Voluntary profit forecast disclosures, IPO pricing and after-market earnings

Transcription

1 Voluntary profit forecast disclosures, IPO pricing and after-market earnings By Paul B. McGUINNESS * Draft: 13 September 2014 * Professor, Department of Finance, The Business Administration Faculty, Room 1212, Cheng Yu Tung Building, The Chinese University of Hong Kong, Sha Tin, New Territories, Hong Kong (Tel.: ) mcguinne@baf.msmail.cuhk.edu.hk. 1

2 Voluntary profit forecast disclosures, IPO pricing and after-market earnings Abstract: Prospectus profit forecasts (PPF) constitute one of the most important accounting disclosures relevant to firm valuation. I examine PPFs in the context of the Hong Kong market setting, in which both IPO activity and PPF disclosure rates appear at notably high levels. My investigation reveals that forecasts typically underestimate future earnings levels (on average, by around 7 per cent). More importantly, ensuing forecast errors bear strong correlation with book-runners initial price determinations (in relation to fixings within the offer price range) and subscribers initial secondary market returns. I also extend the related literature (Cheng and Firth, 2000; Chen, Firth and Krishnan, 2001; Jog and McConomy, 2003; and Chong and Ho, 2007) through a further channel by examining such errors association with longer-term operating and stock returns. Study findings point to a strong connection with longer-term earnings and stock returns. Effects are also apparent between a PPF disclosure dummy and longer-term performance. Finally, I offer support for bivariate signalling effects between PPF disclosure and pre-listing owners retained equity levels (Hughes, 1988; and Li and McConomy, 2004). The signals inverse association suggests that the probability of PPF disclosure is increasing in the size of public float sought. Keywords: Prospectus profit forecasts; profit estimates; post-ipo earnings and turnover; initial and post-ipo stock returns; IPO price determination. JEL Codes: M40; M48; G17 & G30 2

3 Voluntary profit forecast disclosures, IPO pricing and after-market earnings 1 Introduction One of the most important disclosure items relevant to an initial public offering (IPO) is an issuer s prospectus profit forecast (PPF). The decision to disclose a PPF as well as the dollar level of it serve as potentially important signals of firm value (see Trueman, 1986; Hughes, 1986; and Clarkson, Dontoh, Richardson and Sefcik, 1992). The present study investigates the extent to which the PPF disclosure decision and the size of ensuing forecast errors function in explaining post-ipo operating and stock performance. I consider this question for IPO firms listing on the Hong Kong market. This setting is of particular interest given its (1) position as one of the world s leading primary markets and (2) a regulatory-legal environment that facilitates a high rate of PPF disclosure. 1 By addressing the issues above, I significantly extend the boundaries of understanding on the impact of PPFs on firm performance (Cheng and Firth, 2000; Chen, Firth and Krishnan, 2001; Jog and McConomy, 2003; and Chong and Ho, 2007). In terms of the disclosure decision itself, study findings suggest a positive association with post-listing changes in earnings. In contrast to other important PPF environments like Canada (Jog and McConomy, 2004), the disclosure decision bears only limited connection with initial and longer-term stock returns. However, my analysis unveils a substitution effect between the PPF disclosure decision and the proportion of equity retained by pre-listing owners 1. In my study sample, roughly 60 per cent of IPO firms post PPFs. Such a disclosure rate is noticeably higher than in other major IPO markets like the UK, where around 30 per cent of issuers disclose (Jelic, 2008: 4), and the US, where regulations effectively prohibit the release of PPFs (Chong and Ho, 2007).Disclosure rates in Canada (Jog and McConomy, 2003) and Australia (Gallery, Gallery and Linus, 2011) are closer to the Hong Kong rate. Hong Kong also contrasts with other settings, like New Zealand (Bilson, Heaney, Powell and Shi, 2007: 102), where IPO forecasts are mandatory. The Greek market environment is also noteworthy given a switch from a mandatory to voluntary disclosure basis in 2001 (see Gounpoulos, Kraft and Skinner, 2013). Singapore is also of interest given a general decline in the proportion of IPO firms posting forecasts (see Chong and Ho, 2007: 72 for details). 3

4 (Hughes, 1988; and Li and McConomy, 2004). Results suggest that firms seeking larger public float are more likely to post a PPF. This finding is consistent with contentions in the bivariate signalling literature (Hughes, 1988; and Li and McConomy, 2004) that PPF release helps offset the negative signal emanating from a lower than average equity retention level (Leland and Pyle, 1977). In terms of the magnitude of forecast errors, I find a strong association with aftermarket returns and longer-term operating firm performance. Results extend earlier Hong Kong-based evidence (Chen, Firth and Krishnan, 2001) in relation to IPO underpricing. Specifically, my evidence suggests that investors subscription decisions reflect some level of anticipation of ensuing forecast errors. More significantly, the present study demonstrates a strong positive link between forecast error size and the magnitude of post-ipo buy-and hold stock returns (at horizons up to 3-years) and fundamental (earnings and sales revenue) performance. Forecast errors correlate strongly with profit performance between the first and third accounting year-ends post-ipo. As a further important contribution, I also assess how PPFs function in relation to the determination of an issue s final offer price. The overwhelming majority of firms listing on HKEx disclose an offer price range in relevant prospectus documents, with the final offer price pitched at a level within (and not outside) the range. However, determination of final offer price only occurs after the close in the retail offer period. This institutional feature invites the possibility of price fixing below the maximum level. Two important findings emerge from my assessment of the price determination decision. First, IPOs with a PPF are more likely to be priced at the top of the range. Second, there is a significant inverse association between the scale of any price adjustment and the magnitude of earnings forecast errors. My exploration of the price determination process significantly extends the literature on how PPFs function in relation to IPO returns. While the positive link between earnings forecast errors and initial 4

5 secondary market returns is well documented, much less is known about the interaction between earnings forecast errors and initial IPO price determination. Investors anticipation of ensuing forecast errors thus results in two stages of initial price adjustment: (1) from the mid-point in the offer range to final offer price determination; and (2) from final offer price to first secondary market price. In furthering this study s objectives, Section 2 outlines the data design and institutional arrangements surrounding PPF release. Section 3 summarizes the measurement forms and descriptive statistics for prospectus-based earnings projections. Details of the variables and empirical method then follow in Section 4. Finally, section 5 and 6 present results and conclusions. 2 Data and institutional background The present investigation considers all listings via IPO on Hong Kong Exchanges and Clearing Limited s (HKEx s) Main Board market during the 60- month period, 1 January 2005 to 31 December Three factors weigh in my selection of this five-year period. First, it accommodates a relatively large number of issues ((269). Second, the period potentially allows for examination of at least three years data on post-listing operating and stock returns. Third, the sample period contains both bull and bear markets. Two market sentiment dummies (Bull05-07 and Bull09-09) capture such effects. Dummy Bull05-07 controls for the buoyant market conditions between January 2005 and October 2007, while Bull09-09 captures the resurgence of primary and secondary market sentiment between April and December The time-frame enmeshed between the two dummy periods (November 2007 March 2009) relates to bear market effects as wrought by the Global Credit Crunch and subsequent Lehman Brothers fiasco. In arriving at my final sample frame of 269 IPOs, I filter-out all new listings through means other than IPO. This requirement results in the exclusion of issues 5

6 through introduction (where listing is achieved without a corresponding share offer) and transfer listing (from HKEx s second-tier GEM forum to its much larger Main Board. 2 Related Hong Kong-based studies show that PPFs tend to be conservative projections of future earnings (Cheng and Firth, 2000; Chen, Firth and Krishnan (2001). Preliminary analysis of all IPOs in my more recent study period confirms this picture. Accounts in related settings like Singapore (Chong and Ho, 2007) also reveal that issuers and their advisors generally error on the side of caution when posting PPFs. A radically different story plays-out in some other developed market settings. For instance, Jog and McConomy s (2003, p. 140) Canadian IPO evidence reveals optimistic ( conservative ) forecasts in 51 (49) per cent of sample firms. 3 In stark contrast, more than 96 per cent of the PPFs in my study sample constitute conservative projections of a firm s earnings for its first year-end post-ipo. 4 Of the 269 IPO firms, 191 posted either a PPF (n=161) or estimate (n=30) of profits. Thus, almost 60 per cent of issuers came to market with a forward projection of earnings. An estimate of profit arises where the IPO firm s prospectus is dated just after the estimate s financial year-end. To illustrate, Foxconn International released its estimate of earnings, for its 31 December 2004 year-end, in its 17 January 2005 IPO prospectus. This suggests a gap of -24 calendar days between the prospectus and relevant year-end dates. In contrast, a profit forecast has a prospectus disclosure date preceding the financial year-end 2. My focus also dictated the exclusion of three other extraneous listings: (1) a unit IPO, with warrants attaching to offer shares; (2) a dual-listing, where the shares listing on HKEx already had listed status elsewhere; and (3) an issue with a major preferential offer component. 3. See Gounopoulos (2011: ) for useful summary of the international evidence. 4. Jog and McConomy (2003) report that firms posting overly-optimistic PPFs benefit from reduced IPO underpricing but risk the prospect of weaker longer-term returns. Market sanctions against exuberant forecasting likely extend to seasoned equity offer (SEO) funding costs. It is of interest that HKEx imposes an absolute proscription in SEOs during a listed entity s first six months of listing. Market participants may therefore be able to detect an overly-exuberant PPF projection in advance of a newly listed firm s planned SEO (See Rule of HKEx Main Board Listing Rules: 6

7 to which it relates. By way of illustration, Amber Energy Limited posted a profit forecast in its 29 June 2009 prospectus for the forthcoming 31 December 2009 year-end. As with all forecasts, the time gap between prospectus release date and the relevant year-end is positive (= calendar days in the Amber case). Comparison of a firm s prospectus-based projection with its first post-ipo annual report allows for evaluation of the accuracy and bias of a PPF or estimate. Rule of HKEx s Main Board Listing Rules requires annual reports to be posted within four-months of the reporting year-end (i.e., a firm with 31 December 2009 year-end must disclose its 2009 year earnings by latest 30 April 2010). Moreover, PPFs and earnings estimates constitute discretionary disclosure items. HKEx Main Board Listing Rules (Rules ) contain the major provisions relating to such projections. Among other things, forecasts and estimates require some level of sponsor and auditor validation in relevant prospectus documents. From careful study of all prospectus-based disclosures, I note that PPFs and estimates usually appear in the form of earnings attributable to the parent or issuer s (company s) shareowners. This is not the case for a minority of issues however. I thus exercise care to ensure that post-listing disclosures correspond to the precise form of prospectus-based projections. PPFs and earnings estimates almost always appear in the Summary section of relevant prospectus documents, and thus constitute highly-visible projections. For two of the 161 issuers posting forecasts, the PPFs appear in only the Financial Information and associated appendix sections. There is therefore a question as to a forecast s relative visibility. For reasons of completeness, I record a PPF or estimate wherever it appears in in the relevant prospectus. In terms of issue pricing, I observe that the majority of IPO prospectus documents (n=248) stipulate an offer price range. The market convention is that a final clearing price is determined some days after prospectus release. The price fixing 7

8 must be at a level within the range and applies as a clearing price to all subscribers allotted shares. 5 The declaration of final offer price occurs just after the close in the formal retail offer period. In respect of the study sample, 49.44% of issuers had final offer prices below the maximum; and 42.75% had prices at the top of the range. A fixed pricing regime (i.e., a single offer price, rather than offer price range) applied to the remaining 7.81% of issues. Prospectus and financial reports, available via the HKEx website ( provide this study s major source of accountingbased information. Datastream serves as the source for secondary market price and liquidity information. Additionally, HKEx Fact Books and a range of IPO-related announcements (as posted to the HKEx website) offer details on other important listing and post-ipo characteristics. 3 Measurement forms and descriptive statistics for earnings projection (PPF and Estimate ) errors Projection error metrics I determine earnings forecast and estimate errors as follows: PPFerr = ACT PPF [1] ACT ESTerr = ACT EST [2] ACT ACT captures actual earnings for the financial year-end corresponding to a company s forecast (FOR) or estimate (EST) of earnings. PPFerr and ESTerr capture the corresponding proportionate errors. 5. This applies unless the issuer revises the offer range through supplementary prospectus. In such cases, the supplementary document must appear prior to the close in the formal application period, as relevant to the earlier prospectus document. Where a supplementary prospectus alters the offer price terms, the final offer price determination would need to be within the new range. The practice of pricing within (and not outside the prospectus disclosed-range) is much more akin to European than US practice (see Jenkinson, Morrison and Wilhelm, 2006). 8

9 The above construction, which is widely employed elsewhere (see, for example, Cheng and Firth, 2000), allows for direct comparison with related studies. Implicit to [1] and [2] above, is the balanced treatment of negative and positive departures of a given absolute $ amount. 6 Various studies of PPFs also distinguish between the accuracy and bias of forecasts. Given that the overwhelming majority of disclosures in Hong Kong are conservative projections of future earnings, this study s analysis primarily focuses on forms [1] and [2]. However, to capture the accuracy of forecasts, an absolute form of proportional errors features elsewhere (see, for example, Cheng and Firth, 2000). To allow comparison with forecast bias in descriptive statistics, I express the accuracy of prospectus earnings forecasts and estimates as follows: ABS (PPFerr) = ACT PPF [3] ACT ABS (ESTerr) = ACT EST [4] ACT An alternative, as in Jog and McConomy (2003) and Chong and Ho (2007), is to consider PPF minus ACC all divided by PPF. In my study such an approach is problematic due to one issuer posting an earnings forecast of $0. In error constructions [1]-[2], a positive numerator value suggests a conservative forecast or estimate. I observe conservative projections in 184 of 191 cases. For the remaining seven IPOs, one (Hontex) failed to report post-ipo earnings. The other six cases reported earnings at the first post-ipo year-end below projection. Table 1 reports respective mean and median PPFerr values of 7.76 per cent and 6.02 per cent. Such results compare favourably with Chong and Ho (2007: 77) 6. For example, a PPF of $90 million for the 31 December 200x year-end would generate a PPFerr value of +10% if the realized profit is $100 million. Similarly, a PPF of $110 million (given earnings of $100 million at the first year-end post-ipo) suggests a PPFerr value of -10%. 9

10 where respective figures, for Singaporean IPOs between 1990 and 2000, are 6.41 and 4.40 per cent. The results also suggest a contraction in forecast errors relative to earlier Hong Kong-based studies (Cheng and Firth, 2000; and Chen et al., 2001) A number of qualifications apply in interpreting the bias and accuracy prospectusbased earnings projections. First, forecasts and estimates are typically expressed as being not less than a specific figure. A prospectus-based disclosure of earnings therefore constitutes a projection s lower bound. Second, virtually all projections display a positive value. Only one issuer came to market with a forecast earnings loss. 7 One other issuer had a PPF of at least $0. 8 A third important observation concerns the generally short horizon of PPFs. Nearly 48 (63) per cent of PPFs had a time horizons within 90 (120) calendar days. Descriptive Statistics An estimate of earnings is easily discernible from a forecast in terms of time horizon. Variable Horizon, measured as the number of calendar days from the date of prospectus to the financial year-end to which the earnings projection relates, highlights this difference (Table 1). For all 161 PPFs, Horizon has nonnegative value. In contrast, all estimates (n=30) exhibit negative Horizon values. Table 1 also reports descriptive statistics for PPFerr and ESTerr. The respective mean values are 7.76 (n=160) and 3.52 per cent (n=30). These figures compare favourably with Hong Kong-based studies of PPFs (Cheng and Firth, 2000; and Chen et al., 2001). They also complement results on forecast accuracy in other 7. This was GCL-Poly Energy Holdings Limited. The relevant declaration in its 31 October 2007 prospectus states a forecast consolidated net loss not more than (HK$212.1 million). This relates to its 31 December 2007 year-end. Actual profit for this year-end, disclosed in its 2007 Annual Report is -HK$ million, suggesting a PPFerr value of %. 8. This case was for Advanced Semiconductor Manufacturing (ASM). The declaration in its prospectus of 27 March 2006 is for Forecast profit for the six months ending June 30, 2006 not less than RMB0. Its 2006 Interim Report reports a profit of RMB million for this six-month period. Its PPFerr is 100 per cent. However, even a deviation of RMB1 from forecast would generate such a result. Consequently, I truncate PPFerror for ASM to zero. 10

11 Pacific-Basin market settings, such as Australia, New Zealand and Canada. 9 The upshot is that forecast disclosures are on average much tighter, and less biased and more accurate projections of future earnings, than hitherto. For example, Chen et al. (2001) report negative forecast errors for more than 25 per cent of the Red-Chip and H-issuers in their sample. However, for other issuers in their sample, the proportion of negative forecast errors is much lower at around 6 per cent (see Page 235 of Chen et al. (2001) for further details. The subsequent tightening in the distribution of forecast errors and the preponderance of positive errors (in Table 1 of my results) points to a more mature and better-regulated Hong Kong market setting. 10 Table 1 highlights the conservative nature of the forecasts. Only six of the 160 IPO firms with available PPFerr scores generated profits below PPF projection. Of these, the most extreme case entailed a PPFerr value of -187 per cent. 11 The other five cases recorded PPFerr values between and per cent. At the other end of the spectrum, the maximum forecast error is per cent. To allow for better control of outliers I winsorize the lowest four PPFerr values to the fifth lowest value. A similar process applies to the distribution s upper tail. As explained in Table 1, this process entails censoring 5 per cent of the PPFerr distribution (2.5 per cent of each tail). The resulting variable, WPPFerr, appears much better normally-distributed. At the same time, the effect on the mean value of the forecast error distribution is only marginal, given a non-winsorized mean of 7.76 per cent as compared to a Winsorized value of 8.89 per cent. 9. See Cheng and Firth (2000: 424-5) for comparison of study findings for these markets. For respective studies on Australian and Canadian PPFs, refer to Hartnett and Romcke (2000) and Li and McConomy (2004). 10. See McGuinness s (2002) review of earlier related studies, which demonstrates a marked contraction in prospectus forecast errors though the 90s running up to The case is Amber Energy Ltd. (prospectus date: 29 June 2009) which posted a PPF of not less than RMB62 million (p. 7) for the 31 December 2009 year-end. Its 2009 Annual Report reveals a profit of RMB million. Sometime earlier, Amber issued a Profit Warning ( 5 November 2009, alerting investors to the possibility of it significantly undershooting the PPF. 11

12 4 Variables and empirical methods employed In relation to the determinants of forecast disclosure, I propose the following empirical model form for testing: Discl(PPF) i = β 0 + β 1.Horizon i + β 2.H-RC i + β 3.LnTA i + β 4.UWQ i + β 5.EqRet i + β 6.LnPrRev i + β 7.LnUnd i + β 8.PrCh3a1a + β 9. Bull05-07 i + β 10.Bull09-09 i + e i (Equation 1) The time horizon (Horizon) between IPO prospectus date and the issuer s first post-listing earnings disclosure is likely to be important in explaining whether an issuer elects to post a PPF. The longer the horizon, the greater the uncertainty surrounding any projection, and the greater the attendant disclosure risks for issuer and advising agents. Table 1 reports descriptive statistics for Horizon (in calendar days) in relation to both PPFs (n=161) and estimates (n=30). For issuers disclosing only historical earnings (n=78) in listing documents, Horizon measures the gap between a company s prospectus announcement and its first accounting year-end post-ipo. This time horizon is positive for 76 (of the 78) companies. 12 Via dummy H-RC, I also control for whether an issuer is a mainland Chineseincorporated H-issuer or an offshore-incorporated (but PRC state-controlled) red chip counter. Forty-two H- and 14 red chips receive value for this variable, constituting just under 21 per cent of all sample firms. I conjecture that H- and red chip issuers bear greater inclination to PPF disclosure. As argued in Ferguson, Lam and Lee (2002), H-issuers and red chips display a greater propensity to voluntary disclosure than other HKEx-listed firms. They ascribe such results to both SOE strategy and general state-funding requirements. Regulatory demands 12. For two companies, the time gap is negative. For Bauhaus (Prospectus date 29 April 2005; reported historical earnings for 7 months up to 31 October 2004) its first post-ipo earnings disclosure relates to the 31 March 2005 accounting year-end. For China Properties Group, its prospectus, dated 9 February 2007), reports historical earnings for 9 months up to 30 September Its first disclosure of annual earnings post-ipo relates to the 31 December 2006 year-end. 12

13 may also promote a higher PPF disclosure rate, given requirements for listing ascent from both the China Securities Regulatory Commission and HKEx. 13 In capturing the ex-ante uncertainty surrounding after-market earnings, I consider the natural logarithm of an issuer s total assets (in HK Dollars) immediately prior to IPO (LnTA). Evidence in Cheng and Firth (2002, p. 433) points to a negative association between issuer size and forecast errors. 14 Other important factors that potentially account for the presence or not of an issuer-based forecast include the quality (reputation) of advising agents. Strong reputational incentives exist for advising agents, principally issue underwriters (sponsors) and reporting accountants, responsible for ratifying and supporting prospectus forecasts, to perform careful due diligence. The level of diligence exercised is likely to be increasing in the reputation capital of issue advisors. It should also exhibit an inverse association with the level of ex-uncertainty surrounding future earnings (Beatty and Ritter, 1986). I construct categorical variable UWQ to capture advising agent quality. From inspection of underwriters/sponsors disclosed in HKEx Fact Books (2005-9), I identify 14 Big-name players plus the China International Capital Corporation (CICC) as the market s leading sponsors. UWQ takes on value three in cases where two or more big-name sponsors are present and value two where only one big-name is present. A value of one applies in issues with two or more regionallybased sponsors and zero in all remaining cases. As routinely suggested in the relevant literature (see Cheng and Firth, 2000), underwriter quality is central to both the PPF disclosure decision and the accuracy of any ensuing errors. 13. It is of interest that Chen et al. s (2001) evidence points to a greater concentration of negative forecast errors in Red-Chip and H-issuers. 14. An alternative measure of ex-ante uncertainty, SD1-30 is the standard deviation of consecutive daily close-to-close returns, as measured between the close of trade on the first day of listing and the thirtieth. As indicated in Ritter (1984), the standard deviation of post-listing returns proxies for the uncertainty surrounding an issuer s intrinsic value. Ceteris paribus, greater return volatility should exist where opinion divergence about post-ipo earnings is greater. 13

14 Equation (1) also includes control for the level of equity retained by pre-listing owners (EqRet). The bivariate signalling literature (Hughes, 1988; and Li and McConomy, 2004) offers the possibility of a substitution effect between PPF disclosure and equity retention. In effect, issuers wishing to raise more capital (and thus a higher float) may be more inclined to post a PPF to signal underlying firm value. Such disclosure may help in offsetting the negative signal resulting from a lower than usual EqRet level (Leland and Pyle, 1977). I measure pre-listing holders equity retention level after exercise of any associated Green Shoe. 15 Through the signalling properties of earnings projections (Truman, 1986), issuing firms potentially promote higher offer prices and reduce underpricing levels. As argued in Allen and Faulhaber (1989), Grinblatt and Masulis (1989) and Welch, 1989), underpricing itself could also signal firm value. It is therefore of interest to investigate whether the PPF disclosure and underpricing signals acts as complements or substitutes. Jog and McConomy s (2003) empirical evidence is consistent with the latter view. In addition, and in a general sense, release of a voluntary management forecast acts to constrain information asymmetries (Coller and Yohn, 1998). The literature on IPOs reinforces this contention. Specifically, in environments where voluntary PPFs are common, issuers failing to disclose suffer greater IPO underpricing (see Jog and McConomy, 2003, for Canadian IPOs). At the same time it is conceivable 15. I construct EqRet = {[N b N sale(go) N sale(oa) N conv(go) N conv(oa) ]/N a }*100 N b = Number of shares outstanding just prior to IPO; N sale(go) (N sale(oa) ) the number of shares sold in the global offer (in subsequent Green Shoe exercise) by pre-listing owners; N conv(go) (N conv(oa) ) = Number of shares converted into H-form and transferred to NSSF at IPO (at Green Shoe exercise); N a = Number of shares outstanding after Green Shoe expiry. Both N conv(go) and N conv(oa) are zero for non-h issuers (N=228). For the 42 H-issuers with positive (zero) N sale(go) and N sale(oa) values, N conv(go) and N conv(oa) are both zero (positive). The only exception is ICBC which had a secondary offer component, with proceeds accruing to a party other than the NSSF, as well as conversion of state-held domestic stock into H-form. For the six H-issuers with an adjoining A-share IPO, N a =number of shares after both A- and H-IPO completion and subsequent Green Shoe. Of the A-share IPOs, only ICBC s involved OAO exercise. I modify RetEq for two H- issuers to account for state-sanctioned conversion of domestic stock into tradable A-form. In both cases, China Metallurgical Corp. and China Pacific Insurance, shares were channelled to the NSSF. Adjustments to RetEq for Foreign Legal-Person share holdings also feature for five SOE bank H-issues. 14

15 that earnings forecasts could excite investor demand and promote above-thenorm subscription rates with unfulfilled excess demand spilling-over into initial secondary market prices. If this view holds, issuers with prospectus-based earnings disclosures should be in a better position to attract higher subscription rates on average and thus higher initial returns levels. I first control for the price adjustment implicit in the book-runners determination of final offer price. LnPrRev captures such price adjustment: LnPrRev = Ln(FOP/Mid) Ln(M T-5 /M T-10 ) Where FOP is the issue s final offer price; Mid is the mid-price of the issuer s prospectus-disclosed offer price range {=[(minimum price + maximum price)/2]}; and M T-10 and M T-5 are the respective closing prices in the Hang Seng Index ten and five business days prior to listing. LnPrRev incorporates general market movement between 10 and five business days prior to listing. Such adjustment captures the typical time gap between prospectus release (10 days prior) and price determination (five days prior to listing). Variable LnUnd then captures the subsequent price run-up between final offer price determination and secondary market open. I construct this variable as the difference between an IPO stock s opening price on the first day of listing and its final offer price (FOP), all divided by FOP. This measure is then adjusted for contemporaneous movement in the Hang Seng Index between an IPO s price determination date and the open of trade on its first day of listing. To summarize, LnUnd = Ln(P 0pen, 1 /FOP) Ln(M 0pen, 1 /M T-5 ) Where P 0pen, 1 is an IPO firm s secondary market price at the open of the first day of listing; FOP the issue s final offer price, M 0pen, 1 the overall market s Hang Seng Index level at the IPO firm s secondary market open; and M T-5 the closing Hang Seng Index level five business days to the first day of listing. I also examine whether longer-term buy-and-hold returns bear association with (1) the PPF disclosure decision and (ii) the size of ensuing forecast errors. In this respect, I consider buy-and hold returns at one-, two- and three-year post-ipo horizons. The form of such returns is as follows: 15

16 LnBhar (i, t=1 T) = Ln(RI i, T /RI i, t=1 ) - Ln(HSIRI i, T /HSIRI i, t=1 ) RIi is stock i s Return Index level (determined from Datastream). As of the close of the first day of listing (t=1), RI j, t=1 is 100. T values for one-month, 12 months and 36 months post-ipo are respectively 1m, 1y and 3y. Similarly, HSIRI i, t=1 and HSIRI i, T reflect respective RI levels on the Hang Seng Index at the close of stock i s first day of listing (t=1) and the close of post-listing day T. As an alternative to stock return performance, I include variable PrCh3a1a in relevant regressions. The form of this variable is as follows: Prch3a1a = (Net profit at third year-end - net profit at first year-end post-ipo) Absolute net profit for first year-end after IPO Finally, in respect of Equation (1), I control for general secondary market sentiment using variables Bull05-07 and Bull I assign value one to dummy Bull05-07 (Bull09-09) for issues with prospectus announcement dates between 1 January 2005 and 31 October 2007 (1 April 2009 and 31 December 2009). In respect of earnings projection errors, I test the following empirical form: WPPFerr i (and/or ESTerr) = β 0 + β 1.Horizon i + β 2.H-RC i + β 3.LnTA i + β 4.UWQ i + β 3.EqRet i + β 6.Qratio i + β 7.LnPrRev i + β 8.LnUnd i + β 9.PrCh3a1a + β 10.Bull05-07 i + β 11.Bull09-09 i + e i (Equation 2) The dependent projection error variables are as defined earlier in this paper. WPPFerr captures the magnitude of Winsorized prospectus earnings forecast errors, while ESTerr the errors resulting from prospectus-based earnings estimates. Variable definitions for Horizon, H-RC, LnTA, UWQ, EqRet, LnPrRev, LnUnd, PrCh3a1a, Bull05-07 and Bull09-09 apply as in Equation (1). Equation (2) includes an additional effect for earnings growth through Qratio. This variable captures Tobin s Q measure, defined as: Qratio = Issuer s market capitalization + book value of total liabilities Book value of pre-ipo total assets + gross funds generated 16

17 The market capitalization in the ratios numerator reflects an issuing firm s total shares outstanding as of the date of over-allotment option expiry. The gross proceeds figure in the denominator captures funds generated from new IPO and Green shoe shares. As with the market capitalization measure, such funds reflect an issue s final offer price. Finally, in this section, Table 2 reports descriptive statistics of all dependent and explanatory variables featured in the foregoing. 5 Results Table 3 presents bivariate statistics in respect of the association between PPF disclosure and post-ipo profitability. Results show that firms coming to market with a PPF typically outperform all others over a three year-period measured from the accounting year-end immediately prior IPO to the third accounting year-end post-listing. Firms with PPFs (n=161) generally outperform those with Estimates (n=30). In turn, issuers with estimates generally outperform firms without any forward-looking earnings projection (n=78). It is also noteworthy that post-listing profitability s connection with Discl(PPF) is considerably stronger than with IPO underpricing (see Pearson correlations in Table 6). 16 Table 4 (Columns 1-3) complements the bivariate analysis in Table 3 by demonstrating the relation between year-on-year profit change and the forecast disclosure decision dummy, Discl(PPF), after due control for other firm- and market-level characteristics. The various binary logit regressions offer a consistent view of pertinent explanatory factors. Four principal findings emerge. First, forecast disclosure is increasing in underwriter or sponsor quality (UWQ) and firm size (LnTA). Second, it is decreasing in the time horizon between prospectus release and the date of the first accounting year-end post-ipo (Horizon). Third, issuers with stronger post-listing changes in earnings appear more likely to post PPFs (or earnings estimates ). Fourth, a significant inverse association is 16. This finding extends US evidence on the general absence of a link between IPO underpricing and post-listing profit change (Jain and Kini, 1994). 17

18 apparent between the forecast disclosure decision and the proportion of equity retained (EqRet) by pre-listing owners. 17 The first two findings largely accord with expectations. The third result points to PPFs having an important signalling role in relation to future earnings. 18 The fourth suggests a substitution effect between the forecast disclosure decision, Discl(PPF), and pre-listing owners equity retention rate (EqRet). This finding provides an important additional result on the relevance of bivariate signalling models (Hughes, 1988; and Li and McConomy, 2004) to IPO forecast and insider holding disclosures. My findings on this subject are consistent with forecast disclosure, Discl(PPF), compensating for or offsetting the negative signal emanating from a lower than average EqRet level. Results suggest therefore that a PPF disclosure helps promote issues where a relatively large public float is sought. It is also consistent with a PPF serving a liquidity-enhancing role, which may be especially important in issues demanding a larger than typical float. In terms of forecast errors, WPPFerr, results in Table 5 reveal that the principal explanatory factors are Qratio, LnPrRev, Bull05-07 as well as respective performance variables PrCh1a1b, PrCh2a1a, PrCh3a1a and LnBhar3Y. The overall findings in Table 5 are consistent with forecast errors presaging subsequent earnings momentum or positive earnings drift. They offer a useful complement to the literature on so-called post- earnings-announcement-drift (PEAD). 19 My results also confirm the importance of the initial price-fixing process. Moreover, the price fixing adjustment (LnPrRev) appears to impound a greater 17. Keasey and McGuinness s (2008) study of IPOs pitched on HKEx between January 2002 and December 2003 also highlights an inverse association. They report a correlation between the two variables of (see Page 654). 18. My analysis here specifically focuses on whether PPF disclosures and their associated errors help in anticipating subsequent earnings change. In contrast, McGuinness (2014) provides wider examination of post-ipo fundamental and stock return performance for Hong Kong IPOs, identifying strong effects in respect of subscription cascades, an issuer s level of strategic-political importance and uncertainty surrounding initial issue pricing. 19. See Zhang, Cai and Keasey (2014) for recent analysis of PEAD, in respect of trading strategies, and examination of the relevant literature. 18

19 proportion of anticipated forecast error than does the initial secondary market adjustment (LnUnd). In other words, the initial IPO return has a residual effect which is subordinate to the implicit adjustment from mid-price (in the offer range) to final offer price. This last result provides an important extension of the related evidence for Hong Kong (Cheng and Firth, 2000), which only considers forecast error associations with initial returns (and not with the price fixing adjustment). Findings in Table 5 also appear robust to the inclusion of specific calendar year dummies (D2005, D2006, D2007, D2008 and D2009) to control for the year of an issuer s first earnings year-end post-ipo. Columns 5 and 6 of Table 5 reflect the inclusion of such dummies. For reasons of brevity, I do not report such dummies estimated coefficients and t statistics. Nonetheless, results in Columns 5 and 6 of Table 5 reflect their inclusion. It is instructive to note that only the D2008 dummy achieves significance at conventional levels. Perhaps not too surprisingly, given the contraction in general market earnings during 2008, the D2008 dummy coefficient exhibits a negative relation with WPPFerr. Table 5 results are also materially unchanged when adding-in the 30 IPO firms that came to market with prospectus-based earnings estimates Conclusions The present study examines the underlying factors relevant to prospectus earnings projections in a market setting where voluntary forecast disclosure is commonplace. I note that around 60 per cent of issuers post a prospectus profit forecast (PPF). Moreover, such forecasts tend to be conservative projections of future earnings. Furthermore, PPFs tend to be fairly accurate with mean errors around the 7 per cent mark. 20. Inclusion of such cases leads to my use of a co-joined dependent variable, WPPFerr & ESTerr. Again, for reasons of brevity, I do not report such results. But in summary, significant effects remain. One visible but perhaps unsurprising change is a marked increase in the significance level of variable Horizon. 19

20 I also note that PPF disclosure, in addition to being explained by factors such as forecast period horizon, firm size and underwriter quality, bears a strong inverse association with pre-listing owners retained equity levels. This last result provides empirical support for bivariate signalling effects between retained equity and voluntary disclosure (Hughes, 1988; and Li and McConomy, 2004). It is also consistent with PPF disclosure functioning to support higher offer prices in issues demanding more capital (and thus public float). In terms of the magnitude of ensuing forecast errors, I find a strong correlation with book-runners initial price determinations, as measured by adjustments from the mid-point in the prospectus-disclosed offer price range to final offer price fixing. Forecast error size also displays a strong association with subscribers initial returns, as measured between price determination and the open of listing. It is instructive that forecast errors bear a stronger link with initial price fixing adjustments than subsequent initial secondary market returns. This observation suggests that book-runners response to formal subscription activity reveals much about the size of forthcoming forecast errors. In sum, the greater the upward adjustment (from mid-point in the offer range), the larger the size of ensuing forecast errors. This insight significantly extends commentary on the link between PPF disclosure and initial IPO returns (Cheng and Firth, 2000; Chen, Firth and Krishnan, 2001; Jog and McConomy, 2003; and Chong and Ho, 2007). The present investigation also sheds light on forecast errors connection with longer-term operating and stock returns. I detect strong associations in respect of post-ipo profit change. Such findings complement and add insight to the positive effects discernible between the PPF disclosure dummy and earnings change. 20

21 Table 1 Descriptive Statistics PPFs: Profit Forecasts (n = 161) # N Minimum Maximum Mean Median Std. D (%) (%) (%) (%) (%) Horizon (days) PPFerr WPPFerr ^ Abs(PPFerr) STD(AbsPPFerr) Profit Estimates (n = 30) N Minimum Maximum Mean Median Std. D (%) (%) (%) (%) (%) Horizon (days) ESTerr (%) Abs(ESTerr) (%) Definition of variables: PPFerr ESTerr ACT AbsPPFerr (ACT PPF)/Abs(ACT) (ACT EST)/Abs(ACT) Actual earnings for the financial year-end corresponding to a company s forecast (FOR) or estimate (EST) of earnings. PPFerr Notes: (#) One observation is lost from the analysis. Computation of a PPFerr for IPO firm Hontex (as listed on HKEx in December 2009) proved impossible. This company was suspended from trading in February 2010 and eventually forced to withdraw its listing. (^) WPPFerr is the Winsorized form of FFPerr, achieved by setting (i) the four most extreme negative observations to the PPFerr value of the next highest (= %) and (2) the four highest FPPerr values to the next lowest (= 40.10%). The adjustment of the eight observations described above results in Winsorization of 5 per cent of the FPPerr distribution. The resulting Jarque-Bera statistic for WFPPerr is markedly lower (50.07) than for FPPerr (17,748). This indicates that the Winsorizing process imparts much greater normality into the distribution of forecast errors. 21

22 Table 2 Descriptive statistics for key variables Mean Median Maximum Minimum Std. Dev. N Discl(PPF) Discl(EST) Discl(HIST) ESTerr PPFerr WPPFerr PPFerr & ESTerr (combined) WEFEerr & ESTerr (combined) Horizon H-RC LnTA UWQ EqRet Qratio LnPrRev LnUnd LnBhar1Y LnBhar2Y LnBhar3Y PrCh1a1b PrCh2a1a PrCh3a1a PrCh3a1b WPrCh3a1a WPrCh2a1b WPrCh3a1b Bull Bull Variable definitions: Discl(PPF) Dummy with value one for an issuer posting a prospectus forecast of earnings. Discl(EST) Dummy for an issuer releasing a prospectus estimate of year-end earnings. Discl(HIST) Dummy for an issuer without a prospectus forecast or estimate of year-end earnings. PPFerr (ACT PPF)/ ACT ; where ACT is net profit for the year-end corresponding to forecast. ESTerr (ACT EST)/ ACT ; where ACT is net profit for the year-end corresponding to estimate. WPPFerr Winsorized form of FFPerr (with Winsorization of 2.5 per cent of of PPFerr). Horizon Time in calendar days from date of prospectus to first financial year-end post-ipo. H-RC Dummy for mainland Chinese-incorporated H- and Red-chip issuers. LnTA Natural log of firm s total assets in HKD (from prospectus) and excluding IPO proceeds. UWQ Categorical variable for underwriter quality (3 highest quality; 0 lowest). EqRet The percentage of equity retained by pre-listing owners on Green Shoe expiry. Qratio Tobin s Q based on final offer price and after Green Shoe completion. LnPrRev Ln(FOP/Mid) Ln(HSI T-5 /HSI T-10 ). LnUnd Market-adjusted buy-and-hold return (in nat logs) for period from price fixing to listing open. LnBhar1Y Market-adjusted buy-and-hold return (in nat logs) for period from first close to one year later. LnBhar2Y Market-adjusted buy-and-hold return (in nat logs) for period from first close to two years later. LnBhar3Y Market-adjusted buy-and-hold return (in nat logs) for period from first close to three years later. PrCh1a1b PrCh2a1a [(Net profit at 2 nd year-end post IPO - net profit at 1 st year-end)/absolute net profit for 1 st year-end post-ipo]. PrCh3a1a [(Net profit at 3 rd year-end post IPO - net profit at 1 st year-end post-ipo)/absolute net profit for 1 st year-end post-ipo]. PrCh3a1a [(Net profit at 3 rd year-end post IPO - net profit at year-end prior to IPO)/Absolute net profit at year-end prior to IPO]. WPrCh3a1a Winsorized form of PrCh3a1b (bottom 3observations set to ; and top 3to ). WPrCh2a1b Winsorized form of PrCh2a1b (bottom 3 observations set to ; and top 3 to ). WPrCh3a1b Winsorized form of prch3a1a (bottom 3 observations set to ; and top 3 to 4.362). Bull05-07 Dummy for issues with prospectus dates between 1 January 2005 and 31 October Bull09-09 Dummy for issues with prospectus dates between 1 April 2009 and 31 December [(Net profit at 1st year-end post IPO - net profit at year-end prior to IPO)/Absolute net profit for year-end prior to IPO]. 22

23 Table 3 Proportional profit changes from the accounting year-end immediately prior to IPO to accounting year-ends 1-, 2- and 3-years post-listing 1 = Discl(PPF) 0 = Discl(HIST) N Mean Std. Dev. t-test (diff. in means) Median WPrch1a1b *** WPrch2a1a *** WPrch3a2a WPrch2a1b *** WPrch3a1b *** WPrch3a1a *** Notes:*, **, *** Indicates significant t-statistics for difference in mean values at respective 10, 5 and 1 per cent levels and assuming unequal variances. Descriptive statistics for cases (n=161) with prospectus-based Forecast projections WPrch1a1b WPrch2a1a WPrch3a2a WPrch2a1b WPrch3a1b WPrch3a1a N Mean Median Descriptive statistics for cases (n=30) with prospectus-based Estimate projections WPrch1a1b WPrch2a1a WPrch3a2a WPrch2a1b WPrch3a1b WPrch3a1a N Mean Median Descriptive statistics for cases (n=78) with issuers with profits in only historical form WPrch1a1b WPrch2a1a WPrch3a2a WPrch2a1b WPrch3a1b WPrch3a1a N Mean Median WPrCh1a1b Winsorized form of Prch1a1b (bottom 3 observations set to ; top 3 to 9.225). WPrCh2a1a Winsorized form of Prch2a1a (bottom 3 observations set to ; top three to 5.539). WPrCh3a2a Winsorized form of Prch3a2a (bottom 3 observations set to ; top 3 to 8.170). WPrCh2a1b Winsorized form of PrCh2a1b (bottom 3observations set to ; top 3to ). WPrCh3a1b Winsorized form of Prch3a1a (bottom 3observations set to ; top 3to 4.362). WPrCh3a1a Winsorized form of PrCh3a1b (bottom 3 observations set to ; top 3 to ). 23

24 Table 4 Binary Logit regressions explaining factors relevant to the disclosure of prospectus-based earnings projections Discl(PPF) i = β 0 + β 1.Horizon i + β 2.H-RC i + β 3.LnTA i + β 4.UWQ i + β 5.EqRet i + β 6.LnPrRev i + β 7.LnUnd i + β 8.WPrCh3a1a + β 9. Bull05-07 i + β 10.Bull09-09 i + e i (Equation 1) PPF 1 (N=161); 0 (N=108) # PPF (& EST cases = 1) 1 (N=191); 0 (N=78) # PPF (Excluding EST cases) 1 (N=161); 0 (N=78) # PPF (& EST cases = 1) 1 (N=191); 0 (N=78) # Coefficient z-statistic Coefficient z-statistic Coefficient z-statistic Coefficient z-statistic Intercept Horizon * *** *** *** H-RC LnTA ** ** *** ** UWQ *** *** *** *** EqRet ** * * * LnPrRev LnUnd WPrCh3a1a ** ** LnBhar3Y Bull Bull ** Huber/White adj. Yes Yes Yes Yes N 261 # (missing, n=8) 261 # (missing, n=8) 233 # (missing, n=5) 269 # (missing, n=0) McFadden R2 adj Notes: *, **, *** Indicates significant (two-tailed) t-statistics at respective 10, 5 and 1 per cent levels. Dependent variables: Explanatory variables: PPF: Binary variable with value one for an issuer posting a prospectus forecast of earnings, as relevant to an issuing firm s first post-ipo disclosure of profits. This variable assigns value zero to all issuing firms with either an Estimate or an historical earnings figure as its most timely disclosed level of profits at IPO. Horizon: Time in calendar days from the issuer s date of prospectus to the period-end of first financial statement date post-ipo; H-RC: Dummy variable for mainland Chineseincorporated H- and Red-chip issuers; LnTA: Natural log of an IPO firm s total assets, as disclosed in the issuer s prospectus and adjusted into HKD where necessary (amount excludes net proceeds from primary offer shares in the IPO); UWQ: Categorical variable for underwriter quality (=3 where two or more big-name sponsors are present; 2 where only one big-name sponsor is present; 1 in issues with two or more regionally-based sponsors; and 0 for all other issues); EqRet: Percentage of equity retained by pre-listing owners after Green Shoe exercise; LnPrRev: = Ln(FOP/Mid) Ln(HSI T-5 /HSI T-10 ), where FOP is the final offer price, Mid is the mid-price in the offer range and HSI is the Hang Seng Index; LnUnd: Ln(P 0pen, 1 /FOP) Ln(M 0pen, 1 /M T-5 ), where P 0pen, 1 is an IPO firm s secondary market price at the open of the first day of listing; FOP the issue s final offer price, M 0pen, 1 the overall market s Hang Seng Index level at the IPO firm s secondary market open; and M T-5 the closing Hang Seng Index level five business days to the first day of listing; WPrCh3a1a: Winsorized form of PrCh3a1a = (Net profit at third year-end post IPO less net profit for first year-end post-ipo/absolute net profit for first year-end post-ipo)*100; LnBhar3Y: Buy-and-hold return over the first 36 months of listing, utilizing the closing Return Index (RI, from Datastream) of stock and Hang Seng Index between day t=1 and t+36m, where t=1 is the stock s closing RI on the first day of listing and t+36m its close on the 792 nd business day of listing (LnBhar3Y = Ln(RI792/RI1)-Ln(HSIR792/HSIR1); Bull05-07 (Bull090-09): Dummy for issues with prospectus dates between 1 January 2005 and 31 October 2007 (1 April 2009 and 31 December 2009). 24