R&D Capitalization and The Income Smoothing Hypothesis A study of Swedish listed Companies

Transcription

1 Master thesis in Accounting, Auditing and Analysis (2011) 1-29 Supervisor: Jiri Novak R&D Capitalization and The Income Smoothing Hypothesis A study of Swedish listed Companies Annelie Persson and Karen Fuentes Abstract This paper examines whether Swedish listed firms use research and development (R&D) accounting as a tool for income smoothing (hypothesis 1). One controversial accounting issue concerning R&D is that R&D capitalization could be influenced by earnings management purposes due to a subjective accounting treatment. We also examine whether firms degree of fluctuation in return on assets (ROA) has an effect on income smoothing behavior (hypothesis 2). Finally, we investigate if the level of flexibility allowed in the R&D accounting with the different accounting standards, BFN R1, RR 15 and IAS 38 has an effect on income smoothing behavior (hypothesis 3). We study the accounts for 21 firms for the years , 52 firms for and 59 firms for Using multiple regression analysis we find that the income smoothing hypothesis is supported in period two ( ). The regression analysis also indicates that firms with low change in ROA tend to capitalize more R&D when they are less profitable than prior year. Our results also imply that the level of flexibility in different accounting standards does not have an effect on income smoothing behavior and hypothesis 3 is not supported. Keywords: Earnings management; income smoothing; R&D accounting; R&D capitalization; ΔROA I. Introduction We have progressively moved from an industrial society to a knowledge society, in which investment in R&D is crucial in order to maintain the firm s competitive position (Cañibano et al., 2000; Zhao, 2002). Firms undertakes more R&D activities and the reporting of R&D has increased significantly during the recent years and it is a well discussed topic in the accounting research literature, in the context of value relevance (Lev & Sougiannis, 1996; Aboody & Lev, 1998; Lev & Zarowin, 1999; Healy et al., 2002; Zhao, 2002; Oswald & Zarowin, 2007) and earnings management (Oswald & Zarowin, 2007; Markarian et al., 2008; Seybert, 2010). Since firms undertakes more R&D activities than before, the annual reports consist of an increased R&D reporting which has result in various problems for standard setters and users of the financial statements, and one controversial issue concerning R&D reporting is the existing accounting differences around the world (Zhao, 2002; Hoegh-Krohn & Knivsflå, 2000). While the U.S. Generally Accepted Accounting Standards (U.S. GAAP) require full and immediate expense on R&D expenditures (except for the software industry), International Financial Reporting Standards (IFRS), requires capitalization of R&D outlays when specific criterias are met. In addition, the accounting treatment for R&D seems to be an open issue and there is a disagreement among standards setters of whether capitalization 1

2 or expensing of development expenditures should be considered as the most appropriate R&D reporting treatment. There are different views of the relative pros and cons with the respective accounting treatments. Before 1975, U.S GAAP allowed companies to capitalize R&D outlays. However in 1975 the Financial Accounting Standards Board (FASB) required US companies to fully expense R&D and U.S. standard setters argue that expensing is preferable to capitalization because it enhances the objectivity of the financial statements and eliminates management s ability to capitalize project cost that has low probability of success. Furthermore, proponents of this standard argue that the possibility for delaying essential impairment of R&D assets eliminates with the expensing method. In contrast, full expensing has also been criticized and prior studies provide evidence for value relevance loss of the financial reporting with the full expensing method (Healey et al., 2002; Lev & Zarowin, 1999; Lev and Sougiannis 1996; Aboody and Lev 1998; Zhao 2002). Gornic- Tomaszewski and Millan (2005) argues that full expensing does not consider the matching of the present period s expenditures, and the rejection of capitalization of R&D in the balance sheet are the same as excluding the firms most valuable asset. The conclusion drawn by the critics, as mentioned in Oswald and Zarowin (2007) is that the expensing method significantly distorts the relevance of the balance sheet and it has been shown that the immediate R&D expensing under US GAAP combined with an increased R&D intensity is some of the main factors to the irrelevance of the financial information (Lev & Zarowin, 1999). R&D capitalization seems to be value relevant since it reduce information asymmetry between the firm and the market participants (Lev and Sougiannis, 1996) and as mentioned by Healey et al. (2002) managers can use write-downs for unsuccessful projects to reveal useful information to users of the financial report concerning the R&D outcomes. Nevertheless, R&D capitalizing can also create an opportunity for managers to engage into earnings management and then the capitalization method becomes an accounting issue (Cazavan-Jeny & Jeanjean, 2003; Markarian et al., 2008). As noted in Callimacy and Landry (2003 p. 134) the application of the criteria for capitalization is based on management judgement and induces a considerable amount of subjectivity, giving manager opportunity to utilize R&D capitalization for self benefit. Most prior research has concentrated of the value relevance of R&D accounting. However, as discussed by Seybert (2010) value relevance benefits (R&D capitalization) must be weighed against the effect the accounting methods have on management decision. As provided by Markarian et al. (2008) in their study, the decision to capitalize R&D costs is related to the change in profitability and the results suggest that firm, who have decreased profitability tend to capitalize R&D expenditures more often than the ones who have an increased profitability and we find it is interesting to see whether this pattern can also be found for Swedish listed firms. Before harmonizing to IFRS in 2001 and before converting to IFRS in 2005, Swedish accounting standards allowed firms to choose between the two possible R&D accounting treatments; R&D capitalization or expensing of development expenditures. Today Swedish listed firms prepare their R&D accounting in accordance with the recommendation by IAS which is based on the accrual principle. IAS 38 Intangible Asset requires development expenditures to be capitalized as an intangible asset once certain criteria are met, while research cost must be expensed when incurred. This paper examines whether managers choice to capitalize R&D expenses is influenced by earnings management purposes and whether the result supports the income smoothing hypothesis, as in the paper by Markarian et al, Under Swedish accounting rules, R&D activities must be disclosed which makes it interesting to conduct a R&D study based on a sample of Swedish listed companies. 2

3 R&D cash outlays will either be classified as research and will be expensed the year they occur affecting the income statement, or as development and will be activated as an intangible assets in the balance sheet and are expensed over the intangible assets expected useful life. The classification problem of what expenses that should pertain to the research phase and which should be considered as development has been harmonized by Organization for Economic Co-operation and Development (OECD) Frascati definition, which is the internationally recognized definition of R&D (Khadoroo et al., 2003). International Accounting Standard Board (IASB) defines research as the original and planned investigation undertaken with the prospect of gaining new scientific or technical knowledge and understanding while development is defined as the application of research findings or other knowledge into a plan or design for the production of new or substantially improved materials, devices, products, processes, systems or services prior to the commencement of commercial production or use. Although, IASB provides specific guidelines of what expenses that should pertain to research and which that should be classified as development, we believe firms are taking advantage of the classification problem and we hypothesize that firms choice of classification is influenced by earnings management purpose. Given the subjectivity in assessing whether the R&D capitalization criteria are met, it seems that, even under IAS, companies that prefer to expense even when the criteria for capitalization are met, can justify this approach (Callimacy & Landry, 2003; Markarian et al., 2008). Hence, the purpose of this study is to examine whether R&D accounting gives opportunities for income smoothing behavior. R&D accounting is a vivid debate among regulators and the central issue in the debate is whether managers should have the flexibility to capitalize certain R&D costs according to IAS 38 or not (Cazavan-Jeny et al., 2007). A crucial question for standard setters is to determine how much judgement to allow management to exercise in the financial reporting (Healy & Wahlen, 1999). Evidence of R&D accounting and earnings management is therefore primarily of interest for standards setters. A small number of studies have provided evidence for real earnings management; meaning that managers over-invest or under-invest in R&D in order to achieve their earnings goal (e.g. Oswald & Zarowin, 2007; Seybert, 2010), but research who examines the motives behind the choice of R&D classification are even less. Thus, our study contributes to the debate of appropriate accounting treatment of R&D cost by providing empirical evidence for managerial incentives in the choice of R&D classification. We hypothesis that managers decision to capitalize R&D expenditures are affected by earnings management incentives and the investigated method of earnings management involved in this study is smoothing of the earnings stream through accounting procedure choices i.e. classifying development as an intangible asset or as an expense when incurred. Our hypotheses are tested on firms listed on the Stockholm Stock Exchange, NASDAQ OMX. Using multivariate regressions our results indicates that Swedish firms utilize the subjectivity of R&D capitalization to smooth reported earnings in one of two periods tested. The regression model suggest that firms with lower fluctuations in ROA tend to capitalize more R&D when they are less profitable than the prior year and we note that the level of flexibility the different R&D accounting standards allows managers to exercise has no direct impact on the magnitude of income smoothing that takes place. The reminder of the paper proceeds as follow. In section II we introduce R&D accounting in Sweden and the institutional background. Section III reviews relevant literature and develops the hypothesis, following by the research methods in section IV. Section V presents the result, while section VI concludes the study. 3

4 II. Institutional background of R&D accounting in Sweden The applied accounting treatment for R&D expenditures for Swedish firms have historically varied and the development of R&D accounting can be divided into three periods dependent on which R&D accounting standard firms are subjected to; Bokföringsnämndens rekommendationer 1 (BFN R1), Redovisningsrådets rekommendationer 15 (RR 15) or IAS 38. The R&D accounting treatment was first influenced by The Swedish Accounting Standards Board, then by The Accounting Council and later by The International Accounting Standard Council. In each period, the R&D reporting rules has taken a turn in some sense. However, Swedish accounting regulation has always allowed for some flexibility in the treatment of R&D expenditures and R&D capitalization has at all times been present. Thus R&D capitalization has always been a topic for the accounting regulation. The Swedish Accounting Standard Board The first accounting standard for R&D in Sweden were initiated the 1th of January 1988, where Swedish firms with financial year initiated after this date had to conform to the recommendation BFN R1 by the Swedish Accounting Standards Board. The recommendation aimed to treat the accounting information given of the R&D expenditures and its magnitude (BFN R1). Before BFN R1 there had not been any developed accounting treatment for R&D expenditures, instead the legislature had referred to the accounting practice, which led to a decreased comparability between different firms. The aim of the implementation of BFN R1 was to provide an attempt for international harmonization considering the accounting treatment of R&D expenditures. Indicative international recommendations used are the one developed by OECD and International Accounting Standards Council (IASC). Both recommendations agreed upon that the main treatment of R&D expenditures is to expense them immediately when they arise but that capitalization is allowed when certain criteria are fulfilled. The major argument for this accounting treatment was that the certainty regarding the future benefits from R&D in general were to considered as are very small (BFN R1). The capitalization treatment in BFN R1 is recorded in Bokföringslagen (BFL 17:2) and statues that R&D expenditures under certain circumstances can be capitalized as an intangible asset on the balance sheet. The decision should be based on whether the expenditures can be seen as of significant value for the firm during the following years. The guiding definitions of which R&D expenditures that shall be treated as R&D costs are based on the definitions by Statistiska Centralbyrån (SCB), which also are in accordance with the guidelines of OECD and IAS 9, Accounting for research and development activities. BFN R1 divide R&D expenditures into three areas; basic research, applied research and development, as opposed to the later standards, who distinguishes merely between two categories of R&D expenditures. Basic research is defined as organized and systematical search after new knowledge and ideas without any established appliance in mind. Applied research is, opposite from the former, to systematically and organized search for new knowledge and new ideas with a certain established appliance in mind and Development is to systematical and organized use the result from the research and the achieved knowledge in order to create new products, new processes, new systems or essential improvement for existing ones (BFN R1). Concerning the estimated useful life in BFN R1, a minimum criterion is that one fifth of the capitalized value shall be depreciated (17:2 BFL). If the expected future benefits are likely to arise in a shorter period, the depreciation period should be adjusted to these conditions. In extraordinary cases there can be deviations from the minimum criteria and a longer depreciation period shall then be used, this concerns mostly the circumstances when 4

5 the R&D is provided by external financing (Bokföringsnämndens praxis, 1999). This longer period shall in those cases be possible to determine with reasonable certainty and a disclosure of the reason to the deviation shall be provided (Hoegh-Krohn & Knivsflå, 2000). As RR 15 and IAS 38, intangible assets shall be initially recognized at acquisition value and amortized over the intangible assets useful life on a straight line basis according to the main rule. However, in cases a different method is available that better reflect the reality it can be selected under the conditions that it is applied consequently during the intangible assets entire expected useful life. The depreciation period shall be initiated at latest at the date when the R&D project has been completed (Bokföringsnämndens Praxis, 1999). The assessment whether capitalization of R&D assets are justified shall be performed at each fiscal year. If the required conditions for the capitalization no longer is fulfilled a write down shall be performed (Bokföringsnämndens praxis, 1999). This principle is also in accordance with RR 15 and IAS 38. The Accounting Council After the 1th of January 2001 the accounting treatment for R&D by BFN R1 concerning listed firms and their corporate group reporting were replaced by the recommendation of the Accounting Council (RR 15a). As a result of later amendments the required implementation date were moved forward to the financial year initiated after the 1th of January 2002 (RR 15b). However, earlier implementation was possible and also encouraged. The main difference between RR 15 and the previous accounting standard, BFN R1 is foremost that according to BFN R1, firms are allowed to treat certain R&D expenditures as capitalized asset if they fulfill certain criteria. Note that there is no obligation to activate in R&D outlays in BFN R1, it become a requirement under RR 15. The main rule is that the expenditure for R&D shall be expensed when they arise unless they do not meet the criteria of an intangible asset or if they have arise as a part of a corporate acquisition and cannot be reported as a separate asset in the acquisition balance sheet. In those cases they should pertain to the value of the acquired goodwill. The assets pertaining to the research phase shall be expensed; the once belonging to the development phase shall be capitalized and if there is no way to distinguish the phases from each other the expenditures shall be expensed, as the main rule states (RR 15a p ). As BFN R1 and IAS 38, for R&D expenditures to fulfill the criteria for capitalization it must fulfill the required conditions for an intangible asset. An intangible asset is defined as a non-monetary asset without any physical substance that is used in the production of products, for rental to others or for administrative aims. An asset is further defined as a resource for the firm that the company has control of as a consequence of events and that is probable to give the firm economical advantage in the future. Hence, for fulfilling the criteria s of an intangible asset the R&D expenditures shall be identifiable, within the firms control and give the firm future economical benefits. (RR 15a) Additional difference between the accounting standards pertains to the expenditures which are prohibited to capitalize and shall be expensed when incurred. RR 15 p contains a limitation which BFN R1 lacks, and can therefore be seen as more restrictive in that sense. Further difference is the assumption of useful life of the intangible asset (RR 15c, 2000). While the estimated useful life according to BFN R1 is restricted to five years, the applied assumption taken in RR 15 is that the estimated useful life of intangible properties can be extended to maximum 20 years. As a result of experienced difficulties on how to determine the acquired value in a trustworthy way, and in cases where it has been impossible to distinguish the costs pertaining to the internal produced immaterial asset from maintaining, RR 15 divide R&D 5

6 expenditures into two phases; one research phase and one development phase (RR 15c, 2000). Research is defined as a systematical and planned search that could provide new scientific knowledge or new technical insight and knowledge. Development is to conform to the result of the research or other obtained knowledge, to achieve new or significant improved constructions, material, processes, systems, products or services within commercial production or usage (RR 15c, 2000), which are broadly the same definitions as in IAS 38. The International Accounting Standard Council In an effort to harmonize the financial accounting worldwide, listed firms in Sweden have since 2005 been obligated to implement the accounting principles following IFRS, and the recommendation by IASC for R&D accounting (IFRS, 2010). The accounting treatment of intangible assets and consequently the accounting for R&D is presented in the accounting standard IAS 38, Intangible Assets. The objective of IAS 38 is to prescribe the accounting treatment for intangible assets and mandates full expensing of all research expenditures (IAS 38:68), like previous standards. Capitalization of development outlays is only permitted if the asset fulfills the technical feasibility criteria and the firm can demonstrate on the ability for using or selling the asset. (IAS 38:57) Hence, firms have certain requirements to fulfill and must intend and be able to complete the intangible asset and either use it or sell it. Additional requirement is that the managers must be able to demonstrate how the asset will generate future economic benefits, otherwise development cannot be capitalized (IAS 38). More precisely, in order for development to be recognized as an intangible asset it must be (1) clearly identified (i.e. separable from goodwill) and (2) be controlled by the firm (i.e. control over the future economic benefits) and (3) give future economic benefits, which are broadly the same criteria s as in the previous standards. The general requirements provided by IAS 38 of the R&D disclosure states that an internal generated intangible asset must be distinguished from those acquired separately and those acquired through business combination. IAS 38 presents examples of separately classes of intangible assets. However, the standard allows for disaggregation or aggregation of the examples given if this provides the users of the financial statement with more relevant information (IAS 38:119). IAS 38 is quite similar to RR 15 since the latter was an attempt to harmonize to IASC recommendation. However, the standards differs in some respects and the main difference is that IAS 38 permits revaluation of intangible assets to fair value if there is an active market, whereas revaluation of intangible assets is not permitted in RR 15, nor BFN 1. Unlike the recommendations provided in RR 15, intangible assets can in IAS 38 be treated as having an indefinite useful live, while the presumption in RR 15 is that the useful life cannot exceed 20 years (KPMG, 2005). 6

7 Comparison of definitions of R&D activities and the information contained in the different accounting standards are summarized in exhibit 1 and 2. 1 Exhibit 1 Definition Research Development BFN R1 Basic research: organized and systematically searches after new knowledge and ideas without any established appliance in mind Applied research: systematically and organized search for new knowledge and new ideas with a certain established appliance in mind RR 15 IAS 38 Research: systematical and planned searches that could provide new scientific knowledge or new technical insight and knowledge Research; Original and planned investigation undertaken with the prospect of gaining new scientific or technical knowledge and understanding Systematical and organized use of the result from the research and the achieved knowledge in order to create new products, new processes, new systems or essential improvement for existing ones Conform to the result of the research or other obtained knowledge to achieve new or significant improved constructions, material, processes, systems, products or services within commercial production or usage Application of research findings or other knowledge to a plan or design for the production of new or substantially improved materials, devices, products, processes, systems or services prior to the commencement of commercial production or use Exhibit 2 Recommendations Disclosure Amortization Valuation treatment BFN R1 Capitalization is allowed when certain criteria s are fulfilled. Maximum 5 years Initially recognized at acquisition value RR 15 IAS 38 Capitalization must be made when certain criteria s are fulfilled. Capitalization must be made when certain criteria s are fulfilled. Maximum 20 years Indefinite useful live is allowed Initially recognized at acquisition value Initially at acquisition value and revaluation to fair value. Measurement after recognition either in accordance with a cost or revaluation model 1 We summarize the general accounting principles for BFN R1, RR 15 and IAS 38. Exceptions from these accounting principles are allowed in extreme cases. 7

8 III. Earnings management and hypothesis development Definition of earnings management Although earnings management has been a vivid debated topic in the academic field the existing literature only documents a few definitions of the topic. A common known definition of earnings management is given by Schipper (1989, p. 92) who defines it as a purposeful intervention in the external financial reporting process, with the intent of obtaining some private gain. Another definition is provided by Healy and Whalen (1999 p. 368) who gives a definition from the perspective of standards-setters and suggests that earnings management occurs when managers use judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company or to influence contractual outcomes that depend on reported accounting numbers. Regardless of the definition used, the one by Schipper (1989) or the one by Healey and Wahlen (1999), both concern the impact of management judgment on the financial reporting. The implication following the definition of earnings management is that; when managerial judgment is exercised on the financial reporting, there is a possibility for earnings management. Given the following definition, R&D accounting with the capitalization method is subject to possible exertion of earnings management (Callimacy and Landry, 2003; Koch, 1981; Markarian et al., 2008; Seybert, 2010). Why earnings management Mangers might be motivated to distort the figures in the financial reporting for several reasons and the research carried out indicates that managers manipulate earnings using a wide variety of methods. One of the methods for management to exercise earnings management is income smoothing, meaning in short; reducing the fluctuations in the reported income (Markarian et al., 2008). The income smoothing hypothesis suggest that income smoothing is a means used by management to diminish the variability of a stream of reported income numbers relative to some perceived target stream by the manipulation of artificial (accounting) or real (transactional) variables (Koch, 1981 p. 574). Hence, managers can manage earnings by using real operating decisions and/or financial reporting choices (Leuz et al., 2003). This paper focus merely on what Koch (1981) defines as manipulation of artificial variables and more precisely accruals and R&D capitalization. Several explanations for income smoothing are given by prior studies. The motives for managers to smooth income varies and the literature suggests that the underlying incentives for decreasing the fluctuations in reported earnings is to satisfy the external and internal users of the financial reports; external users such as investors and creditors and internal users such as management (Brayshaw & Eldin, 1989; Beidleman, 1973). Previous research suggest that managers may engage in income smoothing in order to influence stakeholders perception of the stability of the firm s earnings, hence the stakeholders assessment of a firm s probability of bankruptcy will change, which consequently decreases the cost of borrowing and increases the selling price of shares (Beidleman, 1973; Foster, 1986; Ronen and Saden, 1981; Trueman and Titman, 1988). It seems that the fluctuations in reported earnings are an essential measure for risk and is, as a consequence a common explanation in the R&D research literature for managers to engage in income smoothing. A further explanation for manage earnings is the firm s desire to pay out a smooth stream of dividends to the owners (Kasanen et al., 1996) and it is suggested that stable earnings allows for a higher level of dividends (Beidleman, 1973). While other research shows that motives for income smoothing have been to gain tax advantages (Hepworth, 1953) and to meet the benchmark target of prior year s earnings (Bauwhede et al., 2003). 8

9 Another driver of income smoothing behavior according to prior research is linked to management benefit and several explanations are provided. One reason is that managers may wish to reduce the variability in earnings over time to establish plans and budget for future periods more correctly since high variable earnings might make this difficult (Beidleman, 1973). Additional motivation is that firm s compensation scheme is related to the reported income affecting the compensation (Watts and Zimmerman, 1978; Brayshaw & Eldin, 1989). Furthermore, the threat of management displacement is another motive for managers to engage in income smoothing, since variation in the firms income might result in displacement of managers (Brayshaw & Eldin, 1989). Prior research imply that income smoothing help managers to provide private information to users of the financial statement (Beidleman, 1973), but also that managers incentives to smooth reported earnings is driven by opportunistic aims (Bauwhede et al., 2003; Foster, 1986; Trueman & Titman, 1988). Beidleman (1973) concludes that managers employ income smoothing to normalize reported earnings, which otherwise would deviate from their normal time trend and the results cannot prove that the incentives to smooth incomes are driven by opportunistic aim, while Trueman and Titman (1988) finds out that managers exercise income smoothing to lower stakeholders perception of the variance of the underlying economic earnings. The difference in their findings is due to managers underlying reason to smooth the reported earnings and the following implication is that income smoothing can be viewed both as a positive and negative strategy, as discussed by Markarian et al. (2008). Income smoothing and R&D capitalization By managing earnings upwards in bad periods and downwards in good periods mangers smooth income, as argued by Trueman and Titman (1988). Their research concludes that if managers have the possibility to choose between two periods to recognize certain income in, he or she may prefer the choice that is expected to give a smoother income stream. Further, Koch (1981) state that the flexibility in the R&D accounting treatment give firms the possibility to systematically influence income from year to year and as a result, dampen the fluctuations in their reported income. Given these findings the subjectivity in the R&D treatments according to IAS 38 and capitalization of development expenditures may not be used in order to increase the value relevance of the figures in the financial statement as discussed by Lev and Sougiannis (1996) and Healey et al. (2002), instead R&D capitalization may be employed for income smoothing purposes as provided by Markarian et al., It should be noted that researches carried out in the U.S suggests that managers use earnings management in a way that increases the informativeness on earnings (Watts & Zimmerman, 1986) and even with the presence of earnings management the earnings reconciliation in the reporting according to IFRS could be considered to be value relevant (Capcun et al., 2008; Healey et al., 2002). Research provided on European countries suggests that earnings management is present and that the highest level of managerial discretion are showed in the countries with the weakest investor protection. In fact, Sweden is considered as one of the European countries with a relative low level of earnings management, due to the country s legal institutions (Capcun et al., 2008; Leuz et al., 2003). In this study, we do not intend to investigate the underlying reason for income smoothing nor the value relevance of reported earnings. Rather, we test whether R&D capitalization is used as a tool for income smoothing. 9

10 Hypothesis development Since previous literature provide evidence for that income smoothing occurs and suggest that R&D accounting is exposed to a possibility for earnings management (Callimacy and Landry, 2003; Koch, 1981; Markarian et al., 2008; Seybert, 2010), we argue that income smoothing takes place through R&D accounting. Although the link between managerial incentive and the choice of R&D accounting treatment is not well pronounced by prior studies, Nelson et al. (2003) suggest that accrual accounting is one of the most common strategies for earnings management. Since the R&D accounting treatment for development expenditures in Sweden is accrual based, the expected consequence will be that managers use R&D accounting to manage earnings. However, it should be noted that previous research suggests that earnings may be managed with a myriad of ways, thus accrual accounting is not the only approach to manage earnings. Because the criteria for capitalization is based on management judgment and allows for substantial amount of subjectivity as noted by Callimacy & Landry (2003), we assume that earnings management occurs through R&D accounting. Firms that prefer immediate expensing over capitalization can justify this approach by stating in the financial statement that the requirement for R&D capitalization according to RR 15 and later by IAS 38 is not fulfilled. Hence, the criteria s for R&D capitalization becomes a requirement only stated in the accounting standards and not always followed by managers in practice. If it is true that; when given managers allowance to exercise judgment in their financial reporting they will use the accounting standards to their own advantage and will choose the accounting policies that smooth their reported income. Since the magnitude of earnings is directly affected by the choice of R&D accounting and the decision whether to capitalize or expense development expenditures affects the profitability profile of the firm, we expect to see a negative relationship between R&D capitalization and a firm s change in profitability. We argue that firms are resistant to show on fluctuations in their earnings and as a consequence choose to smooth their income trough R&D accounting. We assume that managers are more prone to classify development as an intangible asset when the firms are less profitable than prior year (i.e. have lower ROA than the prior year) and as an expense when they are more profitable than the prior year (i.e. have higher ROA than the prior year). In order to report a smoother income, the choice to capitalize or expense development expenditures will be made on basis of the company s profitability development. The study tests the relationship between managerial incentives and the choice of R&D classification and we develop the following hypothesis: H1: There is a negative relationship between a firm s change in profitability and reported R&D capitalization. Firms exercise income smoothing in order to dampen the fluctuations in their reported earnings (Koch, 1981; Markarian et al., 2008) and a logical conclusion would be that the larger the fluctuations are in reported earnings (pre income smoothing behavior), the bigger the incentives are for income smoothing. Since fluctuations in earnings is associated to company risk in several respects (Beidleman, 1973; Foster, 1986; Ronen & Saden, 1981; Trueman & Titman, 1988), and greater fluctuations would indicate a higher level of risk, we assume that the motives for income smoothing are greater for firms with larger fluctuations in their pre-managed ROA. One might expect that firms with less variability in profitability will not find the need to manage their earnings in the same extent as the ones with greater fluctuations, since they are perceived as less risky. Based on the association between fluctuations in earnings and risk, we expect to see a stronger relationship between a firm s 10

11 change in profitability and R&D capitalization for firms with larger fluctuation in ROA than the ones with small fluctuations in ROA and we develop the following hypothesis; H2: The larger the fluctuations are in ROA, the stronger is the relationship with income smoothing behavior. It has been mentioned that in the Swedish setting, R&D accounting treatment has at all times more or less been based on management judgment and we argue that the more judgment management are allowed to exercise in the financial reporting, the more managerial discretion. Since capitalization of development was optional with BFN RI (period one) and became a requirement when specific criteria were met when converting to RR 15 (period two) and IAS 8 (period three), the R&D accounting according to BFN R1 is more subjective and we expect that income smoothing is more visible in this period. We test for whether it is true that when managers gives the opportunity to chose between two accounting methods, in this case R&D capitalization or expensing of development expenditures as in period one, managers chose the classification that smooth the reported income, as suggested by Trueman and Titman (1988). Since the degree of flexibility allowed in the R&D accounting is greater with BFN R1, we expect to see less significant changes in scaled earnings during this period than in the subsequent periods. Based on the different R&D accounting standards, we develop the following hypothesis: H3: There is less variability in scaled earnings in period one than in the subsequent periods. IV. Research methodology Specification of data The data for our statistical analysis is obtained from firms listed on the Stockholm Stock Exchange, currently known as NASDAQ OMX, Stockholm. We exclude financial firms and real estate concerns since they are less substantial on R&D expenditures and are therefore not relevant for our study. Furthermore, we exclude firms who do not report in SEK as there may be differences in the legislation for these firms. Under Swedish accounting rules, firms that invest in R&D must disclose such investment in the annual report and we searched for all annual reports of companies who disclosed R&D investments during each examined year 2000, 2004 and We collected data of the firms total R&D investment, the amount expensed (pre amortization and write-down) respectively capitalized for the examined years. The amount capitalized R&D expenditures for the research year was not always feasible since several companies did not disaggregate R&D capitalization from other intangible assets in the balance sheet nor in the notes 2. In these cases, we have excluded the firms. We are aware of that this may cause sample bias since we only include firms that choose to disaggregate development from other intangible assets and we do exclude the possibility that these firms may manipulate their earnings through R&D accounting. The consequence of the subjective presentation of intangible assets has been that firms in some cases present license for instance as capitalized development, whereas other firms presents license and development separately. We consequently only include firms who explicitly reports activated R&D in the financial report. Firms may undertake R&D activities by generate R&D internally, acquire 2 Recall; presenting intangible assets as an aggregated amount in the balance sheet is in accordance with IAS 38, who allows for aggregation and disaggregation of intangible assets if this gives more valuable information to users of the annual report. 11

12 R&D or obtain R&D in conjunction with business combination, and we only include capitalized internal generated R&D and acquired R&D, while we exclude acquired R&D in conjunction with business combination since the accounting treatment of these R&D expenditures has varied throughout our research periods 3. In order to handle the annual reports in an effective way, we only checked for R&D expenditures in the board of directors report, consolidated statement of comprehensive income, consolidated statement of financial position and and/or in the notes. Since the applied accounting approach for R&D expenditures has varied; we found it relevant to include all the different periods in our study. We understand that the data must be analyzed separately because of the reforms of Swedish accounting standards for R&D, and whether R&D capitalization is a tool for income smoothing is tested cross-sectional. Hence, the applied research periods are; , and The reason for including three year into each research period is based on the Swedish institutional setting for R&D activities. In order to enhance the comparability it was necessary to include the same number of years into each research period. Since Stockholm Stock Exchange is subject to changes (introduction of new companies, delisted companies, mergers and name changes), our sample consist of firms that are in business during the period in which they are included in, meaning that a firm not necessarily need be included in all three periods and may not be traded on Stockholm Stock Exchange today nor traded under the same name today as before. Because the financial information provided by DataStream did not contain all the information needed to conduct our study, we hand-collected and analyzed the original financial statements. For instance, R&D assets could not be identified using the electronic database, since R&D is listed as intangible assets and could not be obtained separately. The annual reports were obtained from the National Library of Sweden and our sample is restricted to the offering provided by the library. In those cases the library lacked certain annual reports, we searched for these in the firm s homepage to the extent this was possible. Due to the above mentioned criteria the sample of firms used in our statistical analysis consists of 21 firms in research period one, 52 firms in research period two and 59 firms in the last research period. Specification of variables In hypothesis 1, we test for whether classifying development as an expense or as an intangible asset is related to a firm s change in profitability and thus could be a tool for income smoothing. We perform a multivariate regression for each research period by using generally the same method as Markarian et al. (2008). We measure the strength of the negative relationship between R&D capitalization and a firm s change in ROA (ΔROA), using the statistical program Statistical Package for the Social Sciences (SPSS) as a tool. We calculate a number of variables to be used in the statistical test for each sample firm and our multiple regressions model is as following: R&DCapitalization i = a +b 1 ΔROA i + b 2 ROA i + b 3 firm size i + b 4 Leverage i + b 5 Growth + b 6 GrowthExpectation i + b 7 R&DTotal i + B 8 Beta i i denote each firm in our sample. 3 Before harmonizing and converting to IFRS, R&D in conjunction with business combination was treated as goodwill. According to IAS 38 intangible assets shall be disclosed separately. 12

13 We calculate the variable of interest, the independent variable and several control variables for each final year of our research period (2000, 2004 and 2009), in order to observe the relationship between the dependent and independent variable. Our dependent variable is R&D capitalization, which is calculated as the capitalized amount of R&D for the investigated year, divided by the firm s total assets. 4 ROA is a profitability measure calculated as the firm's operating income (before R&D depreciation and write-down) in relation to the firm's total assets and our independent variable consists of the change in return on asset in relation to the prior year, ΔROA 5. We included ROA in our regression as our first control variable in order to control for the impact of our sample firms profitability on the dependent variable, since previous literature suggest that one character of high capitalizing firms is that they are less profitable than high expensing firms (Aboody & Lev, 1998). One explanation mentioned in previous research is that firms with negative or low profitability could be motivated to capitalize R&D costs in order to appear financial stronger and that profitable firms avoid capitalization since the reported earnings might be perceived of less quality by the investors (see Cazavan- Jeny & Jeanjean, 2003). Our second control variable is firm size, measured as the natural logarithmic form of the firm s total assets at the fiscal year end. The reason for controlling for firm size is that prior research indicates that large firms tend to spend a substantial part of their R&D costs on basic research, maintenance and upgrades of their products. These costs are expensed and cannot be capitalized according to IAS 38 (see Cazavan-Jeny & Jeanjean, 2003). Leverage measured as debt divided by equity 6 is our third control variable. We include the ratio in our regression since it has been suggested that high leveraged companies have a stronger incentive to capitalize R&D expenditures (see Cazavan-Jeny & Jeanjean, 2003). Furthermore, we assume that firms with the highest level of growth are the ones undertaking a larger amount of R&D activities and we therefore control for growth, measured by the annual change of net sales in the research year compared to the prior year, which represents our fourth control variable. We also expect that investors growth expectations influence the R&D classification decisions, giving us our fifth control variable. We use the book-to-market ratio as a control for investor growth expectations. We expect firms with high (low) book-to-market ratio to have low (high) R&D intensity (see Cazavan- Jeny & Jeanjean, 2003). We introduce R&D total as our sixth control variable, as it is reasonable to expect that firms who undertake more R&D activities might have a higher probability to fulfill IAS 38 criteria for R&D capitalization, than firms investing less. R&D total is calculated as the total R&D investment undertaken by the firm during the research year divided by the firm s total asset (see Markarian et al., 2008). Finally, we include beta as a control variable for firm risk which we obtained from the database Six-Trust on a 24 month basis. 4 We choose the firm s total asset as a deflator since it has been used in prior earnings management studies. For instance the original Jones (1991) deflates accruals by total assets. As suggested by Markarian et al., 2008, an alternative to firm s total asset could be firm s income; however deflating by income could bias the result since this will exclude firms with negative operating income. 5 When we calculate ΔROA we assume that current ROA is pre-managed i.e. before the effects of R&D capitalization i.e. yearly amortization and write-down. ROA is calculated net of R&D capitalization effects since accrual accounting is a well known strategy for earnings management (Nelson et al., 2003). Worth noticing is that companies may use other earnings management techniques, beside R&D capitalization in order to achieve their goals. As a result, we do not assume that our so called pre-managed ROA is unaffected by other earnings management mechanism (see Markarian et al., 2008). 6 Debt to equity ratio measured by total liability divided by total shareholders equity. 13

14 Hypothesis 2 investigates whether the degree of fluctuation in ROA has an effect on firms income smoothing behavior. When testing hypothesis 2 we classify our sample firms into two groups; high-variability firms and low-variability firms. Since our sample consists of firms from different industries, we calculate the medians change in ROA 7 industrial wise in any given period, which is used as a benchmark and is considered as the industries normal level of change in profitability. This criterion implies that some industries are excluded due to a small sample, which prevented us to perform the calculations required. The industries included are; Industrials, Health Care and Information technology, while firms in the Consumer Discretionary and Materials sector are excluded. Furthermore, hypothesis 2 is only tested on firms listed in the last two periods ( and ) since our small sample size in period one ( ) prevented us to classify firms into high- respective low-variability group. In total, the high-variability group includes 22 firms whose change in ROA is distributed above the ninetieth percentile and below the tenth percentile and the low-variability group contains of 81 firms whose change in ROA is distributed within the ninetieth and tenth percentile. In order to test hypothesis 2 we perform a multivariate regression on the two groups using the same variables as in our first hypothesis. Hypothesis 3 tests for whether the degree of subjectivity allowed in the financial reporting by the different R&D accounting standards, BFN R1, RR 15 and IAS 38 have an effect on income smoothing behavior and if the reported income are considered as more smooth in research period one, following BFN R1 than in the subsequent periods following RR 15 and IAS 38. Hypothesis 3 is tested by measuring the variability in scaled earnings. We first calculate a yearly ROA for a three years period for each firm. Thereafter we compute the standard deviation of the yearly ROA for each firm in a given period in order to measure the variability in profitability on a company level. Thirdly, we calculate the average of all individual firms standard deviations in order to measure the average variability in firm profitability for each period. Finally, the standard deviation for the average variability in firm profitability is compared between the different research periods, where a lower average standard deviation is presumed to indicate on more income smoothing. 7 When calculating the median change in ROA for each industry we use the pre-managed ROA i.e. operating income in relation to the firm s total asset before R&D capitalization effects. However, we do not assume that our so called pre-managed ROA is unaffected by other earnings management methods. 14