The Role of R&D Capitalisations in Firm Valuation and Performance Measurement.
We investigate the value-relevance of capitalised R&D on the balance sheet, and the extent to which R&D accruals improve the association between accounting-based measures of firm performance and capital market returns for Australian listed companies. This is a regulatory setting where management discretion in the capitalisation decision is permitted and can be empirically observed. Our results suggest that capitalised R&D on the balance sheets of selective capitalisers is value-relevant; that is, the ability of capitalised R&D to explain information contained in prices (given information conveyed by other components of the balance sheet) is statistically significant. For the same group of firms R&D accruals (particularly the initial capitalisation) improve accounting earnings as a measure of performance but only for the pooled sample using contemporaneous returns. The results for the fully expensing sample are less clear, perhaps due to the small sample size.
RESEARCH AND DEVELOPMENT; ACCOUNTING ACCRUALS; VALUE RELEVANCE; PERFORMANCE MEASURES.
1. Introduction and Background
This paper investigates the usefulness of research and development (R&D) cost accruals in the valuation and performance measurement of Australian firms which apply the discretion to capitalise R&D accorded by AASB 1011 (Australian Accounting Standards Board 1987). It informs the ongoing debate--to continue or disallow this discretion--through an empirical investigation of two questions. First, is the unamortised balance of R&D on the balance sheet value-relevant? That is, is the information conveyed by capitalised R&D consistent with the information set used by the market to value the firm at balance date. Specifically, we test for an association between capitalised R&D and market value of the firm to establish whether capitalised R&D is `believed' to be an asset by the market. The second question is: do R&D accruals improve accounting-based measures of firm performance relevant to investors? To address this, we investigate whether R&D accruals improve the association of accounting measures of performance with stock returns.
Our findings, while tentative, suggest that capitalised R&D is value-relevant and that R&D accruals (and the R&D capitalisation accrual in particular) do increase the association between accounting-based and market-based performance measures. These results are consistent with arguments against the proposed limitation of management discretion, and raise the possibility that such reform may impose additional investor communication costs.
Accounting for R&D costs has long been a controversial issue (Stickels 1996; Lev & Sougiannis 1996; Lee & Sami 1995). While R&D activity can be viewed as an investment of time and money that creates genuine future economic benefits, it can also represent highly risky business strategies with uncertain future benefits. The difficulty in predicting future benefits for particular expenditures arises out of both the inherent uncertainty of R&D programs, and the nature of the commercial marketplace where even `successful' R&D ventures may not prove profitable due to advances in competitors' technologies (Stickels 1996). The accounting controversy centres on whether company management should have the discretion to capitalise any R&D costs.
Allowing discretion in accounting treatments can facilitate the reduction of information asymmetries between management and investors--particularly in industries with high proprietary costs of disclosure, such as those with high levels of R&D (Healy & Palepu 1993, 1995; Horwitz & Kolodny 1981; Clinch 1991). Under this scenario, discretion to capitalise and amortise R&D costs may provide superior measures of firm value and performance compared with a requirement to immediately expense (Lev & Sougiannis 1996; Henderson & Pierson 1995, pp. 423-26).
Mitigating against the credibility of management choice are other features of the financial reporting environment that give rise to managerial opportunism (Healy & Palepu 1993, 1995). These include debt- and equity-related incentives to manipulate accounting numbers. The former arise from the structure of debt contracts; where companies have debt covenants requiring the maintenance of certain income and asset ratios, accounting accruals can be manipulated to reduce the probability of default when the firm is close to specified limits (DeFond & Jimbalvo 1994; Watts & Zimmerman 1990).(1) Equity-related incentives to manipulate accruals can arise from the structure of management remuneration schemes, particularly if compensation is formally linked to reported profit through a bonus plan (Smith & Watts 1982; Healy 1985; Dechow & Sloan 1991). Finally, political cost factors can influence the capitalisation decision--for example, large firms can have incentives to reduce reported profits and hence, political `visibility', by expensing R&D.(2) To the extent that the policy choice is influenced by incentives to manipulate the reported accounting numbers, the ability of the R&D accrual to convey information may be reduced.
The difference in perspectives--efficient signalling implying enhanced, and managerial opportunism implying reduced, value-relevance and utility of R&D accruals--is reflected in different international requirements on the treatment of R&D costs. In the US, the Financial Accounting Standards Board requires (through SFAS no. 2 (FASB 1974)) that all R&D expenditures be expensed as incurred. In contrast, current Australian, New Zealand, Canadian, British and International Standards allow management discretion to defer and amortise research and/or development costs if specific criteria are satisfied. Since 1987, AASB 1011 `Accounting for Research and Development Costs' has governed accounting for R&D expenditures in Australia. This legally binding standard requires that all R&D costs be expensed as incurred except to the extent that they are expected to be recoverable beyond any reasonable doubt (paras 30-31). Additionally, the unamortised balance of R&D must be reviewed annually (para. 33), and any excess above the recoverable amount is to be written off against income for that financial year. A `write-off' is distinct from R&D amortised. The converse of the write-off, or the writing back of R&D costs previously expensed, is specifically prohibited (para. 50).
Australian and international standard setters have recently come under pressure to remove management's discretion with respect to R&D costs, and to mandate full expensing in line with the US requirements (Schmidt 1996; Stickels 1996). These calls are supported by the Australian Securities Commission (ASC) which, in a recent review of practice, concluded that firms' R&D capitalisations were often incomplete and arbitrary, with little evidence given to support their decisions (Schmidt 1996). Our research will provide evidence relevant to this debate.
The paper proceeds as follows. Section 2 describes the research method and data. The results are presented and discussed in section 3, and our conclusions appear in section 4.
2. Research Method
2.1 Tests of the Value-Relevance of Capitalised R&D
We consider the market value of equity as a function of balance sheet items to assess the relative contribution (if any) of capitalised R&D (vis-a-vis other balance sheet items) to explain information contained in stock prices. In an efficient market, if R&D expenditures are unrelated to future economic benefits, firm valuation will be unrelated to the level of capitalised R&D expenditures. Conversely, if firm value is a function of capitalised R&D expenditures, it can be concluded that R&D is systematically related to expected future value consistent with arguments for its recognition as an asset.
Following Barth (1994), model la tests the value-relevance of capitalised R&D by regressing the market value of equity (MVE) on the book value of total assets excluding any capitalised R&D (TALRD), the book value of total liabilities (TL) and capitalised R&D costs (RD).
(1a)(3) MVE = [[Alpha].sub.0]+ [[Alpha].sub.1] TALRD + [[Alpha].sub.2] TL + [[Alpha].sub.3] RD + [Mu]
Predicted Signs: + - +
where: MVE = market value of equity (measured at balance date(4), as closing share price multiplied by number of ordinary shares);
TALRD = book value of total assets excluding any capitalised R&D costs;
TL = book value of liabilities; and
RD = book value of capitalised R&D costs.
To mitigate against heteroscedasticity, all variables are deflated by the number of shares outstanding at year-end, adjusted for stock splits and dividends (Barth 1994, p. 6). It is predicted that assets (liabilities) will have a positive (negative) association with the market value of equity. If capitalised R&D is value-relevant, [[Alpha].sub.3] will be significantly greater than zero. Value-relevance would imply that:
(i) management's signalling of their private information about the likely future benefits associated with R&D expenditure is viewed as credible; and
(ii) opportunistic factors do not outweigh the signalling value of the capitalisation decision.
Model 1 a assumes that the relation between the independent and dependent variables is the same for all firms. In practice, this is unlikely to be true due, at least in part, to different asset mixes. There is argument (and limited evidence) to suggest that individual intangible assets are priced differently to each other and to other assets (McCarthy & Schneider 1995, pp. 78-79; Coombes, Otto & Stokes 1996, pp. 16-17). Consequently, model 1b disaggregates total assets into total tangible assets (TTA), capitalised R&D costs (RD) and other intangible assets (IALRD):
(1b) MVE = [[Alpha].sub.0] + [[Alpha].sub.1] TTA + [[Alpha].sub.2] TL + [[Alpha].sub.3] RD + [[Alpha].sub.4] IALRD + [Mu],
Predicted Signs: + - + +
where: TTA = book value of total tangible assets;
IALRD = book value of other intangible assets (total intangibles less capitalised R&D costs); and
all other variables are as previously defined.
The coefficient predictions are consistent with those in model la. The signs of the coefficients on IALRD and RD should be significantly positive if intangible assets are value-relevant.
2.2 Tests of the Usefulness of R&D Accruals in Performance Measurement
Following Dechow (1994) and Guay and Sidhu (1998), we also investigate the extent to which R&D accruals improve the ability of accounting based performance measures to explain market returns. This approach investigates which performance measure has a higher association with market returns.(5) For firms in our sample which capitalise R&D, we compute performance measures which successively remove the effect of the capitalisation decision from accounting earnings. That is, we remove the effect of current period capitalisation, write-offs, and amortisation of previously capitalised R&D from reported income numbers. In this way, we are able to determine the extent to which these accruals add to or detract from performance measurement.
A comparison of models 2a-2d illustrates the relative effects of the removal of R&D capitalised (RDCAP), R&D written off (WROFF) and R&D amortised (AMORT).
(2a) R = [[Alpha].sub.c] + [[Beta].sub.c] NPAT + [Epsilon]',
(2b) R = [[Alpha].sub.c] + [[Beta].sub.c] (NPAT - RDCAP) + [Epsilon]',
(2c) R = [[Alpha].sub.c] + [[Beta].sub.c] (NPAT - RDCAP + WROFF) + [Epsilon]', and
(2d) = [[Alpha].sub.c] + [[Beta].sub.c] (NPAT - RDCAP + WROFF + AMORT) + [Epsilon]]',
where: R = contemporaneous share market annual returns(6) (adjusted for changes in basis of capitalisation);
NPAT = net profit after tax and before extraordinary items (as reported in
the Profit and Loss Statement);
RDCAP = R&D expenditure incurred during the period and deferred;
WROFF = previously capitalised R&D costs written off in the current period;
AMORT = previously capitalised R&D costs amortised in the current period; and
all other variables are as previously defined.
As in Dechow (1994), all accounting variables are measured on a per-share basis and scaled by beginning of period price to avoid spurious correlations due to size and in order to reduce problems of heteroscedasticity.
The stepwise removal of the three main R&D accruals permits an evaluation of their relative contributions to performance measurement. If R&D capitalisation and the relevant accruals are an efficient way of communicating information to the capital market, the stepwise removal of these components from earnings should reduce the utility of the resulting measure of performance (Guay & Sidhu 1998; Dechow 1994; Watts & Zimmerman 1986; Holthausen & Leftwich 1983). This will be reflected in reduced [R.sub.2] for successive models. Conversely, such a result would not be expected if capitalised R&D is predominantly a function of managerial opportunism. Likewise, to the extent that write-offs and accruals are based on arbitrary allocation procedures, the explanatory power of the models 2c and 2d will not be compromised.
The returns-earnings association of firms which immediately expense all R&D expenditure is also of interest because it facilitates an analysis of the extent to which this decision adds noise to the earnings figures of expensing firms. Lee and Sami (1995) and Lev and Sougiannis (1996) find that R&D is not treated as an expense by the market in the United States environment of compulsory expensing. If this situation extends in part to an Australian context (due to the stringent asset recognition requirements in AASB 1011), the utility of the earnings number for R&D expensing firms may increase when adjusted for the effects of the R&D expensing. Models 2e and 2f estimate the returns-earnings association for firms which immediately expense all R&D expenditure. Model 2e examines this association with reported earnings while model 2f estimates the usefulness of adding back R&D expensed to the earnings number.
(2e) R = [[Alpha].sub.e] + [[Beta].sub.e] NPAT + [Mu]', and
(2f) R = [[Alpha].sub.e] + [[Beta].sub.e] (NPAT + RDDEX) + [Mu]',
where: RDDEX = R&D costs directly expensed; and
all other variables are as previously defined.
If RDDEX is predominantly contributing noise to the earnings figure, the [R.sup.2] on model 2f (which effectively removes the effect of RDDEX) will be higher than that for model 2e. Conversely, if RDDEX is `believed' to be a simple period expense by the market, the [R.sup.2] of model 2f will not be greater than that of model 2e.
2.3 Data Collection
There are two main steps in our data collection. The first identifies industries in which R&D expenditure is pervasive and constructs a measure of `R&D pervasiveness'. The second builds a sample comprised of all firms from within those industries selected as having a high score on this measure.
The `R&D pervasiveness' measure is defined as the percentage of firms within a 3-digit Australian Stock Exchange (ASX) industry code that include `research and development' in their Datadisc text in June 1996.(7) While Datadisc includes all currently listed (and some delisted) Australian companies, it does not include the `Notes to the Accounts' that form part of the financial statements of Australian listed companies. As a consequence, the Datadisc search produces a list that differs from the ideal sample selection criteria in two key respects. First, it does not reveal companies that only disclose R&D information in the `Notes to the Accounts', thereby missing companies of interest. Second, it selects companies where the term is mentioned in the Datadisc `text', but where there is no entry in the company's accounts, thereby including companies with no R&D expenditure. For these reasons, a Datadisc search is only used to identify industry groups where R&D is pervasive. Included in the resulting Datadisc list are 528 firms, 161 of which were delisted as at June 1996. The 3-digit industry codes for the remaining 367 firms were collected from the ASX Industry Classification Report (1996) and the companies sorted by these codes. A count is made of the number of firms within each industry entering this sample of 367 firms (the Datadisc sample) to calculate `R&D pervasiveness' scores. Industry codes are then ranked on this measure.
The second step is to select the sample. Industries (and all firms within them listed on the Australian Stock Exchange(8)) are selected in descending order (of R&D pervasiveness) until a sample size of 200 firms is reached. The 200 firms selected are from 28 industry codes. The `R&D pervasiveness' statistic for the last entering industry code is 59% (i.e. 59% of firms in this industry mention `research and development' in their Datadisc entry). Data is initially collected for 1995 financial year-ends.
Of the 200 firms identified, annual reports for 5 were unavailable, 88 did not disclose any R&D in their financial reports and 13 only disclosed the tax related incentive associated with their R&D expenditure. Of the remaining 94 firms, returns data was unavailable for 5 firms. Therefore, the sample for 1995 consists of 89 firms. The 1994 financial statements were also sought for these 89 firms. Useable data was obtained for 78 of these resulting in a final sample of 167 firm-years. Market data was obtained from the UNSW Banking and Finance database and, where particular data was unavailable, price, dilution factor and dividend information was extracted from the Knight Ridder -- Equinet database. All 167 firm-years are used in the descriptive statistics and in estimating the balance sheet models (1a and 1b). However, 23 observations had returns data for less than one year (start-up firms). Therefore, the regressions utilising returns data (models 2a through to 2f) are estimated on the remaining 144 firm-years.
2.4 Descriptive Statistics
Descriptive statistics for the 167 firm-years in the final sample are presented in table 1.
Table 1 Descriptive Statistics for Firms in the Final Sample(1) Capitalisers n = 118 (63) Variable No. of Mean Median Std. Firm- No. of Mean ($m) ($m) Dev. Years Firms ($m) RDSPD 6.7 1.5 15.9 114 61 7.4 RDCAP 2.8 0.3 8.6 75 47 n.a. RDDEX 3.9 0.7 13.1 83 46 7.4 WROFF 0.6 0 3.6 15 13 61.9 AMORT 1.0 0.0 5.5 60 36 66.2 RDOB 5.1 0.1 17.2 71 41 128.1 RD 6.2 0.7 20.1 83 50 n.a. Total 872 31.0 3,924 118 63 1,282.6 Assets RDSPD/ 0.08 0.04 0.10 118 63 0.03 Total Assets Expensers n = 49 (26) Variable No. of Median Std. Firm- No. of ($m) Dev. Years Firms RDSPD 2.1 10.4 49 26 RDCAP n.a. n.a. n.a. n.a. RDDEX 2.1 10.4 49 26 WROFF n.a. n.a. 1 1 AMORT n.a. n.a. 1 1 RDOB n.a. n.a. 1 1 RD n.a. n.a. n.a. n.a. Total 400.5 2,029.8 49 26 Assets RDSPD/ 0.0 0.1 49 26 Total Assets
Note: (1.) n = number of firm-years (number of firms).
RDSPD = total R&D expenditure in year.
RDDEX = R&D costs directly expensed.
AMORT = previously capitalised R & D amortised in year.
RD = book value of capitalised R&D (year-end).
RDCAP = R&D costs incurred during year and deferred.
WROFF = previously capitalised R&D costs written off in year.
RDOB = opening balance of capitalised R&D.
Total Assets as per reported book value at year-end.
One `capitaliser' from 1994 changed its R&D accounting policy during 1995, amortising and writing-off previously capitalised amounts and thus, enters the `expenser' sample in the second year.
While our discussion here focuses on the number of firm-years, the number of firms these observations represent are shown in parentheses in the table. The sample comprises 118 `capitalisers' which have an explicit policy of selective R&D capitalisation (as disclosed in the note on significant accounting policies) and 49 `expensers' which either state that they routinely expense all R&D as incurred, or express no policy but expense all R&D in the current year. Of the `capitalisers', four did not incur any R&D costs in the relevant period, although they amortised previously deferred costs. Among the 114 selective capitalisers who incurred R&D expenditure in the relevant period, 75 firms capitalised (31 fully, 44 partially), and 83 directly expensed (39 fully, 44 partially) these costs.(9) Seventy-one of the selective capitaliser sample had some R&D on their balance sheets at the beginning of the year (RDOB), 60 of these amortised (AMORT) and 15 wrote-off (WROFF) some portion of that capitalised amount.
The total amount of R&D spending (RDSPD) by sample firms over the two years is $1.157 billion. Of this, $364m is incurred by expensing firms and $793m by selective capitalisers. The mean (median) R&D spending by expensers ($7.4m ($2.1m)) is higher than the corresponding amounts for the selective capitalisers ($6.7m ($1.5m)). However, the expensers are larger, with mean (median) total assets of $1,282m ($400m) compared to $872m ($31m) for capitalisers. Thus, capitalisers spend proportionately more on R&D relative to their size as indicated by the ratio of RDSPD to total assets. The skewness of the sample towards small firms is apparent. For example, although the average amount capitalised (RDCAP) by selective capitaliser firms is $2.8m, the median value is only $0.3m. Similar differences are noted in the other variables of interest for the selective capitaliser sample with means (medians) of $3.9m ($0.7m) for R&D directly expensed (RDDEX), $1.0m ($0.0m) for the annual amortisation charge (AMORT) and $6.2m ($0.7m) for the amount of R&D on the balance sheet at the end of the period. This skewness underscores the deflation of variables in all regressions to mitigate against heteroscedasticity.
3. Results and Discussion
Discussion of the results is divided into two sections. Section 3.1 presents the results of tests of the value-relevance of capitalised R&D. Section 3.2 provides tests of the usefulness of R&D accruals in performance measurement. All reported results are based on a pooled time-series sample after the exclusion of a small number of influential observations defined as those with an absolute RStudent value greater than 3.0 (Easton, Eddey & Harris 1993). All t-statistics reported are White-adjusted. We also discuss the robustness of these results in year-by-year estimations.
3.1 Tests of the Value-Relevance of Capitalised R&D
Panel A of table 2 presents the results of the regressions investigating the value-relevance of capitalised R&D on the balance sheet (model 1 a). The coefficients on total assets excluding R&D (TALRD), total liabilities (TL), and capitalised R&D (RD) are statistically significant in the predicted directions and the adjusted [R.sup.2] is 49%. Although the coefficient on RD is higher than that on TALRD, a test of equality between the two coefficients fails to reject the null hypothesis of no difference (two-tailed p [is greater than] 0.10).(10) While the correlation between TALRD and TL is high (-0.95) as might be expected in such a model, correlations between these two variables and RD (0.19 and 0.26, respectively) are not sufficiently high to suggest problems with multicollinearity.
Table 2 Regressions of the Market Value of Equity on Capitalised R&D and Other Balance Sheet Components(1)
Panel A--Selective capitalisers (n = 114; after excluding 4 outliers with RStudent > |3.0|) (1a) MVE = [[Alpha].sub.0] + [[Alpha].sub.1] TALRD + [[Alpha].sub.2] TL + [[Alpha].sub.3] RD + [micro]' Variable Intercept TALRD TL RD Predicted Sign + - + Coefficient 0.368 1.401 -1.749 2.478 t-statistic (2.945) (4.722) (-3.940) (2.574) 2 tailed p-value 0.004 0.000 0.000 0.011 Adj. [R.sup.2] = 49% Panel B--Selective capitalisers (n = 114; after excluding 4 outliers with RStudent > |3.0|) (1b) MVE = [[Alpha].sub.0] + [[Alpha].sub.1] TTA + [[Alpha].sub.2] TL + [[Alpha].sub.3] RD + [[Alpha].sub.4] IALRD + [micro]' Variable Intercept TTA TL RD IALRD Predicted Sign + - + + Coefficient 0.391 1.387 -1.674 3.087 0.912 t-statistic (3.177) (4.834) (-4.000) (2.745) (2.151) 2 tailed p-value 0.002 0.000 0.000 0.007 0.034 Adj. [R.sup.2] = 49%
(1.) All variables are scaled by number of ordinary shares outstanding at 1995 year-end.
MVE = market value of equity at balance date.
TL = total liabilities.
TTA = total tangible assets.
TALRD = total assets less capitalised R&D.
RD = capitalised R&D on the balance sheet.
IALRD = intangible assets less capitalised R&D.
Panel B disaggregates TALRD into total tangible assets (TTA) and intangible assets excluding R&D (IALRD). This model (1b) also has an adjusted [R.sup.2] of 49%. The coefficients on all variables are statistically significant in the predicted directions.(11) Again, while the coefficient on RD is higher than that on TTA or IALRD, tests of equality between (a) the coefficients on RD and TTA; and (b) the coefficients on RD and IALRD, fail to reject the null hypothesis of no difference (two-tailed p [is greater than] 0.10 and p [is greater than] 0.20, respectively).(12) With the exception of a high correlation between TTA and TL (-0.91), the correlations between variables do not suggest difficulties caused by multicollinearity. The next highest correlation is between RD and IALRD (0.55).
Except where otherwise noted, similar findings are obtained in year-by-year estimations or when market values are measured three months after balance date. In summary, our results suggest that capitalised R&D on the balance sheet reflects information that is relevant to the pricing of company shares.(13)
3.2 Tests of the Usefulness of R&D Accruals in Performance Measurement This section reports the results of regressions which test the extent to which R&D accruals improve or detract from the usefulness of earnings as a performance measure for capitalising firms, and the extent to which a simulated capitalisation improves earnings as a performance measure for expensing firms. The results for capitalisers are presented in panel A of table 3 while those for the expensing group are shown in panel B.
Table 3 Tests of the Utility of R&D Accruals in Contributing to Earnings as a Measure of Performance(1)
Panel A--Selective Capitalisers (n = 98; after excluding 2 outliers with RStudent > |3.0|)(2) (2a) R = [[Alpha].sub.c] + [[Beta].sub.c] NPAT + [Epsilon]' (2b) R = [[Alpha].sub.c] + [[Beta].sub.c] (NPAT - RDCAP) + [Epsilon]' (2c) R = [[Alpha].sub.c] + [[Beta].sub.c] (NPAT - RDCAP + WROFF) + [Epsilon]' (2d) R = [[Alpha].sub.c] + [[Beta].sub.c] (NPAT - RDCAP + WROFF) + AMORT) + [Epsilon]' Model 2b Model 2c Model 2d Model 2a Intercept -0.071 -0.069 -0.072 -0.069 t-statistic (-1.560) (-1.435) (-1.486) (-1.424) Coefficient 0.421 0.093 0.084 0.124 t-statistic (3.405) (1.176) (1.115) (1.358) Adj. [R.sup.2] (%) 8.42 0.51 0.20 0.86 Panel B--Expensers (n = 44; no outliers with RStudent > |3.0|)(2) (2e) R = [[Alpha].sub.e] + [[Beta].sub.e] NPAT + [Epsilon]' (2f) R = [[Alpha].sub.e] + [[Beta].sub.e] (NPAT + RDDEX) + [Epsilon]' Model 2e Model 2f Intercept 0.051 0.073 t-statistic (1.116) (1.407) Coefficient -0.721 -0.788 t-statistic (-3.783) (-2.345) Adj. [R.sup.2] (%) 8.98 6.58 Panel C--Results of the likelihood ratio test developed by Vuong (1989) for non-nested model selection; a significant and negative (positive) Z-statistic indicates that the performance measure in the second (first) model is rejected in favour of the performance measure in the first (second). Comparison of Vuong's Probability Performance Measure in Z-statistic 1st Model v. 2nd Model 2a v. 2b -2.97 (0.002) 2a v. 2c -3.27 (0.001) 2a v. 2d -2.77 (0.003) 2b v. 2c -0.08 (0.465) 2c v. 2d 1.71 (0.043) 2e v. 2f -1.21 (0.113)
(1.) All variables are expressed in per-share terms and scaled by beginning-of-year share price.
R = share market annual returns (adjusted for changes in the basis of capitalisation).
NPAT = net profit after tax and before extraordinary items (as reported in the P&L statement).
RDCAP = R&D costs incurred during the period and deferred.
RDDEX = R&D costs directly expensed.
AMORT = previously capitalised R&D costs amortised in current period.
WROFF = previously capitalised R&D costs written off in current period.
(2.) Observations with incomplete returns data (returns for less than one year) are deleted from this analysis. This leaves 100 (out of a maximum of 114) and 44 (out of a maximum of 49) observations in the capitalising and expensing groups, respectively.
Models 2a-2d in panel A consider the extent to which removal of R&D accruals from earnings affects its utility as a performance measure for capitalising firms. The results suggest that the RDCAP accrual assists in improving the usefulness of accounting-based performance measures. Removing RDCAP from earnings reduces the association of earnings with returns from 8.42% to 0.51%, a difference that is statistically significant using the Vuong statistic (reported in panel C). The stepwise removal of WROFF and AMORT likewise produces performance measures that are statistically inferior to NPAT in terms of explanatory power. The adjusted [R.sup.2] for all models which exclude one or more accruals is less than 1%; further, the relevant Vuong statistics comparing models 2c, as well as 2d, to model 2a indicate that NPAT is more closely associated with returns than those measures which exclude R&D accruals. Note that the largest shift in explanatory power occurs when RDCAP is removed. The other accruals do not have much impact on the utility of the performance measure. However, no significant decrease in [R.sup.2]s is observed in year-by-year estimations or in estimations based on annual returns measured to financial year-end plus three months.
The results for the expenser sample reported in panel B are puzzling. The coefficient on earnings for expensers is actually negative implying that earnings is inversely related to the market measure of performance for the sample of expensers.(14) Reasons for this are not clear. When the simulated R&D capitalisation (RDDEX) is added back to earnings for the expensers (model 2f), the adjusted [R.sup.2] on the earnings-return regression falls from 8.98% to 6.58%.
The results of this study suggest that R&D capitalisations by management are value-relevant and that the R&D capitalisation accrual improves accounting-based measures of firm performance for Australian firms in industries where R&D activity is widespread. Specifically, in tests of value-relevance, capitalised R&D of `selective capitalisers' has a significant association with firm value. This result is robust to year-by-year estimations and the measurement of market value at three months after balance date. In tests of the usefulness of R&D accruals in improving measures of performance we find that they do add to the utility of earnings, improving its association with contemporaneous market returns. The evidence for this is stronger in the sample of capitalising firms and within that group for the R&D capitalisation accrual in particular. However, the results on performance measures are not stable in year-by-year estimations or in estimations using annual returns measured to three months after financial year-end.
Our findings are relevant to the regulatory debate to disallow the capitalisation of R&D costs and require immediate expensing in line with US requirements. We show that the outcome of currently permitted managerial discretion to capitalise R&D costs provides superior measures of firm value and, possibly, firm performance. This suggests caution in reviewing current Australian reporting requirements. It also suggests benefits to further research based on larger samples comprising both a larger number of firms and over a longer period of time.
(1.) The closer a firm is to its debt constraints, the greater are the incentives to capitalise R&D, thereby increasing both the asset base and reported profits. However, this is unlikely to be descriptive in Australia where debt covenants are often written in terms of tangible asset ratios which effectively exclude the effects of R&D policy choice (Whittred & Zimmer 1986; Ramsay & Sidhu 1998).
(2.) There is considerable evidence that large firms tend to expense R&D while small firms are more likely to capitalise it (Percy 1994; Selto & Clouse 1985; Daley & Vigeland 1983; Horwitz & Kolodny 1980, 1981).
(3.) Subscripts have been omitted from the dependent and independent variables for all models for simplicity. An alternative specification would include `earnings' as an independent variable (Amir & Lev 1996). A third alternative is to adopt an `income statement approach' where MVE is modelled as a function of earnings components as in Barth, Beaver and Landsman (1992). We discuss the outcome of these alternative specifications in the results section.
(4.) Our results are similar if market value is measured three months after the balance date.
(5.) A number of US studies investigate the utility of R&D accruals by examining the relationship between abnormal returns and unexpected earnings, decomposed into R&D expenditures and other specific revenue and expense items (Lee & Sami 1995; Bublitz & Ettredge 1989). Potential limitations in estimates of abnormal returns are a possible source of error in these models (Bublitz & Ettredge 1989, p. 123). Further, note that this approach is designed to answer a different research question--it asks whether the release of the R&D numbers has any `news' (information content) beyond market expectations, and thus, any impact on share price.
(6.) Our results are not sensitive to the use of annual returns measured to financial year-end plus three months.
(7.) Datadisc is a database managed by the ASX and includes text and statistics, where available, since 1987.
(8.) Note that at this stage, companies are not restricted to the Datadisc sample of 367 firms used to estimate the proxy of `R&D pervasiveness'.
(9.) Selective capitalisers, in aggregate, capitalised 32% of their R&D expenditure (RDCAP/RDSPD), while directly expensing the remaining 68%.
(10.) However, the null hypothesis of no difference is rejected (p = 0.077) in sensitivity tests based on market values measured at three months after balance date.
(11.) The coefficient on IALRD is not significantly different from zero (p = 0.151) in tests based on 1995 data.
(12.) However, the coefficients on RD and IALRD are statistically different (p = 0.009) in estimations based on 1995 data. Further, the null hypotheses of no difference between the coefficients on RD and TTA, and between RD and IALRD, are rejected (p = 0.022 and 0.018, respectively) in sensitivity tests based on market values measured at three months after balance date.
(13.) We repeated our analysis using an alternative specification similar to that used by Amir and Lev (1996), by including Earnings (NPAT, i.e. net profit after tax and before extraordinary items) as an independent variable in models 1a and 1b. The coefficient on RD remained significant and positive in both models. Following Barth, Beaver and Landsman (1992) we also performed estimations based on an income statement approach with MVE modelled as a function of earnings components--sales, other expenses, R&D directly expensed, R&D amortisations and write-offs (the latter two applying to the capitalising group only). For the capitalising group, only sales and other expenses were statistically significant in the predicted directions (positive and negative, respectively). For the expensing group, sales and R&D directly expensed were both positive and significant, implying that the market does not perceive the R&D expenditure of this group as `expired' but attributes some future value to it.
(14.) Note that this curious result occurs for both years of data in year-by-year estimations and in estimations based on annual returns measured to financial year-end plus three months. Barth and Clinch (1998) report a similar effect among small firms in their sample, which is also drawn from Australian firms.
(Date of receipt of final typescript: September 1998 Accepted by Greg Clinch, Area Editor.)
Amir, E. & Lev, B. 1996, `Value-relevance of nonfinancial information: The wireless communication industry', Journal of Accounting and Economics, vol. 22, pp. 3-30.
Australian Accounting Standards Board (AASB) 1987, AASB 1011: Accounting for Research and Development Costs, AASB, Melbourne.
Australian Stock Exchange (ASX) 1996, Industry Classification Report, ASX, Sydney.
Barth, M.E. 1994, `Fair value accounting: Evidence from investment securities and the market valuation of banks', The Accounting Review, vol. 69, no. 1, pp. 1-25.
Barth, M.E., Beaver, W.H. & Landsman, W.R. 1992, `The market valuation implications of net periodic pension cost components', Journal of Accounting and Economics, vol. 15, pp. 27-62.
Barth, M.E. & Clinch, G. 1998, Revalued financial, tangible, and intangible assets: Associations with share prices and non market-based valuation, Journal of Accounting Research, supp. (forthcoming).
Bublitz, B. & Ettredge, M. 1989, `The information in discretionary outlays: Advertising, research and development', The Accounting Review, vol. 64, no. 1, pp. 108-24.
Clinch, G. 1991, `Employee compensation and firms' research and development activity', Journal of Accounting Research, vol. 29, no. 1, pp. 59-78.
Coombes, R., Otto, C. & Stokes, D. 1996, Economic determinants of the amortisation of identifiable intangible assets, Working paper, University of Southern Queensland.
Daley, L.A. & Vigeland, R.L. 1983, `The effects of debt covenants and political costs on the choice of accounting methods: The case of accounting for R&D costs', Journal of Accounting and Economics, vol. 5, pp. 195-211.
Dechow, P.M. 1994, `Accounting earnings and cash flows as measures of firm performance: The role of accounting accruals', Journal of Accounting and Economics, vol. 18, pp. 3-42.
Dechow, P. & Sloan, R. 1991, `Executive incentives and the horizon problem', Journal of Accounting and Economics, vol. 14, pp. 51-89.
DeFond, M. & Jimbalvo, J. 1994, `Debt covenant violations and manipulation of accruals', Journal of Accounting and Economics, vol. 17, pp. 145-76.
Easton, P., Eddey, P. & Harris, T. 1993, `An investigation of revaluations of tangible long-lived assets', Journal of Accounting Research, vol. 31, pp. 1-45.
Financial Accounting Standards Board (FASB) 1974, Statement of Financial Accounting Standards no. 2: Accounting For Research and Development Costs, FASB, Stamford, Connecticut.
Guay, W. & Sidhu, B.K. 1998, The role of long-term accruals, Working paper, University of Philadelphia and the University of New South Wales.
Healy, P. 1985, `The effects of bonus schemes on accounting decisions', Journal of Accounting and Economics, vol. 7, pp. 85-107.
Healy, P.M. & Palepu, K.G. 1993, `The effect of firm's financial disclosure strategies on stock prices', Accounting Horizons, vol. 7, no. 1, pp. 1-11.
Healy, P.M. & Palepu, K.G. 1995, `The challenges of investor communications: The case of CUC International Inc.', Journal of Financial Economics, vol. 38, pp. 111-40.
Henderson, S. & Pierson, G. 1995, Issues in Financial Accounting, 7th edn, Longman, Melbourne.
Holthausen, R. & Leftwich, R. 1983, `The economic consequences of accounting choice: implications of costly contracting and monitoring', Journal of Accounting and Economics, vol. 5, pp. 75-117.
Horwitz, B. & Kolodny, R. 1980, `The economic effects of involuntary uniformity in financial reporting of R&D expenditures', Journal of Accounting Research, vol. 18, supp., pp. 38- 107.
Horwitz, B. & Kolodny, R. 1981, `The FASB, the SEC and R&D', The Bell Journal of Economics, vol. 12, no. 1, pp. 249-62.
Lee, B. & Sami, H. 1995, The effect of R&D and advertising outlays on earnings response coefficient, Working paper, Concordia University.
Lev, B. & Sougiannis, T. 1996, `The capitalisation, amortisation and value-relevance of R&D', Journal of Accounting and Economics, vol. 21, pp. 107-38.
McCarthy, M.G. & Schneider, D.K. 1995, `Market perception of goodwill: Some empirical evidence', Accounting and Business Research, vol. 26, no. 1, pp. 69-81.
Percy, M. 1994, Accounting and disclosure of research and development expenditure in Australia, Working paper, University of Queensland.
Ramsay, I.M. & Sidhu, B.K. 1998, `Accounting and non-accounting based information in the market for debt: Evidence from Australian private debt contracts', Accounting and Finance, vol. 38, no. 2, pp. 197-221.
Schmidt, L. (ed.) 1996, `R&D costs warning' in Accountancy Hotline, Business Review Weekly, 24 June, p. 106.
Selto, F. & Clouse, M. 1985, `An investigation of managers' adaptations to SFAS No. 2: Accounting for research and development costs', Journal of Accounting Research, vol. 23, no. 2, pp. 700-17.
Smith, C.W. & Watts, R.L. 1982, `Incentive and tax effects of executive compensation plans' Australian Journal of Management, vol. 7, pp. 139-58.
Stickels, G. 1996, `Authorities aim for the objective on R&D', Business Review Weekly, 18 March, pp. 99-100.
Vuong, Q.H. 1989, Likelihood ratio tests for model selection and non-nested hypotheses, Econometrica, vol. 57, pp. 307-33.
Watts, R.L. & Zimmerman, J.L. 1986, Positive Accounting Theory, Prentice Hall, New York.
Watts, R.L. & Zimmerman, J.L. 1990, `Positive accounting theory: A ten year perspective', The Accounting Review, vol. 65, pp. 131-56.
Whittred, G. & Zimmer, I. 1986, `Accounting information in the market for debt', Accounting and Finance, vol. 26, no. 2, pp. 19-33.
Tony Abrahams and Baljit K. Sidhu, Australian Graduate School of Management, The University of New South Wales, Sydney NSW 2052; E-mail: email@example.com
We especially thank G. Clinch (the Area Editor), G. Whittred and an anonymous reviewer for many helpful suggestions. We also received useful comments from: M. Daley, D. Dhaliwal, R. Swieringa, seminar participants at UNSW, the 1997 conferences of the AAANZ (Hobart), the IAAER (Paris), and the Asian-Pacific International Accounting Conference (Bangkok). Finally, thanks to R. Czernkowski and Ian Huntley Pry Ltd for help with data collection. B.K. Sidhu acknowledges financial support from the Australian Research Council.3
|Printer friendly Cite/link Email Feedback|
|Author:||Sidhu, Baljit K.|
|Publication:||Australian Journal of Management|
|Date:||Dec 1, 1998|
|Previous Article:||Ownership Structure and Building Society Efficiency.|
|Next Article:||An Analysis of Intraday Quoted Bid-Ask Spreads in Futures Markets: Evidence from the Sydney Futures Exchange.|