADCA) are constructed by matching each treatment firm with a control firm based on industry and return on assets in period t − 1; and Underwriter Reputation (UR) refers to the ranking of the firm’s underwriter using the ‘Reputation Rankings for IPO Underwriters’ obtained from the Jay Ritter website. c***indicates significance at the 1 per cent level; **indicates significance at the 5 per cent level; and *indicates significance at the 10 per cent level.
To consider the effect of underwriter reputation, the sample is partitioned into a 2 × 2 matrix (Panel B of Table 6), based upon the median underwriter reputation value, as well as the earnings management measures (median DCA = 0.021; median ADCA = 0.020). Following Carter et al. (1998) and Loughran and Ritter (2004), we define high-quality underwriters as those with rankings greater than the median (8) for the full sample. We find that for those IPOs issued by low-quality underwriters, the high-DCA group has a mean BHAR of −48.53 per cent, while for the low-DCA group, at −0.76 per cent, the mean difference in BHAR is −47.77 per cent, with statistical significance at the 5 per cent level. In contrast, for IPOs underwritten by high-quality underwriters, there is no significant difference in BHAR between high- and low-DCA groups. The patterns for ADCA are very similar to those for DCA, with these results suggesting that earnings management does not result in declining long-term performance for those IPO firms engaging the services of more prestigious underwriters.
()Our hypothesis predicts that the involvement of lower-quality underwriters in the issuing of IPOs will lead to a stronger negative relationship between earnings management and the post-IPO performance of the firm’s stocks. We include the interaction term between DCA, URD and the other control variables in Model 1 of Table 7, with the results showing that the coefficient associated with the interaction effect is significantly negative, thereby once again suggesting that underwriter reputation is important with regard to the effect on the relationship between earnings management and long-term post-IPO stock performance. Furthermore, when we replace the earnings management measure with ADCA in Model 2, the evidence remains very similar.10 These findings provide strong support for our hypothesis that underwriter reputation has an important moderating effect on the relationship between earnings management and long-run stock performance.11
| Variablesb | Model 1 (DCA) | Model 2 (DCA) | Model 3 (ADCA) | Model 4 (ADCA) | ||||
|---|---|---|---|---|---|---|---|---|
| Coeff. | t-Statisticc | Coeff. | t-Statisticc | Coeff. | t-Statisticc | Coeff. | t-Statisticc | |
| Intercept | 0.083 | 0.31 | 0.054 | 0.20 | 0.155 | 0.60 | 0.103 | 0.40 |
| EM | −0.275 | −1.64 | −0.036 | −0.27 | −0.257 | −1.60 | −0.034 | −0.27 |
| URD | −0.039 | −0.49 | −0.047 | −0.60 | −0.069 | −0.89 | −0.059 | −0.77 |
| EM × URD | −0.525 | −1.94* | −0.647 | −2.94*** | −0.428 | −1.64* | −0.548 | −2.56*** |
| Underprice | −0.273 | −2.88*** | −0.276 | −2.91*** | −0.328 | −3.58*** | −0.341 | −3.69*** |
| Log(1 + Age) | 0.038 | 1.11 | 0.037 | 1.06 | 0.038 | 1.14 | 0.038 | 1.15 |
| Venture backed | 0.042 | 0.64 | 0.052 | 0.77 | 0.011 | 0.17 | 0.044 | 0.68 |
| ΔCapExp | 0.869 | 2.57*** | 0.863 | 2.55** | 0.761 | 2.16** | 0.806 | 2.26** |
| ΔNI | 0.003 | 2.06** | 0.003 | 2.04** | 0.001 | 0.28 | 0.003 | 0.14 |
| Log(MV) | −0.019 | −0.52 | −0.015 | −0.43 | −0.031 | −0.90 | −0.024 | −0.69 |
| Log(B/M) | 0.098 | 2.41** | 0.095 | 2.32** | 0.082 | 2.08** | 0.084 | 2.10** |
| ΔROA | 0.428 | 4.55*** | 0.435 | 4.62*** | 0.740 | 6.79*** | 0.752 | 6.86*** |
| HighTech | 0.148 | 2.15** | 0.147 | 2.13** | 0.161 | 2.42** | 0.147 | 2.20** |
| Oil | −0.113 | −0.53 | −0.105 | −0.49 | −0.152 | −0.73 | −0.144 | −0.68 |
| Year dummy | Yes | Yes | Yes | Yes | ||||
| No. of observations | 2053 | 2053 | 2032 | 2032 | ||||
| Adjusted R2 | 0.044 | 0.043 | 0.057 | 0.057 | ||||
- aThe sample comprises of 2053 IPOs which took place during the period from 1989 to 2003, on which the following cross-sectional regression is conducted:
-
- b LBHAR is (Log[1 + IPO firm 3-year buy-and-hold return] − Log[1 + Matching firm 3-year buy-and-hold return]). The 3-year buy-and-hold return is computed over a period of 756 trading days from the announcement date of the first post-issue financial statement; EM is discretionary current accruals (DCA) or performance-adjusted discretionary current accruals (ADCA); URD is a dummy variable, which is equal to 1 for a ‘lower’ underwriter reputation, otherwise 0 (firms are placed into the ‘lower’ category if the reputation of the underwriter is <8.0; Underprice is defined as the difference between the offer price and the closing price on the first day of trading; Log(1 + Age) is the natural logarithm of 1 plus the age of the firm; Venture Backed is equal to 1 if the IPO firm has venture capital backing, otherwise 0; ΔCapExp measures the asset-scaled mean capital expenditure in years 1, 2 and 3, less the asset-scaled mean capital expenditure in years –1 and 0. ΔNI is net income growth in the IPO year; Log (MV) is the natural logarithm of the market value of the common equity of the IPO firm; Log(B/M) is the natural logarithm of book value divided by market value; ΔROA is the summary of the change in return on assets in years 1 to 3; High Tech and Oil are dummies for the high-tech and oil industries. The year dummy refers to the offer year. c***indicates significance at the 1 per cent level; **indicates significance at the 5 per cent level; and *indicates significance at the 10 per cent level.
Models 3 and 4 report the results based upon a regression analysis computed according to the Cook D-influence statistic, which excludes the most influential 1 per cent of observations. The coefficients on DCA × URD and ADCA × URD still have a significantly negative correlation with matching firm-adjusted buy-and-hold returns. Our findings also suggest that the negative effect of earnings management on long-run performance found in the prior studies may be largely driven by those IPOs with lower-quality underwriters, because the nature of earnings management in such cases is more related to earnings manipulation.
As regards the control variables, we find that the long-run stock performance has a significantly negative correlation with Underprice, and a significantly positive correlation with ΔCapExp, HighTech, Log(B/M) and ΔROA. The result for the Underprice variable is consistent with the overreaction hypothesis, which posits that IPO firms with higher first-day returns will have lower post-IPO performance (Ritter, 1991). The positive effects of ΔCapExp and HighTech both suggest that those firms with more opportunities will demonstrate better post-IPO performance.
The positive association between the B/M ratio and post-IPO stock performance is consistent with Fama and French (1992), Loughran and Ritter (1995) and Spiess and Affleck-Graves (1995, 1999). The results on the change in return-on-assets (ΔROA) and the change in net income (ΔNI) also suggest that improvement in business operations is an important factor in the evaluation of IPO firms by investors. The results for the other control variables although not statistically significant at all conventional levels, are generally consistent with the theoretical predictions.
4. Tests for robustness
We conduct several additional analyses to test the robustness of our findings. Firstly, in addition to the matched-sample approach, as noted by Barber and Lyon (1997), several studies have used the calendar-time portfolio approach to measure long-run stock performance. In order to test whether our conclusions hold for different measures of long-run performance, we follow Brav and Gompers (1997) to present the annual buy-and-hold abnormal returns under the calendar-time portfolio approach. Specifically, we divide the sample into four sub-groups based upon the median value of underwriter reputation (8.0) and the earnings management measure for DCA (0.021).
We calculate the monthly returns for portfolios taking up a long position in equal amounts of all IPO firms that went public within the previous 3 years, and then calculate the annual return by compounding the monthly returns of the IPO portfolios starting in January and ending in December of each year. The Matching Firm-adjusted Calendar Time Returns are measured by the calendar-time returns of the IPO portfolio minus the calendar-time returns of the matching portfolio. The results for the Wealth Relative, which is computed as 1 plus the Matching Firm-adjusted Calendar Time Returns of the high earnings management group divided by that of the low earnings management group, are shown in Table 8.
| Year | Low-quality underwriter reputation (UR < 8.0) | Wealth relativesb | High-quality underwriter reputation (UR ≥ 8.0) | Wealth relativesb | ||
|---|---|---|---|---|---|---|
| High EM | Low EM | High EM | Low EM | |||
| 1990 | −88.40 | −23.58 | 0.15 | 22.12 | 0.27 | 1.22 |
| 1991 | −46.32 | −13.44 | 0.62 | −21.03 | 14.87 | 0.69 |
| 1992 | −53.62 | −9.74 | 0.51 | −21.42 | −6.69 | 0.84 |
| 1993 | −50.94 | −16.58 | 0.59 | −0.44 | −8.07 | 1.08 |
| 1994 | −36.20 | −14.95 | 0.75 | −12.03 | −2.03 | 0.90 |
| 1995 | −52.99 | 3.43 | 0.45 | 3.74 | 9.68 | 0.95 |
| 1996 | −48.04 | 9.68 | 0.47 | −7.96 | 9.08 | 0.84 |
| 1997 | −42.66 | −20.24 | 0.72 | −11.78 | −9.09 | 0.97 |
| 1998 | −48.58 | 0.82 | 0.51 | 8.44 | 18.02 | 0.92 |
| 1999 | −69.93 | −32.52 | 0.45 | −17.35 | −16.91 | 0.99 |
| 2000 | −29.99 | 9.32 | 0.64 | −5.34 | 9.28 | 0.87 |
| 2001 | −52.14 | 0.92 | 0.47 | −4.24 | 0.61 | 0.95 |
| 2002 | −75.39 | −7.33 | 0.27 | −16.94 | −3.16 | 0.86 |
| 2003 | −50.19 | −19.82 | 0.62 | 6.90 | 13.37 | 0.94 |
| 2004 | −28.33 | 7.88 | 0.66 | −1.72 | 0.04 | 0.98 |
| 2005 | −23.98 | −41.00 | 1.29 | −29.67 | −5.49 | 0.74 |
| 2006 | −55.46 | −17.15 | 0.54 | −8.05 | 21.23 | 0.76 |
| 2007 | −21.95 | −8.77 | 0.86 | −1.05 | −5.10 | 1.04 |
| Mean | −48.62 | −10.73 | 0.59 | −6.55 | 2.22 | 0.92 |
- aThis table shows the percentage of annual buy-and-hold adjusted returns of the matching-firms based upon the underwriter reputation (UR) and earnings management (EM) measures covering the period from 1990 to 2007. The returns for all firms going public are calculated for each month within the past 3 years; the annual return for each year is the compound return of these monthly returns calculated from January to December. EM is measured by discretionary current accruals (DCA). The firms are classified into the ‘High’ earnings management group if DCA is greater than the median DCA and into the ‘Low’ earnings management group if DCA is smaller than the median DCA. UR refers to the ranking of the IPO firm’s underwriter (firms are placed into the ‘low’ reputation category if the reputation of the underwriter is <8.0). bWealth relatives are calculated as (1 + BHARHigh EM)/(1 + BHARLow EM), where BHARHigh EM are the adjusted buy-and-hold abnormal returns on IPOs with high earnings management, and BHARLow EM are the adjusted buy-and-hold abnormal returns on IPOs with low earnings management.
Our hypothesis predicts that the difference between high and low earnings management firms, in terms of calendar-time portfolio abnormal returns, should be greater for IPOs underwritten by low-quality underwriters than those involving high-quality underwriters. As shown in Table 8, the Wealth Relative for IPOs issued by low-quality underwriters is below 1.00 (mean = 0.59) for 17 of the 18 years examined.12 In sharp contrast, the wealth relative for the high-quality underwriter group is close to 1.00 in most of the sample years, with a mean value of 0.92. With the one exception of 2005, the wealth relative for the high-quality underwriter group is consistently greater than that for the low-quality underwriter group. The findings in Table 8 are consistent with those reported in Tables 6 and 7, providing clear support for the hypothesis that the negative effect of earnings management on post-IPO performance is mainly driven by those IPOs associated with low-quality underwriters.
()
()The estimation results for the Fama and French three-factor model are presented in Panel A of Table 9, from which we can see that the intercept is significantly negative for the low-quality underwriter group, thereby implying that firms with higher abnormal accruals experience significantly lower abnormal returns in the post-IPO period.14 In contrast, there is no significant difference from 0 with regard to the intercept for the IPOs of firms underwritten by high-quality underwriters.
| Discretionary current accruals (DCA) | Low-quality underwriter reputation (UR < 8.0) | High-quality underwriter reputation (UR ≥ 8.0) | Difference | ||
|---|---|---|---|---|---|
| Coeff. | t-Statistic | Coeff. | t-Statistic | ||
| Panel A: Fama-French three-factor model | |||||
| αp | −1.076 | −2.08** | 0.125 | 0.30 | 5.80** |
| βm | −0.210 | −1.49 | −0.119 | −1.04 | |
| βs | −0.052 | −0.34 | 0.374 | 2.97*** | |
| βh | −0.172 | −0.92 | 0.394 | 2.57** | |
| Panel B: Carhart four-factor model | |||||
| αp | −1.005 | −1.89* | −0.049 | −0.11 | 3.59* |
| βm | −0.230 | −1.59 | −0.073 | −0.62 | |
| βs | −0.042 | −0.27 | 0.350 | 2.78*** | |
| βh | −0.185 | −0.98 | 0.423 | 2.75*** | |
| βu | −0.065 | −0.60 | 0.155 | 1.70* | |
- aDependent variables are formed for the difference between calendar-time portfolio returns of ‘high’ and ‘low’ earnings management firms (HEMRpt − LEMRpt) for each month from April 1989 to December 2007. The calculation of the portfolio return is based upon the monthly returns of firms offering initial equity in the previous 3-year period, on which the following time-series regression is conducted:
-
-
- HEMR pt (LEMRpt) is the value-weighted return for IPO firms with high (low) earnings management. The independent variables are the excess returns on the market portfolio (Rmt − Rft); the difference between the returns of the value-weighted portfolios of small, medium and big stocks (SMBt); and the difference between the returns of the value-weighted portfolios of high, medium and low book-to-market stocks (HMLt); UMDt is defined as the difference between the stock returns of an equally-weighted portfolio with the highest 30 per cent returns and the stock returns of an equally-weighted portfolio with the lowest 30 per cent returns in months t–12 to t–2; Underwriter reputation (UR) is the ranking of the IPO firm’s underwriter (firms are placed into the ‘Low’ reputation category if the reputation of the underwriter is <8.0). Earnings management is measured by discretionary current accruals (DCA); firms are classified into the high (low) earnings management group if DCA is greater (smaller) than the median DCA. The Fama and French (1993) three-factor model and Carhart (1997) four-factor model are employed to calculate the intercept (αp), and the F-statistic is employed to test the difference intercept between the low-quality and high-quality underwriter groups. b***indicates significance at the 1 per cent level; **indicates significance at the 5 per cent level; and *indicates significance at the 10 per cent level.
We also test the difference between the intercepts of high- and low-quality underwriter groups, and find that such difference is statistically significant. All the evidence suggests that the negative relationship between earnings management and post-IPO performance is stronger for the low-quality underwriter group. The findings of the Carhart four-factor model presented in Panel B of Table 9 also provide further evidence of the positive effect of underwriter reputation, in terms of moderating the negative effect of earnings management on post-IPO stock performance.
Finally, we address the issue of endogeneity in underwriter choice. Following Jo et al. (2007), we rerun the analyses using the ‘two-stage least squares’ (2SLS) technique. In the first stage, the underwriter ranking index is regressed on a set of potential determinants of underwriter choice. In specific terms, we include the measures of earnings management (DCA and ADCA), dummy variables for auditor quality, ROA, the logarithm of market value of equity, shares offered to the total number of shares outstanding, industry dummies based on the two-digit SIC code, and year dummies based on the issue year.
Thereafter, based on the first-stage regression results, we compute the predicted value of underwriter reputation ranking and apply the predicted value of reputation in the second stage of the regression following Equations (1) and (2). Our results are virtually unchanged by the findings of the 2SLS analysis.15
5. Conclusions
As a result of the limited information available on IPO firms, it is often difficult for investors to distinguish between changes in abnormal accruals attributable to managers exercising their discretion and those that are due to exogenous influences from the operational decisions taken by firms. We investigate the role of underwriter reputation in determining the nature of earnings management surrounding IPOs, hypothesizing that prestigious underwriters will carefully protect their reputation by auditing and certifying the financial information of IPO firms, thereby restricting any earnings manipulation. As a result, increases in accruals for IPO firms underwritten by high-quality underwriters are more likely to represent changes in the firm’s operational performance. Conversely, increases in abnormal accruals for those IPO firms associated with low-quality underwriters may be the result of opportunistic earnings inflation.
In those cases where changes in accruals are caused by artificial earnings inflation, we find that a significantly negative correlation between earnings management and the post-IPO performance of the stocks, whereas no similarly strong relationship is expected when changes in accruals result from changes in the operational performance of a firm.
We find that IPO firms associated with more prestigious underwriters have significantly less aggressive earnings management, and that the negative relationship between earnings management and the post-IPO performance of a firm’s stocks is discernible only for firms associated with less prestigious underwriters, with no significant results being found for firms associated with more prestigious underwriters. Our findings suggest that underwriter reputation determines the nature of earnings management; thus, it is important to consider the effect of underwriters in the overall relationship between earnings management and post-IPO stock performance.
