An analysis of Accounting-based valuation models using Australian IPOs Lachlan Tuite The University of Sydney btui2697@usyd.edu.au Abstract: Comparable firm multiples are often supported as the primary method to be used in valuing Initial Public Offerings (IPOs) despite the existence of more sophisticated models. Empirical research has found relatively weak support for comparable multiples in IPO settings and there are few studies that investigate a range of alternatives to this method. This study aims to contribute to the literature by providing such an analysis of a range of alternative models, and will do so through a value relevance framework. Furthermore, a review of the literature unveils some inconsistency concerning the value relevance of management earnings forecasts frequently provided by IPO firms in disclosure documents. This study will be able to provide some resolution of this inconsistency. Finally, this study seeks to examine several criticisms of the Edwards-Bell-Feltham-Ohlson (EBFO) model’s implementation in an Australian IPO setting. 1 1.0 Introduction An Initial Public Offering (IPO) is when an entity first goes to the public to raise funds. IPOs are characterised by a higher level of uncertainty than seasoned equity offerings and other events that might occur during a lifetime of a firm including mergers, acquisitions, and management buyouts. Such increased uncertainty is due to a large information asymmetry between the promoters of the issue and prospective buyers. This makes IPOs an interesting situation in which to explore valuation issues. This study seeks to resolve three related issues that emerge form the literature. Firstly this study examines the effectiveness of various accounting-based valuation models in estimating the value of IPOs. Whilst the comparable multiples approach is widely endorsed by practice and academia, there are other valuation models that are ostensibly better theoretically developed than this approach. For example, the Edwards-Bell-Feltham-Ohlson (EBFO) models are arguably a relatively more complete model, albeit utilising some fairly restrictive assumptions. Surprisingly there are few tests in the literature examining such an issue, much less in an IPO environment. Whilst the proposed study does not test the various models directly, the value relevance approach taken does allow for a statement to be made about the likely effectiveness of the models. Secondly, a review of the literature reveals some inconsistency over the relevance of management earnings forecasts in IPO disclosure documents. The proposed study seeks to resolve this issue. Thirdly, with respect to the EBFO models, there is much criticism of some of the assumptions that are made as well as those additional assumptions that researchers have made in implementing the model empirically. This primarily concerns the linear information dynamic assumed between earnings of different periods and the corresponding effect on the value relation. This study seeks to assess whether such concerns are present within the Australian IPO context. The following section will discuss the literature relevant for the research question outlined above. The subsequent section will develop the testable hypothesises and explain the research method to be used. 2.0 Prior literature In order to lay the foundation for the rest of the paper, two key areas of the literature will now be discussed in some detail; accounting-based valuation and IPOs. 2.1 Valuation There is a vast literature on valuation across three distinct disciplines; accounting, finance and economics. For the purpose of this study, the accounting-based valuation models are of most interest. The three models to be empirically tested in this study are the EBFO models, the comparable multiple models and the residual income valuation model. 2.1.1 EBFO models The EBFO models attempt to explicitly link accounting information to equity valuation through a specific dynamic process. These models start with the basic and widely acknowledged proposition that the equity value of a firm is the summation of the discounted future stream of dividends, the Present Value of Expected Dividends (PVED) model. With the aid of the clean surplus assumption, it is possible to convert the PVED model into a model that accommodates the current book value of equity and abnormal earnings as being the 2 determinants of value. The clean surplus relation states that any change in book value must be attributed to either earnings or to net distributions (i.e. dividends). This particular formulation is not new in the literature, with similar models dating back to the beginning of last century (see Bernard 1995 for more background information). What makes the EBFO models a novel contribution to the literature is the explicit link between value and accounting information that the linear information dynamic provides. The linear information dynamic provides for the relationship between abnormal earnings of different periods. Specifically the relationship between the year ahead abnormal earnings is defined as the current abnormal earnings multiplied by a persistence factor plus a factor for ‘other information’ and an error term. The ‘other information’ variable is similarly defined in an autoregressive-one process 1 . The persistence parameters are then used to estimate firm value in a closed-form process. This separates the EBFO model from the residual income valuation model that will be discussed below. Despite widespread praise of the original Ohlson (1995) model and subsequent efforts, the empirical evidence on the EBFO models is mixed. Dechow, Hutton and Sloan (1999) found limited support the EBFO modelling developed by Ohlson (1995) and extended in the intervening time. However further studies have found fault with the underlying methodology of the EBFO modelling. Criticism has centred mainly on the linear information dynamic. Whilst linear information dynamics allows for relatively parsimonious analysis, there is evidence that such dynamics are not appropriate 2 . Burgstahler and Dichev (1997) found for instance that equity is a convex function of earnings (book value), holding book value (earnings) constant. That is, the relevance of earnings and book value depends on the relative level of earnings and book value. This makes sense intuitively as firms with low or negative earnings will seek to resurrect their fortunes and thus current earnings would not be expected to be a good indicator of firm value, which is contingent on future events. It is not clear how this affects IPOs, characterised as they are by the fact that many such firms seek additional capital to enlarge current operations i.e. raising funds to continue with current production technology as opposed to pursuing an alternate process. In a more direct test of the EBFO models, Myers (1999) did not find support for linear information dynamics. Myers calculated firm value using variations on the EBFO model and assessed the estimated parameters against actual parameters and found a significant variation. Interestingly, Myers also found that the linear information dynamic of the EBFO models did not provide significantly better estimates than book value alone. Following on from Burgstahler and Dichev (1997) and Myers (1999) a number of studies have attempted to develop non-linear information dynamics, based on real-options 3 . The ‘other information’ component in the EBFO models is often omitted due to this variable being a priori unknown. Alternatively limited metrics are used as proxy for this ‘other information’ e.g. Myers (1999) used order backlogs as the proxy for ‘other information’, although Callen and Segal (2005) suggest that this is not a particularly representative surrogate. Whilst initially left undefined, later papers discussed methods of estimating this variable using currently available data, namely earnings forecasts (Liu and Ohlson 2000; Ohlson 2001). More recent models in the EBFO literature have accounted specifically for ‘other information’ (e.g. Begley and Feltham 2002). Furthermore, the study by Callen and Segal (2005) suggests that inclusion of an appropriate ‘other information’ variable can 1 Although there have been some attempts to introduce multi-period lags e.g. Morel (1999) The EBFO model does not rest on linear information dynamics opposed to any other information dynamic being used (Christensen and Feltham 2003). Thus appropriate modifications may be made whilst retaining the original model’s intuition. 3 See for example Yee (2000), Zhang (2000), Biddle, Chen and Zhang (2001), and Ashton, Cooke, and Tippett (2003). 2 3 significantly improve the power of the model, as they find some support for the EBFO model using the suggestions by Liu and Ohlson (2000) and Ohlson (2001). Although the EBFO model does not restrict the number of ‘other information’ variables, to avoid cumbersome analysis, this variable is typically kept to just one representation. 2.1.2 Comparable multiples Comparable multiples are a popular method for estimating firm value due to their simplicity as well as their uncanny accuracy (Richardson and Tinaikar 2004). Multiples for instance do not require the explicit estimation of cost of capital and streams of future dividends or abnormal earnings and are not subject to the same level of very strict assumptions of other valuation models. However this is not to say that multiples lack an underlying theory. There are a number of multiples in common use including the more prominent price-earnings and market-to-book ratios. Other multiples that may be used include sales-to-book, forecast earnings-to-price, and cash-to-price 4 . The use of comparable multiples is predicated on the idea that the relevant ratio captures the market’s assessment of both risk and growth prospects (Zarowin 1990). This is derived from the development of the various models. For example, the price-earnings ratio is based on the assumption that capitalised future earnings are sufficient for estimating price, and thus implicitly make assumptions about risk and growth. Zarowin (1990) did not find evidence that multiples were associated with risk, whilst finding only modest evidence for a significant association with growth. However, Zarowin used Sharpe-Lintner betas to proxy for risk and there has been considerable evidence that such measures of risk are deficient (Fama and French 2004). There does not appear to be any subsequent study that has assessed multiples with respect to other estimations of firm-specific risk. 2.1.3 Residual income valuation model The residual income valuation (RIV) model defines the value of a firm as the summation of the book value of the firm and the discounted stream of abnormal earnings of the firm. Typically a number of periods of abnormal earnings would be explicitly forecasted and discounted before a final forecast is set in an infinite series, which is to be distinguished from the closed form EBFO model. As indicated in section 2.1.1, the residual income model has been present in the literature for some time. Whilst technically any (appropriate) numbers may be substituted for book value and earnings – there is no restriction that dictates that these two variables be book value and earnings – as in this instance accounting summary measures are substituted, this model is labelled an accounting based valuation model. The RIV model relies on forecasts of future earnings capturing all the information relevant in estimating the value of the firm. 2.2 Initial Public Offerings (IPOs) As identified previously, IPOs are characterised by a higher level of uncertainty than other events that might occur in a firm’s life. This makes sense given that in contrast to some of the other major events in the firm’s life identified in the introduction, an IPO is an instance where the firm, coming from relative obscurity, must convince the market to invest. Prior to the IPO, many firms are almost unknown in investor circles and thus there is a high level 4 Liu, Nissim and Thomas (2002) elaborate on further variations. 4 uncertainty surrounding the typical issue 5 . Given the large information asymmetry between promoters and investors, there is a large role for signalling information in the valuation process. The following paragraphs will briefly discuss some significant issues vis-à-vis IPOs in the provision of management forecasts, IPO valuation, and the role of signalling in IPO valuation. 2.2.1 Management earnings forecasts Under Australian corporate law 6 IPO promoters are required to disclose all the information that investors would reasonably require to make an informed decision about whether or not to participate in the share offer. Past studies have shown that provision of management forecasts are almost the de facto standard 7 . However a recent study suggests that the introduction of legislation in 2000 8 , which relaxed the litigation risk faced by promoters that disclose forecasts, resulted in less frequent disclosure subsequent to the introduction of the legislation (Chapple, Clarkson, and Peters 2005). This is consistent with much of the voluntary disclosure literature (e.g. Verrecchia 1983) that proposes fairly intuitively that firms will disclose when the costs of non-disclosure are more than the costs of disclosure 9 . In any case the fact that the management of some IPO firms do disclose management forecasts would seem to indicate that they at least perceive some value in doing so. The evidence however is relatively mixed. Clarkson, Dontoh, Richardson and Sefcik (1992) and Firth (1998), in a Canadian and Singaporean environment respectively, found a significant association between earnings forecasts and value. Firth in fact uses a modified version of Clarkson’s et al model and interacted earnings forecasts with a measure of retained ownership and finds this to be significantly related to firm value also. This would seem to suggest that the act of providing an earnings forecasts induces the firm to emphasize that such a signal is credible by retaining more of the shares in the company. However contrary to the studies listed above, How and Yeo (2001) did not find earnings forecasts to be value relevant. Whilst How and Yeo (2001) performed their test in an Australian environment in contrast to the previous studies, the sample was taken from the period before the introduction of the CLERP 1999 legislation. Given the similar between-country litigation environments at that time this does not appear to be the cause of differentiation. Chapple et al (2005) find significant but relatively weak results for the value relevance of earnings forecast in contrast to How and Yeo (2001) in the period before the introduction of CLERP 1999. Following the introduction of the legislation this relationship is found to be significantly stronger. Perhaps, as Chapple et al (2005) note, this is due only to firms that have a reasonable basis for disclosing forecast earnings doing so. 5 There are of course exceptions where well known firms have publicly listed and where analyst following would be considerably higher and consequently there would be less uncertainty about the firms future prospects. 6 In particular s710 of the Corporations Act 2001 7 See for example How and Yeo (2001) and Lee, Taylor and Taylor (2006) 8 Corporate Law Economic Reform Program Act 1999 – amongst other amendments, this act amended the Corporations Law with respect to liability for inadequate disclosure and the oversight role of ASIC. 9 Here the costs of disclosure and non-disclosure include for example proprietary costs through revealing information to competitors, changes in firm cost of capital, and normal charges associated with disclosure such as audit fees and so forth. Strictly speaking a reduction in litigation risk should increase the probability of disclosure. However as Chapple et al (2005) discuss, the reduction in litigation risk applied to omissions as well as inclusions (i.e. provision of forecast). A reduction of risk with respect to omissions should affect a decrease in disclosure. As Chapple et al found that frequency of disclosure had dropped significantly this would suggest that the later effect is more powerful, and is thus consistent with the discretionary disclosure literature. 5 2.2.2 IPO valuation Industry practice for valuing IPOs favours the comparable multiples approach outlined in section 2.1.2 above due to the simplicity in use and the availability of required data (Kim and Ritter 1999). However recent studies have found that such models have relatively poor explanatory power (Kim and Ritter 1999 in the USA; How, Lam, and Yeo 2007 in Australia). Although as Kim and Ritter (1999) noted, it would be concerning if the explanatory power was very high. Even though comparable firms are chosen due to their similarities to the IPO firm, the firm under analysis would have many firm-specific factors that distinguish it from the comparable firm. Cotter, Goyen and Hegarty (2005) performed a direct test of various valuation models in an IPO context including the price-earnings multiple and the residual income valuation model 10 and subsequently found that neither model provides a particularly accurate estimate of IPO price. There does not appear to be any direct test of the EBFO model in an IPO context in the literature. This is not altogether surprising due to the complexities involved in calculating the parameters required to estimate value using the model in a ‘seasoned’ market setting, notwithstanding the assumptions necessary to translate the EBFO model to an IPO setting 11 . 2.2.3 Signalling and IPOs It has been mentioned a few times in this paper that IPOs are characterised by a large degree of information asymmetries. Such an environment induces the promoters of the IPO to signal information about the quality of their offering as otherwise investors may find it difficult to differentiate between ostensibly similar share offerings. Signals however are only credible where creation is costly 12 . The provision of management earnings forecasts is an example of such a signal, albeit one that appears to be open to more abuse than other signals, perhaps due to a lag between any deception (intentional or otherwise) and punishment, but also due to the difficult task of providing a forecast about future uncertain events. There are heavy penalties for firms that disclose forecasts that are later deemed to be misleading. However, whilst the earnings signal is significant, it is likely that there are a great many other signals that affect firm value in concert with earnings forecasts, both in modifying type relationships such as in Firth (1998) as well as by themselves. A number of other signals have been proposed in the literature that would be expected to have a significant impact on price. These factors include underwriter prestige (Krinsky and Rotenberg 1989), audit quality (Feltham, Hughes, and Simunic 1991), other institutional choices, and retained ownership (Leland and Pyle 1977). However these relationships are not necessarily straight forward as illustrated by Lee, Stokes, Taylor and Walter (2003) with respect to the auditing signal. 3.0 Hypothesis development and method This section will develop the hypothesises corresponding to the research questions outlined in the introduction and the method to be employed to carry this out. The research takes a value relevance approach to testing the various models outlined above. The general models will be 10 Whilst Cotter et al (2005) label the residual income model under the EBFO tag, the model they employ does not contain any information dynamic that is characteristic of the EBFO models. 11 See Myers (1999) for an example of the steps involved. Note that Myers uses some fifteen years of data to estimate the required parameters. This is clearly not available for IPOs, and thus for an operationalisation of this model in an IPO context it is likely that market data would have to be used as a proxy for missing IPO data. 12 See Riley (2001) for a comprehensive review of the signaling and screening literature. 6 presented first, followed by a discussion of key empirical issues including scaling and collection of data. The first research question concerns the apparent effectiveness of various accounting-based valuation approaches in valuing IPOs. As discussed above, whilst the comparable multiples approach is widely endorsed in the literature, there does not appear to be many studies that explicitly test this approach with respect to other methods available 13 . As discussed above, support for the comparable multiples method is through their simplicity and the ability to avoid explicit forecast of a series of earnings and calculation of appropriate discount rate in apparent contrast to the other models outlined in this paper. Despite this notable attraction, the multiples method lacks any detailed modelling of the behaviour of summary accounting information that characterises, in various degrees, the other two models outlined in this paper. The EBFO model, as discussed previously, contains an ‘other information’ component. In most modelling this is featured as a single variable. However as Myers (1999) demonstrated, additional ‘other information’ variables may be modelled with a corresponding increase in complexity. Based on the discussion about signalling factors in the IPO environment, it is appropriate to augment these models with additional factors that are expected to determine value. Many of the following factors would only be expected to influence the initial value and would not persist for long. In the review of prior literature above it was noted that later formulations of the EBFO model suggested that forecasts of future earnings may be used to derive ‘other information’. For instance, the difference between current earnings and one period ahead forecasted earnings would be expected to give some insight into the persistence of current projects and growth of future projects (Begley and Feltham 2002). Other factors are also envisioned to have some informational role in determining value. It is suggested that given auditors have a lot to lose by not performing a sufficiently tight audit, they will not allow the client firm to pursue overly aggressive accounting policies (Feltham, Hughes, and Simunic 1991). This would be expected to induce investors to place more reliance on accounts audited by high quality auditors. A similar case may be made for underwriters, who in their presence convey a degree of certainty about the offer (Hughes 1986). Leland and Pyle (1977) suggest that the level of retained ownership sends a signal to prospective investors. Specifically the willingness to retain some ownership in the firm indicates that they believe future value will be forthcoming i.e. they are all in the same boat. The operating history of a firm indicates that firm’s ability to operate with a degree of success over a period of time and provides an important signal to potential investors. Also, under the same general disclosure section explained above, IPO firms are expected to disclose risk factors faced by the firm. Thus the number of risk factors identified in the disclosure document is expected to influence the investor’s beliefs about firm value. A further measure for uncertainty in the post-issue leverage of the IPO firm, defined as total liabilities over total assets. A direct analysis of the various models would involve directly calculating the various models, estimating the parameters of the EBFO model for instance. However this study takes a value relevance approach. Part of the reason for this decision is time constraints. Whilst estimating the parameters is certainly possible for IPOs, the processes involved are quite complex and time consuming across the entire sample, performing the same process over one or two cases 13 There are however studies that test the comparable multiples model in isolation e.g. Kim and Ritter (1999) and How et al (2007). Cotter et al (2005) extends this somewhat by also investigation the RIV model. 7 however would be less demanding 14 . In assessing the accounting-based valuation models, the models are split into simple and complex models. Simple models are the comparable multiples models, whereas complex models are the RIV and EBFO models. As the RIV models are essentially an earlier stage in the development of the EBFO model, it is appropriate to characterise models in this way. Such characterisation allows the first hypothesis to be stated: H1: Complex models outperform simple models of accounting-based valuation in estimating the value of Australian IPOs. Initially it would be instructive to examine the models with respect to their original theoretical development i.e. leaving the multiples without additional information. As discussed above the EBFO model actively incorporates such elements. However this does not say much about the particular models as only very rarely would such implementation be used in practice and thus using these results to reject or not reject H1 would be akin to setting up a straw man to be knocked down. Cotter et al (2005) suggested that investment banks take a variety of factors into consideration alongside the comparable multiple. Thus, to truly test the comparable multiples model, it is necessary to include the range of other factors identified above with respect to ‘other information’ in the EBFO model. Pt = α1M + α2SEG + α3MEG + α4AUD + α5UND + α6LP + α7HIST + α8RISKF + α9LEV + ε (1) 15 Pt = δ1bvt + δ2xat + δ3ΔF1 + δ4ΔF2 + δ5AUD + δ6UND + δ7LP + δ8HIST + δ9RISKF + δ10LEV + ε (2) where P = either the offer or end of first day listing price, M = relevant multiplier (i.e. either earnings, one-period ahead earnings, book value, and sales), bv = book value, xa = abnormal earnings, SEG = Short-term earnings growth, MEG = Medium-term earnings growth, ΔF# = change between # period abnormal earnings and the prior period, auditor (AUD), underwriter (UND), retained ownership measured as in Leland and Pyle (1977) (LP), operating history (HISTORY), risk factors identified in report (RISKF) and a measure of leverage (LEV defined as total liabilities divided by total assets). The second model above ostensibly tests for the EBFO model, although significant coefficients on delta one and two also provide support for RIV model as well, albeit imprecisely. There is however another way to distinguish between the EBFO and RIV models and determine which of the models provides a better explanation of value. Firstly it is necessary to recall the above discussion about the development of the EBFO model. This development led to a close-form valuation model being the function of several factors including book value, abnormal earnings, ‘other information’ and the associated persistence factors on abnormal earnings and ‘other information’. It is thus suggested that these 14 It is important to stress that across all the models examined in this study a number of assumptions are taken in order to facilitate an empirical analysis. These assumptions would be expected to be improved upon in either a more focused study or in practice and would subsequently improve the power of the respective models. 15 An intercept is not included in this model. Whilst, and as discussed later, it would be expected that there would be other information that would be associated with valuation in combination with the multiple, at this stage this model is testing this multiplier alone. Also note that here the coefficient on M is effectively the multiplier in question. This model will have to be adjusted in order to reflect industry variation in multipliers. 8 persistence factors communicate a more sophisticated understanding about valuation than the simpler RIV model. In order to test this assertion, following Begley and Feltham’s (2002) approach, capitalised earnings are calculated with this measure replacing delta’s two through four of model two. This is then compared with a formulation of the model as presented in model two. The model with the capitalised earnings then effectively identifies as a simple RIV type analysis. It is expected that the EBFO type analysis will yield a better coefficient of determination (r2), indicating the value of modelling persistence and growth as is the case in EBFO type models. It is further possible to provide some statement about the likely fit of linear information dynamics in the IPO case. As outlined a few times now, EBFO modelling allows for a closed-form valuation model to be established, with abnormal earnings for instance being multiplied by a function of various persistence factors and the firm’s discount rate. This structure allows the coefficients found in the regression analysis to be reversed out in order to obtain estimates for the persistence parameters. In the EBFO modelling several restrictions are placed on these parameters in order to reflect certain assumptions e.g. the persistence on abnormal earnings is restricted to less than one, corresponding to the pattern expected of abnormal earnings (declining over time due to competition). Thus it is possible to assess how likely the implied parameters are given the discount rate, and thus to give some statement about the apparent ability of the linear information dynamic to describe the valuation process. This leads to the second hypothesis: H2: The coefficients in the regression testing the EBFO model are within the range of expectations given linear information dynamics. As discussed above in section 2.2.1, there seems to be some conflict in the literature about the relevance of earnings forecasts. Clarkson et al (1992) and Firth (1998), in foreign settings find that earnings forecasts are value relevant. However in an Australian study How and Yeo (2001) found no relevance once several risk factors have been controlled for. Chapple et al (2005) using a similar test design as How and Yeo and again in an Australian environment found earnings forecasts to be value relevant either side of the introduction of CLERP 1999. This is an odd discrepancy as the voluntary disclosure literature suggests that the firm would only provide the forecast if it was in some way beneficial to the firm. This discrepancy makes a return to this issue worthwhile. The inconsistency also perhaps points to a more complex relationship between earnings forecast and value. For example Firth (1998) suggested that the ownership signal is incremental to the forecast information i.e. the importance of the retained ownership signal is contingent on a forecast being provided. Following this intuition, there may be a number of factors have a modifying relationship with earnings forecasts. In fact, any of the factors identified above in model two could potentially interact with the provision of earnings forecasts in such a relationship. From this the following hypothesises are formed: H3a: Management earnings forecasts are value relevant. H3b: There is a modifying relationship between some of the factors identified in models one and two and provision of management earnings forecasts. In order to test H3a and H3b, model two is used as a base; the continuous variables for abnormal earnings are replaced with categorical variables (one for presence and zero for absence), leading to model three presented below. To test the respective hypothesises the appropriate variables will be included or excluded i.e. in order to test H3a the modifying relationship (attached to zeta six) will be dropped from the following regression model: Pt = ζ1bvt + ζ2xat + ζ3EF1 + ζ4EF2 + ζ5LP + ζ6Yh*EF1 + ζ7AUD + ζ8HIST + ζ9UND + ζ10GROWTH + ζ11RISKF + ζ11LEV + ε (3) 9 In section two, the literature review, a few paragraphs were spent on discussing the linear information dynamic as it is this component of the EBFO modelling that appears to have become the leading criticism of the model. The essential message from this literature was that if a firm is performing poorly currently it is unlikely they will for too much longer as it would be expected that the firm would pursue some course of action aimed at performing better in the future. Thus the weighting placed on earnings or book value for instance depends on the firm’s relative performance. Burgstahler and Dichev (1997) found such a relationship as discussed above. However this relationship was found amongst firms that had already listed on a stock exchange. IPO firms however might be expected to be different. IPO firms after all are raising funds with the aim to increase the scale of their operation amongst other things. That is IPO firms are firms that one might categorise as highly adaptive firms, in the Burgstahler and Dichev (1997) terminology. If this is the case then, when adopting a similar test to Burgstahler and Dichev there should be no significant difference between the partitioned variables in the equation below. The equations are followed by the associated hypothesis. P/BVt-1 = γ1 + γ2DM + γ3DH + γ4(xat/bvt-1) + γ5(xat/bvt-1)DM + γ6(xat/bvt-1)DH + ε (4) P/xat = λ1(bvt-1/xat) + λ2(bvt-1/xat)DM + λ3(bvt-1/xat)DH + λ4 + λ5DM + λ6DH + ε (5) where Dx = dummy indicating medium or high (low implied) book value or earnings respectively for model five and six. H4: Using the approach by Burgstahler and Dichev (1997) there is no significant difference between the different partitions indicating no convex relationship on these terms. However this is not to suggest that IPOs are not characterised by non-linear dynamics at all. There might be a number of factors that interact with earnings or book value in a non-linear fashion. There are a number of other possible factors that might cause the relation between value and earnings and book values to be non-linear. At this stage of the research project these factors are still being analysed and so this process will not be discussed here in great detail. Work to date has involved investigating the possible role of distinguishing between tangible and intangible assets as a basis of some non-linear relation, either through a partition similar to models four and five above or through separating book value into those two components and investigating that effect. Similarly the possibility of higher order variables is being investigated with the aim of producing a testable hypothesis and corresponding model. 3.1 Cost of capital Both the RIV and EBFO model require the use of abnormal earnings. Abnormal earnings as its title suggests are earnings in excess of normal earnings. Abnormal earnings are calculated as earnings before interest and tax less a capital charge reflecting both the cost of debt and equity. Interestingly many studies investigating the EBFO models use blanket discount rates (e.g. Dechow, Hutton, and Sloan 1999; Begley and Feltham 2002). This is either done through simply declaring a rate across section and time or by assuming a common market premium and allowing the risk-free rate to change through time. Whilst sensitivity tests performed in some of these analysis’s indicate that this is not a major concern, Beaver (1999) noted that given that risk is known to vary by firm, a more realistic representation of risk is called for and would add more explanatory power to the models. An alternative to the blanket approach is to let cost of capital vary across industry using industry betas as Cotter et al (2005) did. Whilst not perfect, this method does introduce some 10 variation across the sample in any one time period. Calculating beta for individual IPOs is clearly not feasible as beta is of course a measure of the firm’s movement with the market, which IPO firms do not, by definition, have a history of. Another issue overlooked by researchers is the cost of debt. CAPM or similar models calculate the cost of equity, however a firm’s cost of capital is a composite of the cost of debt and equity. Thus researchers implicitly assume that all the companies in the sample are all-equity companies. At this stage the source and type of cost of capital is yet to be determined and will be dependent on the ability to collect the data of over the sample period. Despite the apparent weakness of some concepts of risk it is likely that some trade-off will be necessary in order to conduct empirical tests. 3.2 Scaling issues In research of this type (EBFO and RIV models) scale issues become a concern. As researchers attempt to isolate effects using large data sets it becomes apparent that size appears to be a driver in large amount of events – a not altogether unsurprising conclusion. Therefore, given this is not the factor under analysis, it is important to control for this effect. Barth and Clinch (2007) identified types of scale effects that can impede statistical inferences in capital market tests; multiplicative and additive correlated omitted variables, regression parameters that vary with scale, survivorship and scale-related heteroscedasticity. Barth and Clinch (2007) found that none of the currently proposed diagnostics for detecting scaling issues is particularly effective at doing so. In testing different specifications against simulated data they find that deflating by number of shares performs comparably better than other specifications, although this method does not dominate the other methods. Importantly for this study this specification performed well in mitigating additive scale issues. An IPO is where a firm raises capital and so it is of no surprise that book value affected by the issue will be related to the value of the firm – this is the additive scale problem. Consequently this specification will be used throughout the empirical tests 16 . Other specifications introduced in Barth and Clinch will be used for completeness, although will not comprise the main part of analysis. 3.3 Data Much of the data required for the above analysis will be hand-collected off the relevant IPO prospectus. IPO were identified through examining the Australian Stock Exchange (ASX) new listings. Mining firms were excluded as mining companies typically do not disclose earnings forecasts as other components of the prospectus are more important to potential investors, specifically the geological report. Trusts and pooled development funds were also excluded as these firms have a different structure. The sample period under analysis is from July 2001 till June 2005. A cursory examination suggests that in excess of 200 observations are available in this time frame once the exclusions have been taken into account. The exceptions to collecting data from disclosure documents are the measures of industryspecific risk and the measure of the risk free rate. Currently a source for the industry risk measure has not been found whilst the risk free rate will be one of either the rate of return on thirty-year bonds or the cash rate. Large accounting firms are so identified as being one of the big four/five i.e. Ernst and Young, Deloitte Touche Tohmatsu, KPMG, Arthur Anderson or PricewaterhouseCoopers. Prestigious underwriting firms will be identified as firms that have a national presence and/or 16 As the firms under analysis are IPOs, post-listing shares will be used. 11 that are bank-backed. It is unclear at this stage how many underwriters will then be categorised as not prestigious. A cursory overview of the prospectus documents within the sample indicates that those that employ underwriters generally seem to employ those firms that the above categorisation would label prestigious firms. This appears to be less the case with mining firms that seem to employ more localised underwriting firms. However, as discussed above, mining firms are excluded from the analysis. 4.0 References Ashton, David, Terry Cooke, and Mark Tippett. 2003. An Aggregation Theorem for the Valuation of Equity Under Linear Information Dynamics. Journal of Business Finance and Accounting 30 (3-4):413-440. Barth, Mary E, and Greg Clinch. 2007. Scale effects in capital markets-based accounting research: Stanford University. Beaver, William H. 1999. Comments on 'An empirical assessment of the residual income valuation model'. Journal of Accounting and Economics 26:35-42. Begley, Joy, and Gerry Feltham. 2002. The relation between market values, earnings forecasts, and reported earnings. Contemporary Accounting Research 19 (1):1-48. Bernard, Victor. 1995. The Feltham-Ohlson framework: implications for empiricists. Contemporary Accounting Research 11 (2):733-747. Biddle, Gary C, Peter Chen, and Guochang Zhang. 2001. When capital follows profitability: non-linear residual information dynamics. Review of Accounting Studies 6:229-265. Burgstahler, David C, and Ilia D Dichev. 1997. Earnings, adaptation and equity value. The Accounting Review 72 (2):187-215. Callen, Jeffrey L, and Dan Segal. 2005. Empirical tests of the Feltham-Ohlson (1995) model. Review of Accounting Studies 10 (4):409-429. Chapple, Larelle, Peter M. Clarkson, and Christopher J. Peters. 2005. Impact of the Corporate Law Economic Reform Program Act 1999 on initial public offering prospectus earnings forecasts. Accounting and Finance 45 (1):67-94. Christensen, Peter O, and Gerry Feltham. 2003. Economics of Accounting. Edited by J. S. Demski. Vol. 1 - Information in markets, Kluwer Series in Accounting Scholarship. Massachusetts: Kluwer Academic Publishers. Clarkson, Peter M, Alex Dontoh, Gordon D Richardson, and Stephan E Sefcik. 1992. The voluntary inclusion of earnings forecasts in IPO prospectuses. Contemporary Accounting Research 8 (2):601-626. Cotter, Julie, Michelle Goyen, and Sherryl Hegarty. 2005. Offer pricing in Australian industrial public offers. Accounting and Finance 45:95-125. Dechow, Patricia M, Amy P Hutton, and Richard G Sloan. 1999. An empirical assessment of the residual income valuation model. Journal of Accounting and Economics 26:1-34. Fama, Eugene F, and Kenneth R French. 2004. The Capital Asset Pricing Model: Theory and Evidence. The Journal of Economic Perspectives 18 (3):25. Feltham, Gerry, John S Hughes, and Dan A Simunic. 1991. Empirical assessment of the impact of auditor quality on the valuation of new issues. Journal of Accounting and Economics 14 (4):375-399. Firth, Michael. 1998. IPO profit forecasts and their role in signalling firm value and explaining postlisting returns. Applied Financial Economics 8 (1):29-39. How, Janice, Jennifer Lam, and Julian Yeo. 2007. The use of the comparable firm approach in valuing Australian IPOs. International Review of Financial Analysis 16 (2):99-115. How, Janice, and Julian Yeo. 2001. The impact of forecast disclosure and accuracy on equity pricing: the IPO perspective. Journal of Accounting, Auditing and Finance 16 (4):401-425. Hughes, Patricia. 1986. Signalling by direct disclosure under asymmetric information. Journal of Accounting and Economics 8:119-142. Kim, Moonchul, and Jay Ritter. 1999. Valuing IPOs. Journal of Financial Economics 53 (3):409-437. Krinsky, I., and W. Rotenberg. 1989. The valuation of initial public offerings. Contemporary Accounting Research 5 (2):501-515. Lee, Philip J, Donald Stokes, Stephen Taylor, and Terry Walter. 2003. The association between audit quality, accounting disclosures and firm-specific risk: Evidence from initial public offerings. Journal of Accounting and Public Policy 22 (5):377-400. 12 Lee, Philip J, Sarah J Taylor, and Stephen L Taylor. 2006. Auditor conservatism and audit quality: evidence from IPO earnings forecasts. International Journal of Auditing 10:183-199. Leland, Hayne E, and David H Pyle. 1977. Informational asymmetry, financial structure, and financial intermediation. Journal of Finance 32 (2):371-387. Liu, Jing, Doron Nissim, and Jacob Thomas. 2002. Equity Valuation Using Multiples. Journal of Accounting Research 40 (1):135-172. Liu, Jing, and James A Ohlson. 2000. The Feltham-Ohlson (1995) Model: empirical implications. Journal of Accounting, Auditing and Finance 15 (3):321-331. Morel, Mindy. 1999. Multi-Lagged Specification of the Ohlson Model. Journal of Accounting, Auditing & Finance 14 (2):147-161. Myers, James. 1999. Implementing residual income valuation with linear information dynamics. The Accounting Review 74 (1):1-28. Ohlson, James A. 1995. Earnings, book values, and dividends in equity valuation. Contemporary Accounting Research 11 (2):661-687. ———. 2001. Earnings, book values, and dividends in equity valuation: an empirical perspective. Contemporary Accounting Research 18 (1):107-120. Richardson, Gordon D, and Surjit Tinaikar. 2004. Accounting based valuation models: what have we learned? Accounting and Finance 44:223-255. Riley, John G. 2001. Silver Signals: Twenty-five years of screening and signalling. Journal of Economic Literature 39 (June):432-478. Verrecchia, Robert E. 1983. Discretionary Disclosure. Journal of Accounting and Economics 5 (3):179194. Yee, Kenton K. 2000. Opportunities knocking: residual income valuation of an adaptive firm. Journal of Accounting, Auditing and Finance 15 (3):225-270. Zarowin, Paul. 1990. What Determines Earnings-Price Ratios: Revisited. Journal of Accounting, Auditing & Finance 5 (3):439-454. Zhang, Guochang. 2000. Accounting information, capital investment decisions, and equity valuation: theory and empirical implications. Journal of Accounting Research 38 (2):271-295. 13