Fair Value or Cost Mode Drivers of Choice for Ias 40
European Accounting Review Vol. 19, No. 3, 461– 493, 2010 Fair Value or Cost Model? Drivers of Choice for IAS 40 in the Real Estate Industry A.
QUAGLI? and F. AVALLONE?? ? Department of Accounting and Business Studies (DITEA), University of Genova, Genova, Italy and ? ? Department of Computer and Management Science (DISA), University of Trento, Trento, Italy (Received September 2008; accepted February 2010) ABSTRACT The IFRS mandatory adoption in European countries is an excellent context from which to assess the validity of accounting choice theory, which postulates that information asymmetry, contractual ef? iency (agency costs) and managerial opportunism reasons could drive the choice. With this aim, we test the impact of these factors to explain the adoption of fair value for investment properties (IAS 40) in the real estate industry, taking into account the ‘revaluation’ option offered by IFRS1 and using historical cost without revaluations as a baseline category for comparison purposes. We select a sample of European real estate companies from Finland, France, Germany, Greece, Italy, Spain and Sweden, all ? rst-time adopters of the IFRS. Using a multinomial logistic model, we show that information asymmetry, contractual ef? iency and managerial opportunism could account for the fair value choice. Particularly, the most signi? cant ? ndings are that size as a proxy of political costs reduces the likelihood of using fair value while market-to-book ratio is negatively associated with the fair value choice. On the other hand, leverage, another typical proxy of contracting costs, seems not to in? uence the choice. This evidence con? rms the current validity of traditional accounting choice theory even if it reveals, in such a context, the irrelevance of the usual relations between accounting choice and leverage. . Introduction We analyse if the choice between cost or fair value for investment property under IAS 40 aims at (i) reducing agency costs (contractual ef? ciency Correspondence Address: A. Quagli, Department of Accounting and Business Studies (DITEA), University of Genova, Via Vivaldi 2, 16126 Genova (GE), Italy. E-mail: [email protected] unige. it 0963-8180 Print/1468-4497 Online/10/030461–33 # 2010 European Accounting Association DOI: 10. 1080/09638180. 2010. 496547 Published by Routledge Journals, Taylor & Francis Ltd on behalf of the EAA. 462 A. Quagli and F. Avallone easons), (ii) mitigating information asymmetries, as standard setters claim, or (iii) allowing managerial opportunism, typical motives de? ned by accounting choice theory (Holthausen, 1990; Fields et al. , 2001). Using a multinomial logistic regression, we test these hypotheses using 73 observations from real estate companies located in European countries (Finland, France, Germany, Greece, Italy, Spain and Sweden) which do not allow the fair value method in the pre-IFRS mandatory period in order to eliminate the in? uence of pre-existing fair value adoption. All these ? rms are ? sttime IFRS adopters, enabling us to compare the same accounting choice in a similar situation (? rst-time adoption). The mandatory adoption of IAS 40 (Investment properties) by European listed companies offers a unique opportunity to verify managers’ behaviour in a composite context of accounting choice. In fact, IAS 40 allows two alternative methods for appraisal of investment property assets: the cost method or the fair value method with recognition of fair value changes through pro? t and loss. Additionally, taking into account the IFRS1 ‘fair value as deemed cost’ option, the cost choice could be split into two lternatives: (i) historical cost without revaluation, (ii) historical cost with the IFRS1 option to revaluate investment property. This second option could represent a partial substitute for the fair value method, showing its effects only in equity without in? uencing pro? t and loss. 1 Thus, our model assumes the choice of applying historical cost without revaluating it as the referent outcome category to compare (Y ? 0), and forms logits comparing the choice of using historical cost with IFRS1 revaluations of investment property (Y ? 1) and fair value choice (Y ? 2) to it. Our ? dings suggest that all the rationales described by accounting choice theory (information asymmetry, contractual ef? ciency and managerial opportunism) drive the decision to adopt fair value. Indeed, regarding contractual ef? ciency reasons in particular, we ? nd that the larger the size (proxy of political costs), the less likely fair value is to be chosen, while leverage and consequent lenders’ protection seems to be insigni? cant for the choice. Furthermore, our results show that market-to-book ratio (MTBV) (proxy of information asymmetry) is negatively related to the fair value choice. This ? nding, that con? cts with existing literature, could be accounted for in the real estate industry due to the fact that high levels of MTBV in this context reveal growth opportunities associated with a fair estimation of investment properties and therefore with a low information asymmetry. Managerial opportunism behaviour, measured by a dummy variable for earnings smoothing, seems to have an in? uence on fair value choice. While all these variables seem to have an in? uence on the fair value choice, the same variables do not explain the choice of historical cost with the IFRS1 revaluation option in preference to the cost maintenance approach.
This paper offers various contributions to current literature. Firstly, to the best of our knowledge, it is one of the ? rst papers speci? cally focused on the choice Fair Value or Cost Model? 463 between cost and fair value in the IFRS context. We perform the analysis using a sample of ? rst-time IFRS adopters from several European countries adopting only the cost method in the pre-IFRS phase in order to both not limit the research to the traditional comparison between German and UK ? rms and eliminate the risk of in? uence from past experience.
Secondly, this paper introduces to the accounting choice literature a research designed to analyse the in? uence of multiple motivations (contractual ef? ciency, information asymmetry and managerial opportunism) for a multiple-choice environment (cost, cost with IFRS1 revaluation or fair value through pro? t and loss), testing through a multinomial logistic regression all the possible causes. Previous research, on the contrary, usually overlooks a comparison of multiple motivations (Fields et al. , 2001, pp. 290 – 291).
In other words, compared to existing studies we conduct an analysis using an innovative multiple motivations – multiple choices approach that better captures the complexity of accounting choices in management decisions. Finally, we contribute to the current debate on fair value showing which ? rm characteristics drive the choice of this method. While information asymmetries are the most discussed motives for fair value, we demonstrate the in? uence of contractual ef? ciency motivation as well as managerial opportunism, and the actual choices by ? ms demonstrate only a ‘partial enthusiasm’ towards fair value, even in a sector where liquid markets exist. The paper proceeds as follows. Section 2 concerns the literature related to our analysis. Section 3 goes on to describe the main features of IAS 40 and the preIFRS domestic GAAP of the countries sampled. Section 4 illustrates the development of our hypotheses, while Section 5 provides details on the empirical model design, variable de? nition, sample selection and data. Finally, Section 6 describes descriptive statistics, the main ? ndings and the robustness of the results. . Theory and Relation to Existing Research The choice between fair value and cost is a central topic in the current debate on accounting. Fair value is generally preferred due to the fact that ?nancial statements reveal a higher level of information (CFA Institute Centre, 2008),2 even if its adoption requires speci? c conditions: liquid markets, large database of available prices (Barth and Landsman, 1995; Ball, 2006), as well as new competencies in developing measurement models in the absence of liquid markets, making it possible to enhance estimate reliability (Schipper, 2005).
On the other hand, the reliability of fair value estimates is the most critical point (Martin et al. , 2006; Watts, 2006; Whittington, 2008), with the potential damage brought to the stewardship function of ? nancial statements. More generally, the demand for fair value has to be evaluated in its speci? c country context. The demand for fair value and the related preference for a higher level of information vs. reliability of ? nancial statements in Common law countries is quite different from the same demand in Code law countries (see Ball et al. 2000). 464 A. Quagli and F. Avallone Alternatively, a cost model seems more ef? cient in a contractual perspective because it reduces agency costs generated by creditors’ protection, political visibility, taxation and litigation (Watts, 2003; Qiang, 2007). Recent studies, however, seem to ignore the importance that the analysis of the adoption of IFRS evaluation alternatives could have in providing some more explanations for managers’ accounting choices and, consequently, for the progress of accounting choice theory.
Therefore, the choice between cost and fair value is a central topic in this sense. Following the framework of Francis et al. (2004), fair value and cost affect the properties of accounting numbers in a very different way. Fair value is more value relevant,3 and provides more predictable and timely earnings ? gures because it is more oriented towards future cash ? ows (derivable by the current value of some assets); on the contrary, the cost method approach supports conservatism, smoothness and the accrual quality, due to the recognition of value changes only if realized.
While it is dif? cult to suppose the impact on earnings persistence, depending on the size of fair value changes, the aforementioned aspects will give rise to different accounting behaviours. The information about future cash ? ows derived by fair value will be more appreciated in ? nancial markets (analysts and equity investors), because it will contribute to mitigate information asymmetries. On the other hand, the cost method is less costly and has more utility for income smoothing and contractual ef? ciency for which conservatism is a precious support.
In other words, each of these methods has, at a theoretical level, pros and cons and the actual choice will likely depend on ? rm-speci? c circumstances. The different impact of these two methods strongly implies the need of the accounting choice theory to investigate the topic. A powerful starting point for accounting choice investigation is offered by Holthausen (1990; see also Watts and Zimmerman, 1978; Fields et al. , 2001) who classi? ed in: (i) contractual ef? ciency (agency costs), (ii) information asymmetry and (iii) managerial opportunism, the reasons for accounting choices. i) Expectations derived from the accounting choice theory concerning the impact of fair value on contractual ef? ciency could lead to a supposed negative relationship: the choice of fair value could increase agency costs for several reasons. The greater income ? uctuations induced by fair value compared to the cost model could enhance the perceived risk by investors (European Central Bank, 2004) and, consequently, the cost of capital, as the high level of reported pro? ts could increase political costs due to higher company visibility (Hagerman and Zmijewski, 1979). Additionally, the doubtful veri? bility of fair value compared to cost measures, in some contexts (illiquid markets) could increase litigation and its related costs (Watts, 2003), as well as the fact that fair value through pro? t and loss could anticipate taxation costs. Furthermore, we can infer from the contractual ef? ciency reasons regarding lenders’ protection contrasting hypotheses on fair value preference. On the one hand (Watts, 2003; Qiang, 2007), lenders prefer Fair Value or Cost Model? 465 conservatism (thus the cost method) because it reduces the risk of distributing ? rm value through dividends.
On the other hand, fair value represents the current value of assets and it could be more ef? cient in negotiating for debt covenants. In this sense Christensen and Nikolaev (2008), basing their research on a sample of French and German multi-industry companies, ? nd that the fair value method is preferred by companies with high leverage and they account for this through information asymmetry: the current value of ? xed assets gives more thorough information about the ? rm’s solvency capability. In this sense, IFRS1 revaluation option could be a ‘partial’ substitute of IAS 40 fair value, that is, ? ms could use the conservative cost approach to guarantee lenders’ protection but they could opportunistically revaluate investment assets through IFRS1 to beat covenants or to give a signal about their solvency capability. In other words, while IAS 40 fair value is a ‘long-term strategy’ whose effects are uncertain (fair value could give rise to future revaluations or impairments), the IFRS1 option could be seen as a ‘short-term strategy’, the accounting consequences of which could be made available before its adoption (the revaluations ex IFRS1 option must exist at the transition date, that is, one year before the ? st exercise IFRS compliant). In this sense, this option would encourage opportunistic (and aggressive) accounting behaviour. All these propositions, however, could fail to be applied if we take into account that covenants use, on average, to exclude revaluation reserves in ? nancial ratios. (ii) Looking at asymmetries for market participants, measured by market-tobook ratio (MTBV), fair value could be preferred to cost method because of its higher and updated level of information divulgated to ? nancial statement users.
This is the main argument supporting the fair value primacy from a current standard setters’ viewpoint (Barlev and Haddad, 2003; Ball, 2006; Danbolt and Rees, 2008; Whittington, 2008). For this hypothesis, IFRS1 option could be a partial substitute for IAS 40 fair value, because of its in? uence on equity and, consequently, on MTBV. (iii) When a ? rm is choosing between cost and fair value, the managerial opportunistic accounting behaviour, previously demonstrated by income smoothing practices (Barth et al. , 1999; He? in et al. 2002; Graham et al. , 2005) is less likely with fair value through pro? t and loss, which obliges large earnings impact due to the volatility of market prices. However, the choice of the IFRS1 option in this sense should be irrelevant (thus not competing with fair value through pro? t and loss method), because this accounting option in? uences only equity and has no impact on pro? t and loss. Our objective is to test empirically how these multiple, and in part controversial, reasons (managerial opportunism, contractual ef? iency and information asymmetries) account for the choice of either fair value or the cost model due to the recent mandatory adoption of IFRS. In the typical discussion about IFRS, in 466 A. Quagli and F. Avallone fact, the power of fair value is recognized speci? cally regarding its potential to reduce information asymmetries (Whittington, 2008). Our analysis is based on the assumption that recognition is more value relevant than simple disclosure. Since IAS 40 requires footnote disclosure of fair value investment properties for ? ms adopting cost (see Section 3), it could be assumed that the choice between cost and fair value is not relevant, because the information about fair value is available for ? nancial statement users whatever the accounting policy chosen for investment properties. Nonetheless, our paper poses disclosure not equivalent to recognition according to the prevailing literature4 (for a review see Schipper, 2007). In all probability, the reasons can be found in a different reliability of data included in the footnotes relating to the balance sheet measures (Schipper, 2007). As af? med by Cotter and Zimmer (2003), speci? cally for revaluations of ? xed assets, ‘the value relevance of recognized revaluations is not due to recognition per se, but rather to the fact that the assets being revalued are more reliably measured’ (p. 1). 3. Main Features of IAS 40 and Differences with the Domestic GAAP of Countries Sampled IAS 40 is concerned with investment property that is property (land or a building) held to earn rentals or for capital appreciation or both, rather than for use as a site in which to run a manufacturing business or as a good to sell in the ordinary course of business.
The most relevant feature for our interests in IAS 40 is the evaluation method. IAS 40 permits evaluation of investment properties choosing alternatively: . fair value model, by which an investment property is measured, after an initial measurement, at fair value with changes in fair value recognized in the income statement and with no depreciation; . cost model, with the same rule as in IAS 16 (the property is to be measured after initial recognition at depreciated cost less any accumulated impairment losses).
This feature makes IAS 40 unique within the IFRS because it represents the only case where the two main evaluation criteria, fair value and cost, are alternatively admitted in their ‘pure’ form; the IAS 40 fair value re? ects its changes from one period to another in the income statement and not directly in an equity reserve as established by IAS 16 or IAS 38. As a consequence, managers are conscious that the choice between these accounting methods implies substantial variations in accounting results. As reported in the Basis for Conclusions, in the 2003 IAS 40 revision (par.
BC 12), the IASB discussed whether to eliminate the choice between the fair value model and cost model, thus implicitly enforcing the former as the only evaluation Fair Value or Cost Model? 467 method allowed. However, it was decided to leave the choice between the two approaches for two main reasons: the ? rst was to give preparers and users time to acquire experience before using a fair value model. Obviously, with regard to the practice of fair value assessment the second was to allow time for countries with less-developed property markets and valuation professions to mature.
The IASB planned to reconsider the option of using the cost model at a later date, in the light of ‘fair value supremacy’ pervading the International Accounting Standards. Nonetheless, the fair value primacy is notable for its disclosure clause, requesting the fair value of the investment property for the entities that choose the cost model, this means that an entity is obliged to assess fair value in all cases, which is a logical premise to permitting an easier transition to the fair value method at a later date.
Additionally, the entity has to declare in notes whether it applies the fair value model or the cost model and the methods and signi? cant assumptions applied in determining the fair value, including a statement whether the determination of fair value was supported by market evidence or was more heavily based on other factors (which the entity should disclose) relating to the nature of the property and the lack of comparable market data. The fair value method benchmarked by IAS 40 is a novelty for several European countries.
Our sample looks at domestic accounting rules; it is made up of companies from countries which allow only the cost method for investment property: Germany (Deloitte & Touche, 2001), Finland (KPMG, 2003a), France (KPMG, 2003b), Greece (Tsalavoutas and Evans, 2009), Italy (PWC, 2005), Spain (Perramon and Amat, 2007), Sweden (KPMG, 2005). More speci? cally, in Spain and Italy an asset revaluation credited to equity is permitted only if a special law allows it. In France a revaluation to equity is permitted only if it embraces all ? ed assets and the long-term ? nancial assets. In Greece, it is possible to revaluate ? xed assets to equity every four years following a revaluation index established by the Government. In Germany no revaluations are allowed. Finnish and Swedish GAAP permit a revaluation of properties credited to equity if their fair value exceeds cost in a permanent, signi? cant and reliable way. The choice of countries using only the cost model in the pre-IFRS mandatory phase allows us to eliminate the in? uence of any pre-existing in? ence of fair value adoption. 4. Hypothesis Development Following Section 2, we develop our hypotheses concerning: (i) ef? ciency reasons, in terms of both the reduction of political costs and the lenders’ protection, (ii) information asymmetry and (iii) managerial opportunism. 468 A. Quagli and F. Avallone (1) Contractual Ef? ciency Following the hypothesis that conservatism accounting should reduce agency costs through a greater lenders’ protection (Watts, 2003; Qiang, 2007), we suppose a negative correlation between leverage and fair value method.
We do not conjecture the opposite assumption (Holthausen and Leftwich, 1983) that in order to beat covenants, higher leverage could induce earnings increasing policies (like, in our speci? c context, the choice of fair value through pro? t and loss) because covenants usually do not take into account fair value revaluations (Citron, 1992; Christensen and Nikolaev, 2008). Thus, H1: The probability of choosing fair value decreases if company has a high leverage ratio level before IFRS adoption.
We do not posit any assumption on the relationship between leverage and the choice of historical cost with the IFRS1 option for the aforementioned exclusion of revaluation reserves in ? nancial ratios used by covenants. As already described in the part of Section 2 that looks at political costs, we can suppose from the literature that conservative accounting reduces political costs because the high level of reported pro? ts could affect them due to higher company visibility (Hagerman and Zmijewski, 1979; Watts, 2003). In order to verify the impact of political cost on fair value choice, we adopt the ? m size as an independent variable. The size per se has been mentioned speci? cally as a criterion for actions against corporations since several studies document that the magnitude of political costs is highly dependent on the size of corporation (Watts and Zimmerman, 1978). Thus, we conjecture that the political costs increase according to the company size; the larger it is the higher are the political costs and the lower is the probability that is advantageous to choose a fair value approach. Accordingly, our research proposition is: H2: The probability of choosing fair value decreases with the size of the ? m. Even in this case, we do not suppose any relationship between political costs and the choice of historical cost with the IFRS1 option, because this option has no impact on pro? t and loss. (2) Information Asymmetry If information asymmetry exists in the speci? c context investigated, managers could choose fair value in order to clearly inform the market about the ‘true’ value of the ? rm. So, under the assumption that disclosure is not equivalent to recognition (Schipper, 2007), a positive association between the choice of the fair value method and information asymmetry is assumed.
Fair Value or Cost Model? 469 Many studies (Smith and Watts, 1992; Amir and Lev, 1996) use market-tobook ratio (MTBV) as a proxy for information asymmetry, starting from the intuition that while market value captures the present value of growth opportunities, the book value approximates the value of assets in place. As a result, we posit that MTBV is positively related to information asymmetry and, consequently, positively related to fair value choice. Therefore, we assume: H3a: The probability of choosing fair value increases the more marked is the difference between market value and the book value of equity.
We could also develop a concurrent hypothesis to H3a, on the basis that, in this case, the choice of historical cost with IFRS1 option, in? uencing equity, could be a ‘partial’ substitute of fair value through pro? t and loss. Thus, we expect a positive association between the choice of historical cost with IFRS1 option and information asymmetry, as measured by MTBV ratio. H3b: The probability of choosing historical cost with IFRS1 option increases the more marked is the difference between market value and book value of equity. (3) Managerial Opportunism
From the theory we derive that managerial opportunistic accounting behaviour is demonstrated by income smoothing practices (Barth et al. , 1999; He? in et al. , 2002; Graham et al. , 2005) and we thus suppose that fair value through pro? t and loss with its volatile changes contrasts smoothing policies. So, a negative association between fair value choice and pre-IFRS earnings smoothing is expected. Hence: H4: The probability of choosing fair value decreases if managers reduce the variability of reported earnings using accruals.
We do not suppose any relationship between managerial opportunism estimated by earnings smoothing and the choice of historical cost with IFRS1 revaluation, because this option has no impact on pro? t and loss. 5. Research Design Empirical Model and Variable De? nitions Two statistical procedures are used in our analysis: (i) the non-parametric Mann – Whitney two-sample rank-sum test is used to analyse the difference in explanatory variables between the group of ? rms that have adopted the fair value model or cost model with the IFRS1 revaluation and the group that have chosen the cost 470
A. Quagli and F. Avallone model (the cost group has been taken as a referent category). Additionally, (ii) we use a multinomial logistic regression model (MNL) to test the relationship between the ? rm accounting choice for investment properties and the hypothesized explanatory variables. Under the multinomial logistic model with three outcome categories (0, 1 and 2), p covariates and a constant term (b) denoted by the vector x, two logit functions are described as follows (Hosmer and Lemeshow, 2000): g1 (x) = ln[P(Y = 1| x)/P(Y = 0| x)] = b10 + b11 X1 + b12 X2 + . . . + b1p Xp (1) and 2 (x) = ln[P(Y = 2| x)/P(Y = 0| x)] = b20 + b21 X1 + b22 X2 + . . . + b2p Xp . (2) It follows that the conditional probabilities of each outcome category given the covariate vector are: P(Y = 0| x) = 1/1 + eg1 (x) + eg2 (x) P(Y = 1| x) = eg1 (x) /1 + eg1 (x) + eg2 (x) (3) P(Y = 2| x) = eg2 (x) /1 + eg1 (x) + eg2 (x) . Our model assumes the choice to use historical cost without revaluating as the referent or baseline outcome category to compare (Y ? 0), and forms logits comparing the choice to use historical cost with the IFRS1 revaluation of investment properties (Y ? 1) and fair value choice (Y ? 2) to it.
Furthermore, the model assumes the following relation between the proposed explanatory variables and the fair value accounting choice: ln[P(Y = FV| x)/P(Y = COST| x)] = b0 + b1 LEV + b2 SIZE + b3 MTBV + b4 SM + b5 CNT + b6 EPRA + b7 ACT + 1 (4) where b ? CHOICEi ? bFV; dependent variable equal to 2 if the ? rm i adopts fair value model under IAS 40 in ? rst-time adoption (FTA), 1 if ? rm i adopts the historical cost and uses IFRS1 to revalue investment properties and 0 if the ? rm i adopts the historical cost without revaluating; Fair Value or Cost Model? LEVi ? SIZEi MTBVi ? ? SMij ? CNTi ? EPRA ? ACT ? 471 he average debt to asset ratio for ? rm i, measured over two years before FTA; log of the average total asset over the two years before FTA; market-to-book value of ? rm i calculated over the last month of the FTA year since the market is in? uenced by the IFRS immediately after the FTA year; dummy variable coded 1 if ? rm i has an earnings smoothing index . the average index of earnings smoothing in country j (? rm’s country of domicile) and 0 otherwise; dummy variable coded 1 if ? rm i has an external market capitalization on GNP . the average external market capitalization on GNP for his legal country of origin (from La Porta et al. 1997) and 0 otherwise; dummy variable coded 1 if ? rm i is a member of the European Public Real Estate Association (EPRA) and 0 otherwise; ratio between total rents and total operating income estimated over the ? scal year preceding the IFRS mandatory adoption. Following Leuz et al. (2003) and Burgstahler et al. (2006) our proxy to capture earnings smoothing policies in the pre-IFRS period is computed as the ratio of the standard deviation of operating income divided by the standard deviation of cash ? ow from the operation, both measures being computed over the four years before IFRS mandatory adoption.
The ratio is then multiplied by 2 1 so that higher values are associated with higher earnings smoothing policies. Moreover, in order to capture the real signi? cance of the smoothing ratio (only values around zero denote strong earnings smoothing activities but the more the values decrease the more the smoothing signi? cance disappears), in our analysis for each ? rm we only measure the distance from the average value of the same ratio for the country of origin as measured in Burgstahler et al. (2006). So, the resulting dummy variable is equal to 1 if the ? m has an earnings smoothing index higher than the average index estimated for the country of origin and 0 otherwise. This procedure enables us to capture the peculiarity of each country due to the different local GAAP adopted before IFRS (Leuz et al. , 2003; Burgstahler et al. , 2006). We control for three variables we conjecture to affect the fair value choice by including them as independent variables in the model. Controlling for both the country of origin and the EPRA (European Public Real Estate Association) membership allows us to include two exogenous factors that could affect the fair value choice.
The former factor is considered because the differences in the nature of ? nancial systems around Europe are innate factors for international divergences in accounting (Nobes, 1998), thus in? uencing the fair value choice as well. The 472 A. Quagli and F. Avallone latter factor is considered because the EPRA’s Best Practices Committee encouraged the members to adopt fair value accounting to enhance uniformity, comparability and transparency of ? nancial reporting by real estate companies (EPRA, 2006). Additionally, it makes sense to control for the ? m activity since the business segments within the real estate industry could be considerably different (long-term investments, trading activity, development or services). With reference to the country (CNT), we do not use the distinction between Code Law Countries and Common Law Countries (Ball et al. , 2000), because our sample is entirely made up of Code Law Countries. Since accounting practices usually adhere to ? nancing systems (systems based on banks are generally more conservative than systems based on markets), we decided to capture the country effect with the level of ? ancial market development. So, following Nobes (1998), we theoretically classify countries included in our sample in two groups: countries where the role of ? nancial markets is more developed (capital market-based systems) and countries where ? nancial markets are less developed (credit-based systems). We can assume that the adoption of the fair value method should be easier in capital market based systems, where the indirect cost of information production should be lower and the more developed market could better appreciate the informative content of fair value estimates. In order to summarize ? ancial market development, we use the same variable and values as in La Porta et al. (1997). Speci? cally, we ? rstly computed the ratio of stock market capitalization held by minorities to gross national product. Hence, the higher ratio value is associated with highly diffused equity and, as a consequence, with more ? nancially developed markets. Therefore, we adopt a dummy variable coded 1 if the ? rm has an external market capitalization on GNP higher than the average external market capitalization on GNP for its legal country of origin (from La Porta et al. , 1997) and 0 otherwise.
The stock market capitalization held by minorities is computed as the product of the aggregate stock market capitalization and the average percentage of common shares not owned by the top three shareholders in the 10 largest non-? nancial, privately owned domestic ? rms in a given country. The lack of availability of certain data forced us to use the same values estimated by La Porta et al. With reference to the EPRA membership, we only use a dummy variable (EPRA) that takes a value of 1 for ? rms that are EPRA members and 0 otherwise. Lastly, we control for ? rm activity (ACT).
Particularly, since real estate companies could operate in many businesses (renting out investment properties, services, trading of investment properties and development), we use a variable to discriminate the ? rms which generally rent out investment properties from ? rms that operate in trading, services and development. Thus, we use the ratio between total rents and total operating income as a proxy of ? rm activity. So, the high values of the ratio suggest that the renting activity may be considered the company’s core business while low values of the ratio express the opposite.
Both rents and total operating income are hand-collected from ? nancial statements for the ? scal year preceding the IFRS mandatory adoption and the latter Fair Value or Cost Model? 473 has been computed as the sum of rents, services, realized gains/losses on investment property sales and other operating revenues. In terms of empirical predictions, we conjecture a positive relationship between the fair value choice (CHOICE) and both ? nancial market development (CNT) and EPRA membership (EPRA). The present work makes no prediction with respect to the other control variable (ACT).
Table 1, Panel A presents the proxies used for independent variables and the predicted sign of each relation between covariates and fair value choice for investment properties under IAS 40. Moreover, Table 1, Panel B only shows the relations between independent variables and the choice to use historical cost with IFRS1 revaluation, if theoretically signi? cant. Sample and Data Our study focuses on a sample of real estate ? rms from countries where a systematic use of fair value model was not allowed for investment property assets by pre-IFRS domestic GAAP.
A sample of 76 companies was selected from a population of 216 European real estate companies listed in their own country of origin in December 2007 in the following stock markets: Finland, France, Germany, Greece, Italy, Spain and Sweden. In December 2007, the Datastream International database revealed 216 real estate ? rms from the countries that were analysed (235 items, of which 19 were paid rights, preferred share, etc. ). This sample was then screened against a set of conditions: (i) the availability of the full version of the ? rst ? ancial statement complying with IFRS, obtained from the corporate website or via a speci? c request to Investor Relators, (ii) investment property assets on the balance sheet (as de? ned by IAS 40) not equal to zero, and (iii) the full data availability in the Datastream International database. Of the original 216 ? rms, 40 had neither website nor IR contact, 26 had ? nancial statements not complying with IFRS in the period of analysis (2005– 2007), 7 had no investment properties, 27 failed to respond and 40 ? rms did not have complete availability of data in ? nancial statements or in the Datastream database.
Thus, only 76 ? rms had suf? cient information for the above-mentioned explanatory variables to be included in the sample. Table 2, Panel A shows the sample selection procedure. The described procedure clearly illustrates that our sample consists of the maximum number of companies for which it is possible to obtain suf? cient information for the analysis, starting from the initial number of companies identi? ed in the database (N ? 216). Nevertheless, our analysis could have introduced a selection bias if an association between ? rms’ disclosure policies (e. g. assuring the availability of the full ? ancial statement on the corporate website or replying to a speci? c request) and the accounting choice had existed. In order to remove any doubts, we test whether there is a difference in drivers of choice used in our analysis between ? rms that provide an annual report or disclose it 474 A. Quagli and F. Avallone Table 1. Proxies and predicted signs for explanatory variables. The variables are grouped according to the main hypotheses for fair value choice and for the choice to use historical cost with IFRS1 revaluation Hypotheses Predicted sign Proxies Explanatory variables Panel A: explanatory variables and fair value choice 1) Contractual ef? ciency The probability of choosing (H1) 2 Debt/asset LEV fair value model decreases (leverage) with higher leverage The probability of choosing (H2) 2 Log of total asset SIZE fair value model decreases with the size (2) Information asymmetry The probability of choosing (H3a) + Market-to-book MTBV the fair value model value increases the higher is information asymmetry (3) Managerial opportunism The probability of choosing (H4) 2 Earning SM the fair value model Smoothing decreases with the extent to Index (dummy which corporate insiders variable) reduce the variability of eported earnings (earnings smoothing) (4) Control variables Firm’s country of origin + External cap/ CNT (? nancial markets GNP (dummy development) variable) EPRA members (European + Yes/no (dummy EPRA Public Real Estate variable) Association) Firm activity ? Total rents/total ACT operating income Panel B: explanatory variables and historical cost with the IFRS1 option (2) Information asymmetry The probability of choosing (H3b) + Market-to-book MTBV the historical cost with value IFRS1 option increases the higher is the information asymmetry after request (the sample) and those that do not.
Of course, we could only test the difference between the variables we collected from the Datastream database because we do not have access to the ? nancial statements of non-disclosing ? rm. Thus, we do not control if a difference exists in ? rm activity (ACT) between sampled and non-sampled ? rms. Fair Value or Cost Model? 475 Table 2 . Sample selection procedure and breakdown by country Number Panel A: sample selection procedure European Real Estate Firms listed in their own country of origin in December 2007 in the following stock markets (source: Datastream): Finland, France,
Germany, Greece, Italy, Spain and Sweden (countries where systematic revaluation of investment properties was not allowed before the IFRS adoption) Excluding the ? rms: – not reporting under IAS/IFRS in the period of analysis (2005– 2007) – with no investment property assets (or with investment properties equal to zero) – with neither website nor IR contact – failing to respond – with insuf? cient data to estimate equation (3) (in ?nancial statements or in Datastream database) Per cent 216 100% 2 26 12% 27 3% 2 40 2 27 2 40 18. 5% 13% 18. 5% Final sample 76 35% Panel B: breakdown of sampled ? ms by country and the number (percentage) of companies selecting fair value, cost with the IFRS1 revaluation or cost method Country No. of sampled Weight Fair value Cost with the Cost (%) companies (%) (%) IFRS1 (%) Finland France Germany Greece Italy Spain Sweden Total 4 26 22 4 8 4 8 76 5 34 29 5 11 5 11 100 4 (100) 11 (42) 12 (55) 3 (75) 2 (25) 0 (0) 8 (100) 0 (0) 4 (16) 4 (18) 1 (25) 1 (12) 3 (75) 0 (0) 0 (0) 11 (42) 6 (27) 0 (0) 5 (63) 1 (25) 0 (0) Panel B shows the breakdown by country of the sample and the proportion of companies that select fair value, cost with IFRS1 revaluation or cost method without revaluating in each country.
We considered companies listed in: Helsinki (Finland), Paris (France), Frankfurt and Munich (Germany), Athens (Greece), Milan (Italy), Madrid (Spain) and Stockholm (Sweden). Of the original 67 non-disclosing ? rms (which have neither website nor IR contact or failed to respond), 34 ? rms did not have complete data availability on the Datastream database. Thus, only 33 ? rms had suf? cient information to be included in the test. For non-disclosing ? rms, we collected data using the same rules as applied in the sample and considering 2005 as a reference date unless companies were still not listed.
In that case the reference date has been considered as the listing year. The results show that disclosing ? rms (the sample) are not statistically different from the non-disclosing ? rms in terms of the explanatory variables we selected except for the size ( p-value of 0. 000). 476 A. Quagli and F. Avallone This result is consistent with the literature that shows disclosure levels are usually positively correlated with ? rm size because of the decrease in the cost of disclosure (Lang and Lundholm, 1993). However, we keep the variable in the analysis for two reasons. Firstly, the ? m size (our proxy for political costs) could have both a possible negative relationship with fair value choice and a positive relationship with earnings smoothing (Watts and Zimmerman, 1978). If we do not include the ? rm size in the analysis, a signi? cant negative relation between earnings smoothing and fair value choice could be observed even if the size were the true explanatory variable. Secondly, even if a difference between sampled ? rms and non-disclosing ? rms exists in terms of size, summary statistics show a deviation in size within the sample that does not affect the results of the analysis.
With reference to the control variables, among the non-sampled ? rms only one company is an EPRA member. This is an expected result because the EPRA’s objective is to establish best practices in reporting and to provide high-quality information to investors. The result, however, does not introduce a selection bias in the analysis because the sample is made of both EPRA’s members (34 ? rms, around 45% of the sample) and companies that are not (42 companies, 55% of the sample). For these reasons, the results validate our sample and suggest that the sample selection did not introduce a bias into the analysis.
We relied on two sources for obtaining data for tests: (i) the ? rst ? nancial statement compliant with IFRS and (ii) the Datastream database. The former source enables us to verify the ? rms’ fair value or cost method choice for investment properties (IAS 40), the choice of ‘fair value as deemed cost’ under IFRS1 and to hand-collect from notes the portion of revenue that is a result of rental activities. The latter source provides all the accounting and non-accounting data we need to de? ne the other explanatory and control variables.
Non-accounting data includes market-to-book ratio while the accounting data consists of leverage (debt to asset ratio), total asset, operating income and cash ? ow from the operation (the last two accounting numbers have been used to estimate the earnings smoothing ratio) and the revenues that come from rents. Since the aim of this study is to ? nd out why fair value might be preferred to cost under IAS 40, we have commonly used data which is not in? uenced by the choice. In order to make sense of this key assumption, we referred to different periods for market records and information collected from ? ancial statements when collecting data. Market data refers to the end of the FTA year because the market is in? uenced by IFRS immediately after the FTA year. In other words, immediately after the FTA ? nancial data under IFRS is actually disclosed in ? nancial statements (which explains why the market-to-book value is collected during the last month of the ? rst-time adoption ? scal year). Financial data was collected over the two ? scal years before the FTA. Two years of ? nancial data rather than one year is considered to be more representative of a ? rm’s general characteristics and, in particular, able to reduce the effects
Fair Value or Cost Model? 477 that might occur from any unusual or abnormal data from a single year. Only the earnings smoothing ratio required a longer period of time; we used a four-year time period before the FTA for both operating income and cash ? ow from operation in order to estimate the related standard deviations. These two values were then compared to detect any earnings smoothing propensity. Financial information about the Swedish ? rms is converted into euros on the date of download from Datastream. Market data was automatically converted by the Datastream database. 6. Analysis of Results Summary Statistics
Table 2, Panel B shows the sample by country breakdown and displays both the number and the proportion of companies that select fair value, fair value with IFRS1 or cost model, respectively, in each country. At ? rst glance, Table 2, Panel B seems to reveal some national patterns in explaining the selection between fair value, historical cost with IFRS1 and cost model without revaluating investment properties. Despite the relatively small number of companies selected in some countries, it has still been possible to observe that companies from Finland, Greece and Sweden are extremely prone to adopting the fair value method.
Conversely, Italian companies seem to prefer historical cost without revaluating, Spanish companies have a preference for historical cost with the IFRS1 choice to revalue investment properties, while companies from France and Germany, the main countries in our study in terms of number of companies examined, do not show an a priori preference. Thus, the results justify our choice to control for a country variable through the multivariate analysis. Table 3 presents summary statistics for the full sample of 76 ? rms.
It should be noted that the two variables, market-to-book value (MTBV) and leverage (LEV), give rise to outlying observations implied by the values in the minimum and maximum columns of the table. One ? rm in particular had problematic values of both MTBV (value below zero) and LEV (value above one), due to a negative book value and this observation was removed from the analysis. Additionally, we isolate the outlying observations by means of the three sigma (standard deviation) rule (Barnett and Lewis, 1994), thus separating companies which have x ? m(x) ? 3s(x) (5) where s(x) is the standard deviation of the variable (x).
To remove the possible effects of the outliers on the results, we present both the nonparametric analysis and the multinomial logistic regression excluding these values (N ? 73). 478 Variable Explanatory variables: LEV SIZE MTBV SM Control variables: CNT EPRA ACT Mean Std. dev. Minimum Q1 Median Q3 Maximum 0. 5881 12. 7876 1. 4739 0. 3684 0. 2802 1. 6774 1. 1350 0. 4855 0 8. 2765 2 0. 17 0 0. 4829 11. 9451 0. 965 0 0. 6015 12. 9058 1. 3 0 0. 7235 13. 9800 1. 615 1 2. 07 16. 6882 8. 94 1 0. 4736 0. 4473 0. 4999 0. 5026 0. 5005 0. 3492 0 0 0. 1834 0 0 0. 4551 1 1 0. 7778 0 0 0 1 1 1 LEV ? leverage; SIZE ? og of total asset; MTBV ? market-to-book value; SM ? earnings smoothing (dummy); CNT ? ?nancial market development (dummy); EPRA ? EPRA member (dummy); ACT ? ?rm activity. A. Quagli and F. Avallone Table 3. Summary statistics of explanatory variables for sampled ? rms (n ? 76) Fair Value or Cost Model? 479 Nonparametric Mann – Whitney Test To begin by analysing the characteristics of the ? rms that adopt the fair value method or the historical cost with the IFRS1 revaluation in comparison to those that adopt the historical cost without revaluation, we use a Mann–Whitney twosample rank-sum test.
In view of the small size of the three groups, a nonparametric alternative to a conventional t-test is justi? ed because of the less challenging assumptions it requires, although this test has some limitations of its own, including being somewhat less powerful than the t-test. Table 4 shows evident differences across our independent variables, some of which appear statistically signi? cant. Consistent with the information asymmetry hypothesis (H3a), the output shows that there is a statistically signi? cant difference in MTBV between real estate ? ms that choose the fair value method and real estate ? rms that adopt historical cost without revaluation (difference signi? cant at 0. 000 level). The analysis of both means and median for fair value and cost groups makes the direction of the difference clear (for the fair value group, a mean of 1. 203 and a median of 1. 11 against 1. 775 and 1. 49 for the cost group). The output exhibits a negative relation between the MTBV and the fair value choice, contrary to the prediction derived by the traditional meaning of MTBV as proxy for information asymmetry.
In fact, the usual interpretation of high MTBV ratios as a signal of information asymmetry is based on the existence of growth options well known by managers, not revealed by accounting rules and, consequently, not identi? ed by investors. In theory, more growth options for high-tech ? rms in particular, are supposed as a consequence of a large bulk of intangibles whose recognition in ? nancial statements is not allowed, even though investors can estimate their importance (Smith and Watts, 1992; Amir and Lev, 1996).
However, in the real estate industry the relevance of intangibles seems less important than in high-tech ? rms. The main assets are investment properties, whose fair value could be easily estimated by ? nancial analysts. In this context, the meaning of high MTBV ratios might be in direct con? ict with the original intuition. In the cost accounting systems before IFRS adoption, higher values of MTBV ratios revealed growth opportunities associated with a fair estimation of investment properties and therefore with a lower information asymmetry. Conversely, lower MTBV ratios for real estate ? ms adopting the cost method could feasibly be the effect of information asymmetries on investment properties value and managers could prefer to use fair value method to reduce these asymmetries. In more precise terms, under the assumption that disclosure is not equivalent to recognition (Schipper, 2007), lower MTBV ratios estimated before the IFRS adoption for real estate ? rms adopting historical cost should be the result of information asymmetries on investment properties value. Thus, lower MTBV ratios could justify the managers’ preference to the fair value method in order to reduce the asymmetries.
This reasoning makes it possible to demonstrate the validity of the hypothesis (H3a), even if the sign of the variable is opposite to the traditional interpretation of the relationship between MTBV and information asymmetry. 480 A. Quagli and F. Avallone Table 4. Mann– Whitney two-sample rank-sum test. Fair Value Group vs. Cost Group (NFV ? 38; NCOST ? 16) and Cost with IFRS1 revaluation vs. Cost Group (NIFRS1 ? 19; NCOST ? 16) Group Explanatory variables: LEV SIZE MTBV SM Control variables: CNT EPRA ACT Z-Statistics Pr . |Z| FV vs. COST IFRS_1 vs. COST FV vs. COST IFRS_1 vs. COST FV vs. COST IFRS_1 vs. COST FV vs. COST
IFRS_1 vs. COST 2 0. 114 1. 192 0. 682 0. 762 3. 543 1. 258 0. 814 2 1. 185 0. 909 0. 233 0. 495 0. 446 0. 000??? 0. 208 0. 415 0. 235 FV vs. COST IFRS_1 vs. COST FV vs. COST IFRS_1 vs. COST FV vs. COST IFRS_1 vs. COST 2 2. 018 2 1. 931 2 1. 007 0. 040 2 3. 523 2 0. 364 0. 043?? 0. 053? 0. 314 0. 968 0. 000??? 0. 715 This table presents the Mann–Whitney two-sample rank-sum test for both explanatory and control variables. ? , ?? and ??? indicate statistical signi? cance at less than 10%, 5% and 1% level, respectively. The sample (excluding the outliers) comprises 73 companies from seven countries, split into three groups: ? ms that adopt the fair value model (NFV ? 38), ? rms that choose the historical cost and use the IFRS1 option to revalue investment properties (NIFRS1 ? 19) and ? rms that adopt the cost model without revaluating (NCOST ? 16) for investment properties under IAS 40. LEV ? leverage; SIZE ? log of total asset; MTBV ? market-to-book value; SM ? earnings smoothing (dummy); CNT ? ?nancial market development (dummy); EPRA ? EPRA member (dummy); ACT ? ?rm activity. Furthermore, both the ? nancial market development (CNT) and the ? rm activity (ACT) appear statistically signi? ant as well, with a difference signi? cant at 0. 043 and 0. 000 levels, respectively. The analysis of the means and median for CNT (mean of 0. 5526 and median of 1 for the fair value group against a mean of 0. 25 and median of 0 for the cost group) also shows a direction for the difference consistent with our assumption. Particularly, more developed ? nancial markets (estimated as in La Porta et al. , 1997) with the ratio of stock market capitalization held by minorities to GNP) seem to facilitate the adoption of fair value. Hence, the companies from countries where the role of ? ancial markets is more developed (capital market based systems) appear to view the fair value method more favourably than companies from countries where the markets are less developed (credit-based systems). With respect to the ? rm activity (ACT), we made no prediction of the sign. Both the output and the analysis of the mean and the median (mean of 0. 6558 and median of 0. 7507 for the fair value group against a mean of 0. 3005 and median of 0. 2756 for the cost group) show a positive direction of the difference. The result suggests that the predominant activity of the ? rms that choose the fair
Fair Value or Cost Model? 481 value model seems to be investment properties’ rental instead of other activities such as development and trading. Renting out properties implies a longer time period than other activities like development or trading, where assets would typically be sold in a shorter time. Thus, we could interpret the relation with fair value choice as the ? rms need to show the market value of their properties on the balance sheet when their realization will be in a longer time (rental activity). This would reduce the information asymmetry otherwise existing if properties were evaluated at cost.
Conversely, when the business is more concentrated on development and trading, the need for fair value recognition is less strong, due to a shorter time horizon for the realization of these assets. Further explanatory variables, such as leverage (LEV), dimension (SIZE) and earnings smoothing (SM), appear not to be signi? cant in the univariate analysis. However, even if not signi? cant it seems interesting to highlight that for both the size (SIZE) and the earnings smoothing (SM) the analysis of the mean and median reveals differences coherent with our research proposition, hence larger size and earnings smoothing for ? ms adopting historical cost (for size, mean of 12. 781 and median of 12. 905 for the fair value group against mean of 13. 167 and median of 12. 932 for the cost group; for earnings smoothing, mean of 0. 263 and median of 0 for the fair value group against mean of 0. 375 and median of 0 for the cost group). Except for the ? nancial market development (CNT), neither explanatory nor control variables seem to explain the managers’ choice to adopt the historical cost with the IFRS1 option to revalue investment properties rather than opting for historical cost without revaluation.
As for fair value choice, the analysis of the means and median for CNT (mean of 0. 578 and median of 1 for the IFRS1 group against a mean of 0. 25 and median of 0 for the cost group) shows a direction for the difference consistent with the idea that in countries where the role of ? nancial markets is more developed (capital market based systems), companies seem to view the revaluation of investment properties allowed by IFRS1 more favourably than in countries where the markets are less developed (creditbased systems). Multivariate Analysis
Before presenting the results of the multinomial logistic regression, we report the Spearman (rank) correlation coef? cients for the variables (Table 5). Considering the following multinomial logistic regression analysis, the dependent variable has been split into three variables: (i) CHOICE, equal to 0 if companies adopt the historical cost, 1 if companies adopt the historical cost with the IFRS1 revaluation and 2 if ? rms embrace the fair value; (ii) FV vs. COST that only regards as fair value choice (Y ? 1) and historical cost (Y ? 0) and (iii) IFRS1 vs.
COST that only takes into account the choice to adopt historical cost with the IFRS1 revaluation (Y ? 1) and the historical cost (Y ? 0). With reference to the dependent variable, Table 5 con? rms the previous univariate 482 Variables CHOICE FV vs. COST IFRS1 vs. COST LEV SIZE MTBV SM CNT EPRA ACT CHOICE – – – 0. 0552 2 0. 0620 2 0. 4343??? 2 0. 1728 0. 1890 0. 1462 0. 4707 FV vs. COST IFRS1 vs. COST – – 0. 0122 2 0. 0978 2 0. 4745??? 2 0. 0983 0. 2499? 0. 1356 0. 472??? – 2 0. 2045 2 0. 1306 2 0. 2131 0. 2033 0. 3311? 2 0. 0068 0. 0625 LEV SIZE – 0. 1780 0. 1657 2 0. 1818 2 0. 312 0. 0895 2 0. 1136 – 0. 0915 0. 0781 0. 2874?? 0. 3638??? 0. 0239 MTBV SM CNT EPRA ACT – 0. 0633 – 2 0. 0826 0. 2091? – 0. 1176 0. 2734?? 0. 0400 – 2 0. 2627?? 2 0. 0525 0. 4447??? 0. 1659 – This table provides Spearman (rank) correlation matrix for both explanatory and dependent variables. Considering the following multinomial logistic regression analysis, dependent variable has been split into three variables: CHOICE, equal to 0 if companies adopt historical cost, 1 if companies adopt historical cost with the IFRS1 revaluation and 2 if ? rms adopt the fair value; FV vs.
COST that only regards the fair value choice (1) and the historical cost (0) and IFRS1 vs. COST that only takes into account the choice to adopt the historical cost with IFRS1 revaluation (1) and the historical cost (0). Values indicated in bold show statistically signi? cant relationship between variables. ? , ?? and ??? indicate statistical signi? cance at less than 10%, 5% and 1% levels, respectively (two-tailed). Pearson correlation shows similar results. LEV ? leverage; SIZE ? log of total asset; MTBV ? market-to-book value; SM ? earnings smoothing (dummy); CNT ? ?nancial market development (dummy); EPRA ?
EPRA member (dummy); ACT ? ?rm activity. A. Quagli and F. Avallone Table 5. Spearman (rank) correlation matrix Table 6. Multinomial logistic regression results Panel A: model summary – goodness of ? t Number of obs. ? 73 LR chi2 (14) ? 41. 81 Prob . chi2 ? 0. 0001 Pseudo-R2 ? 0. 2799 Log-likelihood ? 2 53. 766523 Panel B: estimated coef? cients Variable Hypothesis 1 LEV SIZE MTBV SM CNT EPRA ACT Constant LEV SIZE MTBV SM CNT EPRA ACT Constant – – (H3b) – – – – 2 (H1) (H2) (H3a) (H4) Predicted sign + 2 2 + 2 + + ? Coeff. 2 0. 6117228 2 0. 4409383 2 0. 6115429 0. 2741504 1. 900273 0. 8488012 2 0. 708886 6. 279216 1. 734055 2 0. 6789767 2 1. 662586 2 1. 692808 1. 510263 2. 449269 2. 263975 8. 836272 Std. err. 1. 681102 0. 2748586 0. 5957133 0. 8804852 0. 9408805 1. 124734 1. 421061 3. 866769 1. 715306 0. 289514 0. 6609104 0. 9636362 0. 949826 1. 124299 1. 353768 3. 969725 z 2 0. 36 2 1. 60 2 1. 03 0. 31 2. 02 0. 75 2 0. 68 1. 62 1. 01 2 2. 35 2 2. 52 2 1. 76 1. 59 2. 18 1. 67 2. 23 P . |z| 0. 716 0. 109 0. 305 0. 756 0. 043?? 0. 450 0. 494 0. 104 0. 312 0. 019?? 0. 012?? 0. 079? 0. 112 0. 029?? 0. 094? 0. 026 95% conf. interval 2 3. 906621 2 0. 9796511 2 1. 779119 2 1. 451569 0. 0561811 1. 355637 2 3. 756117 2 1. 299512 2 1. 627883 2 1. 246414 2 2. 957947 2 3. 5815 2 0. 3513614 0. 2456834 2 0. 3893613 1. 055755 2. 683176 0. 0977746 0. 5560336 1. 99987 3. 744365 3. 053239 1. 81434 13. 85794 5. 095992 2 0. 1115397 2 0. 3672257 0. 1958842 3. 371888 4. 652855 4. 91731 16. 61679 483 (Continued ) Fair Value or Cost Model? LOGIT 484 Panel C: estimated odds ratios LOGIT Variable Odds ratio 1 LEV SIZE MTBV SM CNT EPRA ACT LEV SIZE MTBV SM CNT EPRA ACT 0. 5424156 0. 6434324 0. 5425132 1. 315413 6. 68772 2. 336844 0. 3787464 5. 663571 0. 5071357 0. 1896479 0. 1840021 4. 527923 11. 57988 . 621254 2 Std. err. 0. 9118557 0. 1768529 0. 3231823 1. 158201 6. 292345 2. 628327 0. 5382217 9. 714756 0. 1468229 0. 1253403 0. 1773111 4. 300739 13. 01925 13. 02494 z 2 0. 36 2 1. 60 2 1. 03 0. 31 2. 02 0. 75 2 0. 68 1. 01 2 2. 35 2 2. 52 2 1. 76 1. 59 2. 18 1. 67 P . |z| 0. 716 0. 109 0. 305 0. 756 0. 043?? 0. 450 0. 494 0. 312 0. 019?? 0. 012?? 0. 079? 0. 112 0. 029?? 0. 094? 95% conf. interval 0. 0201083 0. 3754421 0. 1687867 0. 2342026 1. 057789 0. 2577831 0. 0233743 0. 1963449 0. 2875341 0. 0519254 0. 0278339 0. 7037294 1. 278495 0. 6774894 14. 63149 1. 102714 1. 743742 7. 388093 42. 8214 21. 18385 6. 137025 163. 3658 0. 8944559 0. 6926533 1. 216386 29. 13348 104. 884 136. 6346 Choice ? 0 (historical cost) is the base outcome. This table presents coef? cients/odds ratios from multinomial logistic regression (MLN). Our model assumes the choice to use the historical cost without revaluating as the baseline outcome category to compare (Y ? 0), and forms logits comparing the choice to use the historical cost with the IFRS1 revaluation of investment properties (Y ? 1) and their fair value choice (Y ? 2) to it. We present Wald statistics, log-likelihood and McFadden pseudo-R2. , ?? and ??? indicate signi? cance at less than 10%, 5% and 1% level, respectively. LEV ? leverage; SIZE ? log of total asset; MTBV ? market-to-book value; SM ? earnings smoothing (dummy); CNT ? ?nancial market development (dummy); EPRA ? EPRA member (dummy); ACT ? ?rm activity. A. Quagli and F. Avallone Table 6. Continued Fair Value or Cost Model? 485 analysis results. Our proxy for information asymmetry, the market-to book ratio (MTBV), has a strong negative association with fair value choice, thus discriminating the fair value model group from the cost model group.
The result con? rms the above-mentioned interpretation of this sign. Furthermore, both the ? nancial market development (CNT) and the ? rm main business (ACT) condition the choice as well. Conversely, the choice to adopt the historical cost with IFRS1 revaluation is not accounted for by the explanatory variables except for the ? nancial market development (CNT). With reference to independent variables, Table 5 shows that some statistically signi? cant