THE VALUE RELEVANCE OF INVESTMENT PROPERTY FAIR VALUES

Transcription

1 THE VALUE RELEVANCE OF INVESTMENT PROPERTY FAIR VALUES Isabel Costa Lourenço 1 Assistant Professor Accounting Department, ISCTE Business School José Dias Curto Assistant Professor Quantitative Methods Department, ISCTE Business School Abstract 2 This study investigates whether the recognized cost, recognized fair value and disclosed fair value of investment property are priced differently by investors. An analysis is also made into whether the recognized fair value of investment property in four European countries with dissimilar characteristics is priced by investors differently from each other. Our sample is composed of listed real estate firms in France, Germany, Sweden and U.K. subject to the mandatory adoption of IFRS since The empirical results suggest that investors distinguish the recognized cost, the recognized fair value and the disclosed fair value of investment property, but they do not distinguish valuation implications of the recognized fair value of investment property in the four countries under analysis. KEYWORDS: Investment Property, Fair Value Model, Value-relevance 1 Corresponding author: Isabel Lourenço, Avenida Forças Armadas, Lisboa, Portugal, isabel.lourenco@iscte.pt. 2 We gratefully acknowledge participants at the 2007 European Accounting Association Congress (Lisbon) for their helpful comments on a preliminary version of this paper. 1 Electronic copy available at:

2 1. INTRODUCTION There is no international consensus on how investment property should be measured after recognition. Alternatives include the cost model and the fair value model. The key question underlying this issue is whether the fair value model provides more relevant information without a loss of reliability. There are some reservations about extending a fair value model to investment property, especially because certain property markets may not be sufficiently mature for a fair value model to work satisfactorily. The efforts to find international consensus on the appropriate model to measure investment property after recognition is hampered by a gap in our knowledge about the relative value relevance of different types of fair value estimates of investment property. This paper addresses this gap in the accounting literature by investigating: (i) whether the recognized cost, recognized fair value and disclosed fair value of investment property are priced differently by investors; and (ii) whether the recognized fair value of investment property in France, Germany, Sweden and the U.K. are priced by investors differently from each other. To this end, we estimate four valuation models based on the work of Ohlson (1995) and Feltham and Ohlson (1995), after some desegregations which allow to include the recognized cost and disclosed fair value of investment properties accounted for under the cost model, as well as the recognized fair values of investment property accounted by the fair value model, overall and in each of the countries under analysis. Our conclusions are based essentially on coefficient comparisons. The empirical study is conducted in the European setting where the international accounting standard IAS 40 allows firms to choose either the fair value model or the cost model (with disclosure of fair value in the notes) as their accounting policy for investment property. We select four European countries with dissimilar characteristics, namely France, Germany, Sweden and U.K. We analyse only the real estate industry, where investment property represents the majority of firms assets. The results of this research suggest that investors distinguish recognized amounts of investment property accounted for under the cost model and the fair value model. They also distinguish the 2 Electronic copy available at:

3 disclosed fair value and the recognized cost of investment property accounted for under the cost model as well as the disclosed fair value of investment property accounted for under the cost model and the recognized fair value of investment property accounted for under the fair value model. However, investors do not seem to distinguish the recognized fair value of investment property in the four countries under analysis. This research contributes to the literature examining attributes of fair value estimates for non-financial assets (e.g. Barth and Clinch, 1998) by comparing the way as investors valued recognized fair values of investment property in four countries with dissimilar characteristics. The remainder of the paper is organized as follows. Section 2 provides background information to the research. Section 3 describes the research design. Section 4 presents the findings and section 5 presents the summary and concluding remarks. 2. BACKGROUND The New Accounting Rule for European Listed Companies The European listed companies are required to apply IFRS instead of national GAAP as the basis for presenting their consolidated financial statements for the annual accounting period beginning on or after 1 January Thus, listed companies that report investment property should now apply the International Accounting Standard 40, Investment Property (IAS 40), issued in 2000 and amended in While this new regulation could lead to a significant change in the accounting practices in some countries, it is not expected to cause change in other countries. The IAS 40 defines investment property as a property (land or a building, or part of a building or both) held to earn rentals or for capital appreciation or both, rather than for use in the production or supply of goods or services or for administrative purposes or for sale in the ordinary course of business. This standard allows firms to choose the cost model and the fair value model as their accounting policy to measure investment property after recognition. 3

4 Investment property accounted for under the cost model is measured in the same way as occupied property, i.e., IAS 16 requirements on the cost model should be applied. Thus, investment property is carried on the balance sheet at its cost less any accumulated depreciation and any accumulated impairment losses. However, IAS 40 still requires firms to disclose the fair value of their investment property in the notes. Investment property accounted for under the fair value model is carried on the balance sheet at fair value. When applying this model, depreciation is forbidden and all changes in the fair value should be recognized in the profit or loss in the period in which they arise. Hence, all listed firms reporting investment property should now provide its fair value, either in the notes (cost model) or directly in the balance sheet (fair value model). They are also encouraged, but not required, to determine the fair value of investment property on the basis of a valuation made by an independent appraiser. Almost all the firms used in our analysis follow this recommendation 3. The IASB believes that information about both the fair value of investment property and the changes in its fair value is highly relevant to users of financial statements. However, when preparing IAS 40, the IASB decided to include a choice between the cost model and the fair value model for two main reasons (Exposure Draft 64). Firstly, this gives preparers and users time to acquire experience in using a fair value model. Secondly, it allows time for countries with less-developed property markets and valuation professions to mature. As IAS 40 enables firms to report investment property using either the cost model or the fair value model and requires disclosure of fair value for those firms using the cost model, a rare opportunity arises to investigate whether recognized cost, recognized fair value and disclosed fair value are priced differently by investors. The Old Accounting Rules for European Listed Companies 3 About half of the Swedish firms determine the fair value of investment property using an internal valuation model but, in order to provide further assurance and validation of the internal valuation, a significant part of their investment property is also valued by an external appraiser. 4

5 Prior to the adoption of IFRS, investment property was accounted for under national GAAP. The accounting policies used to measure investment property after recognition varied considerably across European countries. For example, in France, Germany and Spain investment property had to be treated as property, plant and equipment and measured by the cost model. In Italy and the U.K. investment property had to be treated separately from property, plant and equipment; however, while firms in the U.K. were required to use the fair value model, in Italy it was prohibited. In Sweden and the Netherlands, investment property also had to be treated separately from property, plant and equipment but firms were allowed to apply either the cost model (with disclosure of investment property fair values) or the fair value model. For more detailed information on the pre-ifrs domestic GAAP treatment for investment property see Muller, Riedl and Sellhorn (2008). As a result, firms in some European countries were already used to applying the fair value model or at least to disclosing investment property fair values while reporting investment property fair values is really something new in some other countries. Our analysis includes two countries from each one of these groups (Sweden and U.K. from the first group and France and Germany from the latter group). Related research This research contributes to the literature examining attributes of fair value estimates for nonfinancial assets. Barth and Clinch (1998) find that voluntary tangible assets revalued amounts for Australian firms are associated with share prices. They also provides evidence that, for non-financial and mining Australian firms, property (included in property, plant and equipment) measured at cost and at fair value are valued differently by investors. Owusu and Yeoh (2006) show that investors do not relatively value investment property for which unrealized gains were recognized in the income statement higher than those for which unrealized gains were recognized in revaluation reserve. Our study adds to this literature by comparing the way as investors valued recognized fair values of investment property in four countries with dissimilar characteristics. 5

6 Furthermore, Danbolt and Rees (2008) provide evidence that fair values are biased where valuation is ambiguous (tangible assets) and are more reliable where they are unambiguous (financial assets). Dietrich et al. (2001) provides evidence that fair value estimates by U.K. property firms employing external appraisers are less biased and more accurate than those reported by firms employing internal appraisers. Muller and Riedl (2002) provides evidence that the market perceives fair values estimates as more reliable when external as opposed to internal appraisers are employed, reflected in lower bidask spreads for firms employing external appraisers. Our study adds to this literature by analysing one setting where in almost all the cases the fair value of investment property is determined by independent appraisers. 3. RESEARCH DESIGN Sample and data The empirical analysis is conducted in the European setting where the international accounting standard IAS 40 allows firms to choose between the fair value model and the cost model for their accounting policy for investment property. As investment property represents the majority of firms assets in the real estate industry, we analyze only those firms classified as real estate firms in the worldscope database. This study uses only the listed firms required to apply IFRS since the accounting period beginning on or after 1 January The sample period comprises the first three years of mandatory adoption of IFRS. We select four European countries with dissimilar characteristics: an Anglo-Saxon country (U.K.), a Nordic country (Sweden) and two continental countries, one of which has a medium level of shareholder protection (France) while the other has a low level of shareholder protection (Germany). 4 Joos and Lang (1994) identify Germany and the U.K. as the originators, and arguably the most extreme examples of the two primary accounting philosophies worldwide, the Anglo-Saxon and the 6

7 Continental models. France is considered one intermediate example of those two models. Based on the antidirector rights index constructed by La Porta (1997; 1998) as a proxy for the level of shareholder protection in each country, the U.K. is classified as a country with a high level of shareholders protection, France and Sweden are countries with a medium level of shareholders protection and Germany is the country with the lower level of shareholders protection. Before the mandatory adoption of IFRS, firms in the U.K. and Sweden were already used to applying the fair value model, unlike Germany and France where local GAAP required all firms to apply the cost model. Firms with no data available for all the variables are excluded from the sample. To mitigate effects of influential observations on the estimated coefficients, we exclude all the observations in which absolute value of the DFBETA indicator is higher than 2. The DFBETAS are computed for all the regressions and for all the variables including the dummies. This results in the deletion of 3% of the observations. The final sample is composed of 224 firm-year observations distributed as follows: United Kingdom (93), France (55), Germany (40) and Sweden (36). The accounting data are from the firms annual reports, which were hand-collected from the firms websites. The share prices and the number of shares are from the Worldscope database. Empirical models Our aim is to determine both whether the recognized cost, recognized fair value and disclosed fair value of investment property are priced differently by investors, and whether recognized fair value of investment property in France, Germany, Sweden and the United Kingdom are priced by investors differently from each other. In line with previous research on the value relevance of accounting data that relies on the comparison of coefficients of different items on the balance sheet (e.g. Barth et al., 1998, and Landsman et al., 2008), we support our analysis on the following model: 4 Only in these four European countries more than 10 firms are found with information available for each of the 7

8 P (1) it = α 0 + α1 ASSETSit + α 2 IPit + α3 LIABILITIESit + α 4 NIit + ε it where P is the share price as of three months after the fiscal year-end 5, ASSETS represents the total assets excluding the recognized amount of investment property, IP is the recognized amount of investment property and NI is the net operating income, and all the variables are on a per share basis. The share price is a summary measure of information that is relevant to investors, while book value (assets minus liabilities) and earnings are measures that summarize the information reflected in financial statement. In order to assess whether the recognized cost and recognized fair value of investment property are priced differently by investors, we replaced the variable IP in (1) by two interaction terms resulting from the product of IP with two dummy variables, COST and FAIR VALUE, that assume the value 1 when the firm uses the cost model or the fair value model, respectively, and 0 otherwise. This procedure gives rise to the following model: P (2) it = α 0 + α1 ASSETSit + α2 IP x COSTit + α3 IP x FAIR VALUEit + α4 LIABILITIES it + α5 NIit + ε it We predict that all the coefficients in (2) are positive, except the liabilities coefficient that we expect to be negative and statistically significant, i.e., all the associated variables are value relevant. To check whether investors price the recognized amounts of investment property accounted for under the cost model and the fair value model differently, we test the equality of the interaction terms PI x COST and PI x FAIR VALUE coefficients. To assess whether the recognized cost and disclosed fair value are priced differently by investors, we introduced a new variable in the equation (2): the difference between disclosed fair value and recognized amounts of investment property accounted for under the cost model (IP_COST_DFVC), giving rise to the following model: three years analyzed in this study. 5 Untabulated findings reveal that our inferences are not sensitive to using prices as of fiscal year-end or as of three months after fiscal year-end. 8

9 P (3) it = α 0 + α1 ASSETSit + α2 IP x COSTit + α3 IP _ COST_DFVCit + α4 IP x FAIR VALUEit + α5 LIABILITIES it + α6 NIit + ε it We also predict that all the balance sheet estimated coefficients in (3) are statistically significant. If the recognized cost and disclosed fair value of investment property accounted for under the cost model are priced differently by investors, the IP_COST_DFVC coefficient should be statistically different from zero. To assess whether the recognized fair value and the disclosed fair value of investment property are priced differently by investors, we replace the variable IP x COST in (2) by a new variable: the disclosed fair value of investment property accounted for under the cost model (IP_COST_DISCLOSED FV), giving rise to the following model: P (4) it = α 0 + α1 ASSETSit + α2 IP_COST_DISCLOSED FVit + α3 IP x FAIR VALUEit + α4 LIABILITIES it + α5 NIit + ε it We also predict that all the estimated coefficients in (4) are statistically significant. We test the equality of IP_COST_DISCLOSED FV and IP x FAIR VALUE coefficients to check whether the disclosed fair value of investment property accounted for under the cost model and the recognized amount of investment property accounted for under the fair value model are priced differently by investors. Finally, and in order to assess whether the recognized amounts of investment property accounted for under the fair value model in France, Germany, Sweden and in the United Kingdom are priced by investors differently from each other, we replaced the interaction term PI x FAIR VALUE in (4) by four new composite variables resulting from the product of that interaction term with four dummy variables, FRANCE, GERMANY, SWEDEN and U.K., that assume the value 1 when the firm s country is the one selected, and 0 otherwise. This procedure gives rise to the following model: Pit = α0 + α1 ASSETSit + α2 IP_COST_DISCLOSEDFVit + α3 IP x FAIR VALUE x FRANCEit + α4 IP x FAIR VALUE x GERMANYit + 9

10 α 5 IP x FAIR VALUE x SWEDENit + α6 IP x FAIR VALUE x UKit + α7 LIABILITIES it + α8 NIit + ε i (5) We also predict that all the estimated coefficients in (5) are statistically significant, i.e., the associated variables are value relevant. We test the equality of the coefficients on interaction terms IP x FAIR VALUE x FRANCE, IP x FAIR VALUE x GERMANY, IP x FAIR VALUE x SWEDEN and IP x FAIR VALUE x UK in order to see if the recognized amount of investment property accounted for under the fair value model in each of this countries is priced differently by investors. Due to the heteroscedasticity problems of the observed errors, and following Easton and Sommers (2003), we estimate the models via a weighted least squares regression (WLS - the weight variable is the share price). The resulting regression specification no longer suffers from the coefficient bias and heteroscedasticity found in the unweighted regression. Furthermore, models (2), (3), (4) and (5) are estimated with country fixed effects. In each one of the models, we add the dummy variables FRANCE, GERMANY and SWEDEN which assume the value 1 when the firm s country is the selected country and 0 otherwise. These binary variables capture the unobserved variation between countries and thereby eliminate any bias which could arise due to different levels of sophistication in the four stock markets. The country reference is the U.K. and therefore the overall estimated intercept corresponds to it. The estimated coefficients for the dummies France, Germany and Sweden are added to the overall intercept in order to compute the intercepts for each of the other three countries. 4. FINDINGS Descriptive statistics Table 1 presents descriptive statistics for the variables that form the regression models and the number of nonzero observations for each variable. Variables amounts are all per share values and they are presented in euros. 10

11 These statistics indicate that, despite the revaluations, on average investment property represents the majority of firms assets. The mean values for the recognized amount of investment property accounted for under the cost value model and the fair model are and euros per share respectively. The mean value for the other assets, excluding investment properties, is euros per share. Table 1 also reveals a higher number of observations for recognized amounts of investments property accounted for under the fair value model than under the cost model (183 against 41 observations). The mean of disclosed fair values of investment property accounted for under the cost model is substantially higher than their recognized amounts (61.43 against euros per share). Thus, the adoption of the fair value model would have a significant economic effect for these companies. TABLE 1 Regression results Recognized cost versus recognized fair value Panel A of table 2 shows the equation (2) WLS estimation results including the estimated coefficients for the country fixed effects dummy variables. The Wald test for the interaction effects coefficients comparison and its significance appear in panel B. TABLE 2 Table 2, panel A, shows that the estimates for all the balance sheet coefficients are statistically significant and have the expected sign. Thus, both investment properties recognized at cost and recognized at fair value are value relevant. However, the p-value (0.000) associated with the Wald test 11

12 coefficients comparison presented in panel B, implies the rejection of the null hypothesis that the coefficients of investment properties recognized at cost and recognized at fair value are identical. This result suggests that investors distinguish recognized amounts of investment property accounted for under the cost model and the fair value model. Our finding is consistent and stronger than the one of Barth et al. (1998). This was already expectable as we analyse real state firms, while those authors use conventional firms for which revaluations have a marginal effect. Recognized cost versus disclosed fair value Table 3 shows the equation (3) WLS estimation results including the estimated coefficients for the country fixed effects dummy variables. TABLE 3 As shown in Table 3, the estimates for all the balance sheet coefficients are again statistically significant and have the expected sign. Not surprisingly, the estimate for the difference between disclosed fair value and recognized amounts of investment property measured by the cost model is positive and statistically different from zero. This result suggests that investors distinguish disclosed fair value and recognized cost of investment property accounted for under the cost model. Recognized fair value versus disclosed fair value Panel A of table 4 presents the equation (4) WLS estimation results including the estimated coefficients for the country fixed effects dummy variables. The Wald test for the interaction effects coefficients comparison and its significance appear in panel B. TABLE 4 12

13 Table 4, panel A, shows that both the recognized amount of investment property accounted for under the fair value model and the disclosed fair value of investment property accounted for under the cost model are value relevant. However, the p-value (0.000) associated with the Wald test coefficients comparison shown in panel B, leads us to reject the null hypothesis that the coefficients of the disclosed and recognized fair value of investment property are equal. Based on this result, we can conclude that investors distinguish the fair value disclosed in the notes and recognized in the balance sheet. Cotter and Zimmer (2003) analyse Australian real estate firms and find that managers are more likely to recognize (rather than just disclose) the current value of real estate when the current value estimate is more reliable. They also show that the market discounts disclosure compared to recognition of real estate revaluations. Further, Muller, Riedl and Sellhorn (2008) also find that European listed firms are more likely to choose the fair value model instead of the cost model to measure investment property after recognition when the property market in which they operate has higher liquidity. Our result is consistent with findings of these two previous studies. Recognized fair value in France, Germany, Sweden and U.K. Table 5 shows the equation (5) WLS estimation results including the estimated coefficients for the country fixed effects dummy variables as well as the Wald test for the interaction effects coefficients comparison and its significance. TABLE 5 As shown in Table 5, the estimates for all the balance sheet coefficients are statistically significant. Thus, investment properties recognized at fair value are value relevant irrespective of the firm s country. Further, the p-values associated with the Wald pairwise coefficients comparison tests (0.551; 13

14 0.674; 0.290; 0.926; 0.824; 0,766) presented in panel B of Table 5 reveals that we do not reject the null hypotheses that the coefficients of the recognized amount of investment property accounted for under the fair value model in two different countries are equal. This result suggests that investors do not distinguish valuation implications of recognized amounts of investment property accounted for under the fair value model in the four countries under analysis. Prior studies support the idea that strong shareholder protection attenuates management opportunism while weak shareholder protection exacerbates it and, thus, managers are more likely to manipulate accrual in a weak shareholder environment than in a strong one. Hung (2001) provides evidence that the use of accrual accounting negatively affects the value relevance of accounting numbers in countries with a weak level of shareholder protection. Further, Soderstrom and Sun (2007) argument that cross-country differences in the value relevance of accounting numbers are likely to remain following IFRS adoption because the value relevance of accounting numbers is a function of the firm s overall institutional setting, including the legal and political system of the country in which the firm resided. However, our results seem to suggest that in the European real estate industry, where the fair value of investment property recognized by almost all the firms is determined by independent appraisers, investors do not distinguish valuation implications of the recognized amounts in the four countries with dissimilar characteristics under analysis. 5. SUMMARY AND CONCLUSIONS There is no international consensus on how investment property should be measured after recognition. The key question underlying this issue is whether the fair value model provides more relevant information without a loss of reliability, ie, more value relevant information. This paper addresses a gap in our knowledge about the relative value relevance of different types of fair value estimates of investment property by exploring the European setting where listed firms are allowed to choose either the fair value model or the cost model (with disclosure of fair value in the 14

15 notes) as their accounting policy for investment property. We select four European countries with dissimilar characteristics, namely France, Germany, Sweden and U.K. The results of this research suggest that investors distinguish recognized cost and recognized fair value of investment property. They also distinguish the two alternative ways to report investment property s fair values (recognize or disclose). Finally, investors do not seem to distinguish the recognized fair value of investment property in France, Germany, Sweden and U.K. 15

16 REFERENCES Barth, M. and Clinch, G. (1998) Revalued financial, tangible and intangible assets: associations with share prices and non-market-based value estimates, Journal of Accounting Research, pp Cotter, J. and Zimmer, I. (2003) Disclosure versus recognition: the case of asset revaluations, Asian- Pacific Journal of Accounting and Economics. Danbolt J. and Rees, W. (2008) An experiment in fair value accounting: UK investment vehicles, European Accounting Review, pp Dietrich, J., Harris, M. and Muller, K. (2001) The reliability of investment property fair value estimates, Journal of Accounting and Economics, pp Easton, P. and Sommers, G. (2003) Scale and scale effect in market-based accounting research, Journal of Business Finance and Accounting, pp Feltham, G. and Ohlson, J. (1995) Valuation and clean surplus accounting for operating and financial activities, Contemporary Accounting Research, pp Hung, M. (2001) Accounting standards and value relevance of financial statements: an international analysis, Journal of Accounting and Economics, pp IASB (2003) Investment Property, International Accounting Standard 40, London. Joos, P. and Lang, M. (1994) The effect of accounting diversity: evidence from the European Union, Journal of Accounting Research, pp Landsman, W., Peasnell, K. and Shakespeare, C. (2008) Are asset securitizations sales or loans?, Accounting Review, pp Muller, K., Rield, E. and Sellhorn, T. (2008) Choosing cost versus fair value: international evidence from the European real estate industry upon adoption of IFRS, Paper presented at the AAA annual meeting, Anaheim, CA. Muller, K. and Riedl (2002) External monitoring of property appraisal estimates and information asymmetry, Journal of Accounting Research, pp

17 La Porta, R; Lopes-de-Silanes, F., Shleifer, A., Vishny, R. (1997) Legal determinants of external finance, The Journal of Finance, pp La Porta, R; Lopes-de-Silanes, F., Shleifer, A., Vishny, R. (1998) Law and finance, The Journal of Political Economy, pp Ohlson, J. (1995) Earnings, book values and dividends in equity valuation, Contemporary Accounting Research, pp Owusu-Ansah, S. and Yeoh, J. (2006) Relative value relevance of alternative accounting treatments of unrealized gains: implications for the IASB, Journal of International Financial Management and Accounting, pp Soderstrom, N. and Sun, K. (2007), IFRS adoption and accounting quality: a review, European Accounting Review, pp

18 TABLE 1 Descriptive statistics for the variables used in the empirical tests Mean Median Std. Dev. Max. Min. n PRICE ASSETS LIABILITIES NET INCOME INVESTMENT PROPERTY COST FAIR VALUE FRANCE GERMANY SWEDEN U.K DISCLOSED FV DFVC PRICE = share price as of three months after fiscal year-end. ASSETS = Total assets excluding recognized amount of investment property. LIABILITIES = Total liabilities. NET INCOME = Net operating income. INVESTMENT PROPERTY COST = recognized amounts of investment property accounted for under the cost model. FAIR VALUE = recognized amounts of investment property accounted for under the fair value model. DISCLOSED FV = disclosed fair value of investment property accounted for under the cost model. DFVC = difference between disclosed fair value and recognized amounts of investment property accounted for under the cost model. All variables are deflated by the number of shares outstanding. n indicates number of observations with nonzero amounts. 18

19 TABLE 2 Recognized cost versus recognized fair value of investment property Panel A: Regression results PRICEit = α0 + α1 FRANCEit + α2 GERMANYit + α3 SWEDENit + α4 ASSETSit + α5 IP x COSTit + α6 IP x FAIR VALUEit + + it α 7 LIABILITIE Sit + α8 NET INCOME + ε it Coef. Std. Error t-statistics p-value Adj. R 2 C FRANCE GERMANY SWEDEN ASSETS IP x COST IP x FAIR VALUE LIABILITIES NET INCOME Panel B: Test of coefficients equality results Restriction Wald test p-value α 5 = α PRICE = share price as of three months after fiscal year-end. FRANCE = Dummy variable that assumes 1 when the firm s country is France and 0 otherwise. GERMANY = Dummy variable that assumes 1 when the firm s country is Germany and 0 otherwise. SWEDEN = Dummy variable that assumes 1 when the firm s country is Sweden and 0 otherwise. ASSETS = Total assets excluding recognized amount of investment property. LIABILITIES = Total liabilities. NET INCOME = Net operating income. IP = recognized amount of investment property. COST = Dummy variable that assumes 1 when the firm use the cost model and 0 otherwise. FAIR VALUE = Dummy variable that assumes 1 when the firm uses the fair value model and 0 otherwise. All variables are deflated by the number of shares outstanding. 19

20 TABLE 3 Recognized cost versus disclosed fair value of investment property Regression results PRICEit = α0 + α1 FRANCEit + α2 GERMANYit + α3 SWEDENit + α4 ASSETSit + α5 IP x COSTit + α6 IP_COST_DFVCit + α 7 IP x FAIR VALUEit + α8 LIABILITIESit + α9 NET INCOMEit + ε it Coef. Std. Error t-statistics p-value Adj. R 2 C FRANCE GERMANY SWEDEN ASSETS IP x COST IP_COST_DFVC IP x FAIR VALUE LIABILITIES NET INCOME PRICE = share price as of three months after fiscal year-end. FRANCE = Dummy variable that assumes 1 when the firm s country is France and 0 otherwise. GERMANY = Dummy variable that assumes 1 when the firm s country is Germany and 0 otherwise. SWEDEN = Dummy variable that assumes 1 when the firm s country is Sweden and 0 otherwise. ASSETS = Total assets excluding recognized amount of investment property. LIABILITIES = Total liabilities. NET INCOME = Net operating income. IP = recognized amount of investment property. COST = Dummy variable that assumes 1 when the firm use the cost model and 0 otherwise. FAIR VALUE = Dummy variable that assumes 1 when the firm uses the fair value model and 0 otherwise. IP_COST_DFVC = difference between disclosed fair value and recognized amounts of investment property accounted for under the cost model. All variables are deflated by the number of shares outstanding. 20

21 TABLE 4 Recognized fair value versus disclosed fair value of investment property Panel A: Regression results PRICEit = α0 + α1 FRANCEit + α2 GERMANYit + α3 SWEDENit + α4 ASSETSit + α5 IP_COST_DISCLOSED FVit + + it α 6 IP x FAIR VALUEit + α7 LIABILITIESit + α8 NET INCOME + ε it Coef. Std. Error t-statistics p-value Adj. R 2 C FRANCE GERMANY SWEDEN ASSETS IP_COST_DISCLOSED FV IP x FAIR VALUE LIABILITIES NET INCOME Panel B: Test of coefficients equality results Restriction Wald test p-value α 5 = α PRICE = share price as of three months after fiscal year-end. FRANCE = Dummy variable that assumes 1 when the firm s country is France and 0 otherwise. GERMANY = Dummy variable that assumes 1 when the firm s country is Germany and 0 otherwise. SWEDEN = Dummy variable that assumes 1 when the firm s country is Sweden and 0 otherwise. ASSETS = Total assets excluding recognized amount of investment property. LIABILITIES = Total liabilities. NET INCOME = Net operating income. IP_ COST_DISCLOSED FV = disclosed fair value of investment property accounted for under the cost model. IP = recognized amount of investment property. FAIR VALUE = Dummy variable that assumes 1 when the firm uses the fair value model and 0 otherwise. All variables are deflated by the number of shares outstanding. 21

22 TABLE 5 Recognized fair value of investment property in France, Germany, Sweden and in the U.K. Panel A: Regression results PRICEit = α0 + α1 FRANCEit + α2 GERMANYit + α3 SWEDENit + α4 ASSETSit + α5 IP_COST_DISCLOSED FVit + + α 6 IP x FAIR VALUE x FRANCEit + α7 IP x FAIR VALUE x GERMANYit + α8 IP x FAIR VALUE x SWEDENit + + it α 9 IP x FAIR VALUE x UKit + α10 LIABILITIESit + α11 NET INCOME + ε it Coef. Std. Error t-statistics p-value Adj. R 2 C FRANCE GERMANY SWEDEN ASSETS IP_COST_DISCLOSED FV IP x FAIR VALUE x FRANCE IP x FAIR VALUE x GERMANY IP x FAIR VALUE x SWEDEN IP x FAIR VALUE x UK LIABILITIES NET INCOME Panel B: Test of coefficients equality results Restriction Wald test p-value α 6 = α α 6 = α α 6 = α α 7 = α α 7 = α α 8 = α PRICE = share price as of three months after fiscal year-end. FRANCE = Dummy variable that assumes 1 when the firm s country is France and 0 otherwise. GERMANY = Dummy variable that assumes 1 when the firm s country is Germany and 0 otherwise. SWEDEN = Dummy variable that assumes 1 when the firm s country is Sweden and 0 otherwise. UK = Dummy variable that assumes 1 when the firm s country is UK and 0 otherwise. ASSETS = Total assets excluding recognized amount of investment property. LIABILITIES = Total liabilities. NET INCOME = Net operating income. IP_ COST_DISCLOSED FV = disclosed fair value of investment property accounted for under the cost model. IP = recognized amount of investment property. FAIR VALUE = Dummy variable that assumes 1 when the firm uses the fair value model and 0 otherwise. All variables are deflated by the number of shares outstanding. 22