Recognition versus disclosure: evidence from fair value of investment property

Abstract

The application of International Accounting Standard 40, Investment Property, in the European Union created a unique setting to study the implications of a decision to recognize versus disclose financial statements’ items, because in this setting recognized and disclosed investment-property-related amounts share a common measurement base, i.e., fair value. I use this setting to (1) explore a firm’s choice to recognize versus disclose fair values of investment properties, (2) test whether recognized and disclosed amounts are valued equally by investors, and (3) determine whether these amounts exhibit equivalent associations with future financial outcomes. To correct for self-selection concerns and assure I compare analogous amounts, I develop a selection model and construct investment-property-related amounts that differ only in whether their components are recognized or disclosed. I find that (1) contractual and asset-pricing incentives help to explain the recognition versus disclosure choice, (2) investors place smaller valuation weights on disclosed amounts, and (3) recognized and disclosed amounts exhibit statistically equivalent associations with future changes in net rental income and cash flows from operations. Taken together, the evidence suggests that managers are opportunistic in making the recognition versus disclosure choice and that even when recognized and disclosed amounts share an equivalent measurement base and are equally relevant for future financial outcomes, investors weight disclosed information less heavily in determining a firm’s value.

1 Introduction

The application of International Accounting Standard (IAS) 40, Investment Property, in the European Union (EU), with mandatory adoption of IFRS, effective January 1, 2005, created a unique setting where recognized and disclosed investment-property-related amounts share a common measurement base, i.e., fair value. IAS 40 allows firms to choose between the cost and fair-value models to account for investment properties, while requiring firms that apply the cost model to disclose the fair values of these assets in the notes. For firms that, prior to the transition to IFRS, applied the cost model for investment properties without disclosing their fair values, the provisions of IAS 40 offered, at the transition to IFRS,¹ a choice between recognition and disclosure of fair values of investment properties and of related revaluation gains.² Hence firms that choose to apply the fair-value model (i.e., adopt the recognition regime) exhibit higher book values of equity and more volatile net incomes than firms that choose to apply the cost model (i.e., adopt the disclosure regime).

I use the setting created by first-time adoption of IAS 40 to (1) explore factors associated with a firm’s choice to recognize versus disclose fair values of investment properties and of related revaluation gains, (2) test whether recognized and disclosed amounts are valued equally by equity investors (hereafter, “market valuation tests”), and (3) determine whether these amounts exhibit equivalent associations with future financial outcomes, such as future changes in net rental income (hereafter, “value relevance tests”). Addressing these questions in the context of investment properties, as reported under IAS 40, has the potential to provide new, substantiated insights into the drivers of a choice of recognition versus disclosure and the market valuation and value relevance of recognized and disclosed amounts in a setting where, regardless of the reporting form, the measurement base is held constant.

The setting studied here has several distinctive features. First, it permits investigation of both statement of financial position (SFP) and income statement (I/S) amounts (i.e., fair values of investment properties and changes in these fair values) and not one type of amount in isolation. Second, since investment properties are assets with identifiable financial outcomes, it allows an investigation of differences in association between recognized and disclosed amounts with related future financial outcomes. This feature enables, for the first time, to test whether the potential differences in market valuation of recognized versus disclosed amounts are due to differing value relevance of these amounts for future outcomes. Third, disclosure requirements of IAS 40 permit construction of analogous SFP and I/S amounts, so that the market valuation and value relevance tests are based on comparing SFP and I/S amounts that differ only in whether their components are recognized or disclosed. Fourth, fair values of investment properties and related revaluation gains or losses often represent a substantial fraction of firms’ total assets, allowing the drawing of insights based on material SFP and I/S amounts.³

To explore factors associated with a choice of recognition versus disclosure, I build on accounting choice literature (e.g., Watts and Zimmerman 1978, 1979; Aboody et al. 2004; Choudhary et al. 2009), which prescribes two sets of incentives as potential drivers of the accounting choice subject of this study. Accordingly, my first hypothesis is that firms with stronger contractual incentives (e.g., those closer to violating debt covenants that are tied to SFP amounts) are more likely to adopt the recognition regime. My second hypothesis is that firms with stronger asset-pricing incentives (e.g., those with more potential to improve the reported SFP and I/S amounts) are more likely to adopt the recognition model. My third and fourth hypotheses, stated in the null form, concern the market valuation and value relevance of recognized versus disclosed amounts.

My research strategy for testing for differences in market valuation and value relevance of recognized versus disclosed SFP and I/S amounts consists of two elements. First, building on my insights regarding factors associated with a choice of recognition versus disclosure, I develop a selection model which I use, as prescribed by Heckman (1979), to control for potential self-selection issues. Second, holding measurement base constant at fair value, I construct analogous SFP and I/S investment-property-related amounts, which differ only in whether their components are recognized or disclosed.⁴

To test my hypotheses, I hand-collect investment-property-related data items that are necessary to construct analogous SFP and I/S amounts from annual financial statements of publicly traded firms in four large EU economies whose domestic pre-IFRS GAAP prescribed the cost model for investment properties (without disclosing fair values). The countries are France, Germany, Italy, and Spain. To increase the power of my tests, I focus on a homogeneous group of firms that are classified as either “Real Estate and Investment Services” or “Real Estate Investment Trusts” on the Thomson Financial Worldscope database (hereafter, “Worldscope”).⁵ At the transition to IFRS, about 47 % of sample firms adopted the disclosure regime, and there do not seem to be notable differences in either the composition of their investment property portfolios or measures that reflect the materiality of investment-properties-related activities relative to other activities.

Consistent with my first and second hypotheses, I find that contractual and asset-pricing incentives help explain the choice. Regarding contractual incentives, firms with higher leverage (a proxy for the proximity of the firm to debt covenant violation) and firms with a more dispersed ownership (a proxy for the extent to which firms’ managers are monitored) have a significantly higher probability of adopting the recognition regime. Regarding asset-pricing incentives, my results indicate that firms whose operating income relative to cash flows from operations was less smooth prior to the transition to IFRS and whose potential investment-property-related gains were larger have a significantly higher probability of adopting the recognition regime.

With respect to differences in market valuation, controlling for self-selection concerns, I find evidence that equity investors place smaller valuation weights on disclosed than on recognized analogous SFP and I/S amounts. This suggests that market participants do not, on average, perceive disclosed information as a substitute for recognized information in determining a firm’s value. Regarding the value relevance tests, controlling for both self-selection concerns and potential differences in reliability (e.g., Dietrich et al. 2001; Muller and Reidl 2002), I show that recognized and disclosed analogous SFP and I/S amounts have statistically equivalent associations with 1- and 2-years-ahead changes in net rental income and cash flows from operations. This suggests that disclosed and recognized fair values of investment properties are equally relevant for future financial outcomes and that differences in relevance cannot explain the market’s undervaluation of disclosed information in determining a firm’s value.

Taken together, the evidence from my research suggests managers are opportunistic in making the recognition versus disclosure choice and that, even when recognized and disclosed amounts share an equivalent measurement base and are equally relevant for future financial outcomes, investors weight disclosed information less heavily in determining a firm’s value.

My inferences are the same when I exploit one of the unique features of the study’s setting and use a within-firm-changes design to test my third and fourth hypotheses. The within-firm-changes design helps to mitigate some self-selection concerns as well as potential differences in reliability related to omitted firm characteristics, which might be constant over time. Specifically, on average, sample firms that switch to recognition from disclosure experience an increase in valuation weights placed on analogous SFP and I/S amounts, even though the association between these amounts and directly related future financial outcomes remains the same. In addition, my inferences are robust to different model specifications, inclusion of additional control variables, and proxies.

This study contributes to three streams of prior research. First, it extends the accounting choice literature by illuminating factors associated with the choice to recognize versus disclose financial statement amounts when (1) essentially the same information is provided regardless of the choice, (2) both SFP and I/S amounts are affected by the choice, and (3) the choice concerns material amounts of financial statements.⁶ Second, this study adds to the recognition versus disclosure literature by testing market valuation of both SFP and I/S analogous amounts (1) that constitute a substantial fraction of a firm’s total assets and net income, (2) whose information content is independent of the firm’s choice of recognition versus disclosure, (3) that share a common measurement base (i.e., fair value), and (4) that have identifiable future financial outcomes (e.g., net changes in rental income).⁷ This paper’s analyses suggest that the common finding in prior literature that investors value disclosed amounts less heavily in determining a firm value cannot be attributed to their being informationally different (e.g., measured differently), constituting a negligible part of firms’ total assets and net incomes, or having different relevance for future financial outcomes.⁸ Third, this study extends asset revaluation literature, which had focused on testing for associations between recognized amounts and future outcomes, by investigating whether the relation between revalued amounts and directly related future financial outcomes differs for analogous SFP and I/S recognized versus disclosed amounts.

The remainder of the paper proceeds as follows. Section 2 provides a detailed description of the reporting requirements of IAS 40. Section 3 discusses related research and develops the hypotheses. Section 4 outlines the research design, while Sect. 5 describes the sample and provides descriptive statistics. Section 6 presents the primary findings, Sect. 7 reports findings from additional analyses, and Sect. 8 concludes.

2 Reporting requirements of IAS 40

IAS 40 defines investment property as property (land or a building, part of a building, or both) held by the owner or by the lessee under a finance lease to earn rentals or for capital appreciation or both, rather than for (1) use in the production or supply of goods or services or for administrative purposes or (2) sale in the ordinary course of business (IAS 40, para. 5).

After initial recognition at cost, IAS 40 permits firms to choose between the fair-value and cost models for recognition purposes and requires that they apply the chosen model to all investment property.⁹ Under the cost model, investment property is accounted for using the cost model as prescribed by IAS 16, Property, Plant and Equipment (PP&E), that is, at its initial cost less any accumulated depreciation and any accumulated impairment losses. Firms that choose to apply the cost model must disclose the fair value of investment property in the notes. Under the fair-value model, investment property is measured at fair value, with all changes in fair value recognized in profit or loss (i.e., all changes are reflected in changes of fair value and no depreciation is recorded under this model). After initial choice, firms can switch from the cost to the fair value model, but a switch from the fair value to the cost model is effectively prohibited.¹⁰

The reporting choices under IAS 40 affect both SFP and I/S reported amounts. Firms adopting the fair value model report, on the face of the I/S, revaluation gains/losses related to investment properties. Firms adopting the cost model usually do not report depreciation charges on the face of I/S in an identifiable way; however, these amounts can be extracted from the notes in which they provide separate information about the investment properties. Both types of firms disclose in footnotes the details of developments in investment property values (i.e., initial value for all firms; depreciation charges, impairment losses, and reversals for firms adopting the cost model; and changes in fair values for firms adopting the fair-value model) as well as rental revenues and related operating expenses incurred in obtaining them.

Compared to firms that adopt the cost model for investment properties, firms that choose to apply the fair-value model for investment properties exhibit higher book values of equity and more volatile net incomes. This is because characteristics of investment properties and the requirements of IAS 36, Impairment of Assets, together create a situation where, under the cost model, the firm recognizes revaluation losses but not gains, while under the fair-value model, the firm recognizes revaluation gains and losses. Figure 1 of “Appendix 2” illustrates the SFP and I/S effects of the choice of the accounting model.

Before the adoption of IFRS in the EU, investment property was accounted for under the domestic accounting standards of the firm’s country of domicile. Italy explicitly required that investment property be accounted for under the cost model (PwC 2008), while this was also the de facto requirement in France, Germany, and Spain, because they did not have a separate standard for investment property (KPMG 2003; Deloitte and Touche 2001).¹¹ Under the cost model, investment properties are presented on the SFP at cost less any accumulated depreciation and impairment losses.¹² None of the EU countries had domestic standards allowing or requiring the fair-value model, under which revaluation gains or losses are recognized in profit or loss.

For firms which, prior to the transition to IFRS, applied the cost model for investment properties without disclosing their fair values (e.g., France, Germany, Italy, Spain), the provisions of IAS 40 offered, at the transition to IFRS, a choice between recognition and disclosure of fair values of investment properties and of related revaluation gains. This choice, combined with the fact that under IAS 40 all firms provide essentially the same information about their investment properties,¹³ enables me to undo the accounting for reported SFP and I/S investment-property-related amounts to construct analogous SFP and I/S amounts. As illustrated in Fig. 2, “Appendix 2”, the information content of these amounts is independent of a firm’s choice of recognition versus disclosure, and they differ only in whether their components are recognized or disclosed (i.e., whether they belong to firms adopting the recognition regime or those adopting the disclosure regime).

3 Hypotheses development

3.1 Accounting choice literature: hypotheses related to factors associated with the choice of recognition versus disclosure

Accounting choice is defined as any decision whose primary purpose is to influence (either in form or in substance) the output of the accounting system in a particular way (Fields et al. 2001).¹⁴ Within this definition, the choice of recognition versus disclosure, subject of this study, is classified as an accounting method choice permitted by GAAP.¹⁵ The accounting choice literature identifies two main categories of incentives—contractual and asset pricing—that can explain the choices among accounting methods permitted by GAAP (e.g., Watts and Zimmerman 1978, 1979; Holthausen 1990; Fields et al. 2001; Aboody et al. 2004; Choudhary et al. 2009). I build on this literature in developing hypotheses concerning factors associated with the choice of recognition versus disclosure and in developing a selection model to be used in market valuation and value relevance tests.¹⁶

3.1.1 Contractual incentives

Contractual incentives stem mainly from the presence of agency costs and the absence of complete markets. Under the opportunistic contractual perspective, managers choose accounting techniques to reduce the extent to which accounting-based debt covenants are binding or to increase their compensation.¹⁷ Prior accounting choice literature provides evidence that firm-specific factors such as leverage (as a proxy for the proximity of the firm to violation of its debt covenant), the presence or absence of a management earnings-based compensation plan, and size (as a proxy for political attention) are associated with accounting choices. For example, Bowen et al. (1981) find that firms with financial ratios closer to likely debt constraints are more likely to elect to capitalize interest expenses. Dhaliwal (1980) reports that firms in the oil and gas industry that use the full-cost method and argue against the successful-efforts method have higher leverage ratios.¹⁸

Because the choice under IAS 40 concerns recognition versus disclosure of fair value of investment property (which is not lower than cost less accumulated depreciation and impairment losses), firms choosing the recognition regime enjoy higher book values of equity and higher total asset values. Such a reporting choice might allow managers to avoid violation of debt covenants that are tied to SFP amounts. This reasoning leads to the following hypothesis:

H1a

Ceteris paribus, the probability of a firm choosing to recognize the fair value of investment property and related revaluation gains is increasing in its proximity to violation of debt covenants that are tied to SFP values.

Prior research finds that managers select accounting methods opportunistically to increase their compensation (e.g., Matsunaga and Park 2001; Aboody et al. 2004; Choudhary et al. 2009). Because under the recognition regime the reported amounts are subject to substantial managerial discretion, managers whose compensation package is more closely related to reported amounts are more likely to opportunistically choose the recognition regime.¹⁹ Prior research also shows that outside monitoring counteracts the private incentives of senior managers (e.g., La Porta et al. 1998; Pollock et al. 2001; Muller et al. 2008; Choudhary et al. 2009). Specifically, firms with stronger external governance exert more control over management, among other things, by influencing the accounting method chosen (to limit management influence on reported numbers). Also, firms with stronger outside monitoring are more likely to use compensation contracts that are not based on reported amounts, thereby reducing management’s compensation-related incentives to adopt the recognition regime. This reasoning leads to the following hypothesis:

H1b

Ceteris paribus, the probability of a firm choosing to recognize the fair value of investment property and related revaluation gains is decreasing in the level of monitoring imposed on its management.

3.1.2 Asset pricing incentives

Asset pricing incentives are driven by information asymmetries between informed managers and less well-informed investors and creditors. Under the opportunistic information perspective, because accounting choices are made by self-interested managers, they might be used to increase (or reduce) earnings in a given period, to smooth earnings over time, to avoid losses or earnings declines, or a combination of these, among other strategies, to influence stock prices.²⁰

Although the choice under IAS 40 does not have direct cash flow implications, prior research shows that accounting method choices are associated with changes in stock prices (e.g., Aboody 1996; Espahbodi et al. 2002). In addition, Graham et al. (2005) report that corporate executives believe net income is the most important financial metric for public firms and they desire to report smooth earnings or meet earnings targets—to the extent that they are willing to sacrifice long-term firm value to achieve these goals. Because recognition of fair value of investment property increases earnings volatility, I raise the following hypothesis:

H2a

Ceteris paribus, the probability of a firm choosing to recognize the fair value of investment property and related revaluation gains is decreasing in its income smoothing.

At the adoption of IAS 40 (or at the switch from the cost to fair value model), the choice of recognition versus disclosure of fair value of investment property provides firms with an opportunity to increase the book value of equity and current profit or loss. Because firms make this choice after observing the fair values of investment properties and related revaluation gains, the opportunity to substantially increase the book value of equity and current profit or loss (i.e., adopt the recognition regime) is likely to be more tempting whenever such gains are high. Thus I offer the following hypothesis:

H2b

Ceteris paribus, the probability of a firm choosing to recognize the fair value of investment property and related revaluation gains is increasing in available revaluation gains.

3.2 Recognition versus disclosure and asset revaluation literatures: hypotheses related to market valuation and value relevance of recognized versus disclosed amounts

Generally, studies that examine whether disclosed and recognized amounts are valued equivalently by equity investors find that investors weight recognized amounts more heavily in determining a firm value.²¹ Aboody (1996) finds that oil and gas firms recognizing a write-down in connection with a decrease in oil prices experience a negative stock market reaction, whereas there is no significant stock market reaction for firms disclosing, but not recognizing, a write-down. Espahbodi et al. (2002) examine the equity price reaction to the pronouncements related to accounting for stock-based compensation. The authors document that firms exhibit significant abnormal returns around the issuance of exposure drafts proposing recognition of stock-based compensation costs and around the event reversing the decision to require disclosure only. Ahmed et al. (2006) use a sample of banks that have some recognized and some disclosed derivatives before SFAS 133 and find that the valuation coefficients on recognized derivatives are significant whereas the valuation coefficients on disclosed derivatives are not. Using the context of subsequent events, Michels (2015) finds that market prices are more sensitive to recognized values than to disclosed values.²²

Asset revaluation literature focuses primarily on the relation between fixed asset revaluation amounts and share prices, returns, or both. Easton et al. (1993) and Barth and Clinch (1998) find for a sample of Australian firms that, although revaluation increments have weak explanatory power for returns, the revaluation reserve has significant explanatory power for prices. They interpret their findings as evidence that Australian revaluations are value-relevant but not always timely. Aboody et al. (1999) focus on UK firms and find that upward revaluations of fixed assets are significantly positively related to changes in future firm performance, measured by operating income and cash flow from operations, and to annual returns and prices, indicating that revaluations are value-relevant for asset value changes.

Investment property assets, as reported under IAS 40, have three of the unique features: (1) they have identifiable, directly related future outcomes (e.g., net rental income); (2) by incorporating disclosed information, it is possible to construct analogous SFP and I/S amounts whose informational content is independent of the firm’s choice of recognition versus disclosure; and (3) they constitute a substantial fraction of the total assets and reported earnings, that is, meaningful SFP and I/S amounts. Equity investors might place equal or different valuation weights on recognized versus disclosed amounts depending on their associations with future firm performance. This is because stock prices and returns summarize investors’ assessments of firms’ asset values and expectations about future performance implied by these values (e.g., Barth and Clinch 1998; Aboody et al. 1999).²³

However, because in this setting recognized and disclosed amounts share a common measurement base—fair value—and have the same information content (i.e., the information content of analogous SFP and I/S amounts is independent of the recognition versus disclosure choice), controlling for potential differences in reliability (e.g., Choudhary 2011; Bratten et al. 2013),²⁴ there is, a priori, no reason that the two types of amounts will differ in their relevance for future related outcomes. Also, the recognized and disclosed amounts investigated in this study represent meaningful fractions of total SFP and I/S amounts, implying that equity investors are likely to pay attention to them (e.g., Hirshleifer and Teoh 2003) and invest time in studying them (e.g., Barth et al. 2003). This reasoning leads to the following hypotheses related to market valuation and value relevance, stated in the null form:

H3

Equity investors place equivalent weights on recognized and disclosed fair values and related revaluation gains in determining a firm’s value.

H4

Recognized and disclosed fair values and related revaluation gains exhibit equivalent associations with directly related future financial outcomes.

4 Research design

My research design consists of two elements: (1) exploration of factors associated with a choice of recognition versus disclosure and development of a selection model to correct for self-selection concerns in tests of H3 and H4, and (2) construction of analogous SFP and I/S amounts, to ensure tests of H3 and H4 are based on comparing analogous amounts.

4.1 Factors associated with the choice of recognition versus disclosure

To test H1a, H1b, H2a, and H2b and to develop a selection model, I conduct univariate analyses and estimate the following cross-sectional probit model:²⁵

$$\begin{aligned} \Pr (REC = 1) & = \varPhi (\beta_{0} + \beta_{1} LEV^{*} + \beta_{2} BLOCKHOLD + \beta_{3} INCSMOOTH \\ & \quad + \beta_{4} TRANSGAIN + \beta_{5} IP\_ACTIVITY + \beta_{6} Log(MVE) + \beta_{7} MTB^{*} \\ & \quad + \beta_{8} ROA + \beta_{9} REIT + \beta_{k} \sum\nolimits_{k} {IP\_TYPE} \\ & \quad + \beta_{p} \sum\nolimits_{p} {YEAR\_FE} + \beta_{q} \sum\nolimits_{q} {COUNTRY\_FE} ) \\ \end{aligned}$$

(1)

where Pr(·) denotes probability, and Φ(·) is the cumulative distribution function of the standard normal distribution. REC is an indicator variable that equals 1 if firm i chooses the recognition regime and 0 otherwise. All variables are measured as defined in “Appendix 1”.

To test H1a, I examine whether the coefficient on $LEV^{*}$, β ₁, is positive. $LEV^{*}$ is a firm’s leverage adjusted for the choice of recognition versus disclosure to include, for firms choosing the disclosure regime, the fair value of investment property and related revaluation gains. I follow prior literature and use leverage as a proxy for the proximity of the firm to debt covenant violation (e.g., Press and Weintrop 1990; DeAngelo et al. 1994; DeFond and Jiambalvo 1994; Aboody et al. 2004; Alissa et al. 2013).²⁶

To test H1b, I explore whether the coefficient on BLOCKHOLD, β ₂, is negative. BLOCKHOLD is the total percentage of a firm’s shares held by investors who hold 5 % or more of those shares. I follow prior literature and use the proportion of shares held by large blockholders as a proxy for the level of monitoring or outside control imposed on the firm’s managers (e.g., Muller et al. 2008; Choudhary et al. 2009; Core et al. 2014).

To test H2a, I investigate whether the coefficient on INCSMOOTH, a proxy for a firm’s income smoothing, β ₃, is positive. INCSMOOTH is the ratio between the standard deviation of a firm’s operating income and the standard deviation of its cash flows from operations during the last 4 years, minus its country of domicile real estate firms’ average, to control for country-specific GAAP or local effects.²⁷ If firms’ past income smoothing is related to the choice of recognition versus disclosure, the coefficient on INCSMOOTH will be positive. That is, firms with less smooth earnings (larger values of INCSMOOTH) in the past will be more inclined to adopt the recognition regime than firms with smoother earnings.

To test H2b, I check whether the coefficient on TRANSGAIN, β ₄, is positive. TRANSGAIN is the sum of net-of-tax difference between fair value and carrying amount of investment property at the beginning of the year of transition to IFRS or adoption of the recognition regime and $REVGAIN^{*}$, scaled by beginning-of-year market value of equity.

Equation (1) also includes control variables. First, as a control for the possible influence of the materiality of a firm’s investment-property-related assets/revenues with respect to its total assets/revenues and the composition of its investment property portfolio on its choice, Eq. (1) includes IP_ACTIVITY and IP_TYPE variables, respectively. Second, following prior literature’s finding that size is an important determinant of accounting method choice (e.g., political costs), Eq. (1) includes a natural logarithm of size, Log(MVE). Third, as a control for possible effects of risk, investment opportunities, firm profitability, and firm type, Eq. (1) includes $MTB^{*}$, ROA, and REIT. Finally, to control for potential country-level variation in IFRS implementation (e.g., Ball 2006; Daske et al. 2013) and related accounting method choices as well as time specific effects, Eq. (1) includes country- and year-fixed effects, COUNTRY_FE and YEAR_FE.

Because H1 and H2 state that the choice to recognize versus disclose fair values of investment properties and related revaluation gains is correlated with several factors related to contractual and asset pricing incentives, the setting studied in this paper raises self-selection-related concerns, which can affect the results concerning tests of H3 and H4. To address this concern, I use a two-stage procedure (Heckman 1979).²⁸ That is, I use Eq. (1) as a selection model to correct for self-selection in subsequent second-stage analyses (i.e., tests of H3 and H4). In particular, I calculate an inverse Mills ratio, InvMillsRatio, for each firm-year, separately for firm-years reporting under the recognition regime and for those reporting under the disclosure regime (and include the InvMillsRatio as a control variable in Eqs. (2)–(5)).

4.2 Analogous SFP and I/S investment-property-related amounts

To construct analogous investment-property-related amounts whose information content is independent of the recognition versus disclosure choice, I first identify reported SFP and I/S amounts that are affected by the firm’s choice of recognition versus disclosure. For firms choosing the recognition regime, such amounts include (1) fair value of investment property (SFP-related amount), (2) revaluation gains/losses (I/S-related amounts), and (3) disposal gains/losses (I/S-related amounts). For firms choosing the disclosure regime, such amounts include (1) the carrying amount of investment property (SFP-related amount), (2) depreciation charges (I/S-related amount), (3) impairment losses/reversals (I/S-related amount), and (4) disposal gains/losses (I/S-related amount). Next, I combine the SFP and I/S amounts that are affected by the choice of recognition versus disclosure with disclosed amounts, or undo the accounting, and achieve analogous SFP and I/S amounts. For example, for firms choosing the disclosure regime, I combine the SFP and I/S amounts that are affected by the choice of recognition versus disclosure with the following disclosed amounts: (1) fair value of investment property, (2) cost of investment property, (3) accumulated depreciation charges, and (4) current period subtracted accumulated depreciation due to sell.

The SFP-related analogous amounts include the (1) IPFV, which is the recognized (disclosed) fair value of investment property for firms adopting the recognition (disclosure) regime and (2) $TA^{*}$, computed as follows:

$$TA^{*} = \left\{ {\begin{array}{*{20}l} {TA - IPFV} \hfill & {{\text{if}}\;REC = 1} \hfill \\ {TA - IPBV} \hfill & {\text{Otherwise}} \hfill \\ \end{array} } \right.$$

The I/S-related analogous amounts include the (1) $NI^{*}$, which is the portion of a firm’s net income that is not affected by the choice of recognition versus disclosure, and (2) $REVGAIN^{*}$, denoting revaluation gains/losses of both types of firms, computed as follows:

$$\begin{aligned} NI^{*} & = \left\{ {\begin{array}{*{20}l} {NI - (REVGAIN + DISPGAIN) \times (1 - TAXRATE)} \hfill & {{\text{if}}\;REC = 1} \hfill \\ {NI - \left( {DISPGAIN - IPDEP\_CHRG - IMPAIRLOSS} \right) \times \left( {1 - TAXRATE} \right)} \hfill & {\text{Otherwise}} \hfill \\ \end{array} } \right. \\ REVGAIN^{*} & = \left\{ {\begin{array}{*{20}l} {(REVGAIN + DISPGAIN) \times (1 - TAXRATE)} \hfill & {{\text{if}}\;REC = 1} \hfill \\ {(\Delta IPFV - \Delta IPCOST + DISPGAIN - IPSUBT\_DEP) \times (1 - TAXRATE)} \hfill & {\text{Otherwise}} \hfill \\ \end{array} } \right. \\ \end{aligned}$$

“Appendix 2” provides a simplified example, which illustrates the construction of these amounts and shows how the information content of IPFV, $TA^{*}$, $NI^{*}$, and $REVGAIN^{*}$ is independent of the firm’s choice of recognition versus disclosure and these amounts differ only in whether they belong to firms choosing the recognition regime or disclosure regime.²⁹

4.3 Market valuation and value relevance tests of recognized versus disclosed amounts

To test H3, I estimate two regression models, treating them as second-stage analyses given the selection model developed above. First, I estimate the following pooled cross-sectional equation relating market value of equity to accounting amounts and allowing for different coefficients on the fair values of investment properties and related revaluation gains, depending on whether they relate to recognition or disclosure regime firm-years:

$$\begin{aligned} MVE_{i,t} & = \beta_{0} + \beta_{1} TA_{i,t}^{*} + \beta_{2} TL_{i,t} + \beta_{3} NI_{i,t}^{*} + \beta_{4} IPFV_{i,t} + \beta_{5} REVGAIN_{i,t}^{*} \\ & \quad + \beta_{6} DISC_{i,t} + \beta_{7} (DISC_{i,t} \times IPFV_{i,t} ) + \beta_{8} (DISC_{i,t} \times REVGAIN_{i,t}^{*} ) \\ & \quad + \beta_{9} InvMillsRatio_{i,t} + \beta_{p} \sum\nolimits_{p} {YEAR\_FE_{t} } + \beta_{q} \sum\nolimits_{q} {COUNTRY\_FE_{i} } + \varepsilon_{i,t} \\ \end{aligned}$$

(2)

where MVE is market value of equity 3 months after end of year, and DISC is an indicator variable that equals one for disclosure regime firm-years.³⁰ I use market value of equity 3 months after end of year to ensure all the accounting information is available to the market.³¹

One can view Eq. (2) as based on the Ohlson (1995) valuation model, where book value of equity is the difference between total assets and total liabilities (TL), and total assets and net income are disaggregated into (1) amounts that are not related to the choice of recognition versus disclosure of investment properties (e.g., $TA^{*}$, $NI^{*}$) and (2) those that are related to that choice but constructed in a way that achieves analogous amounts (e.g., IPFV, $REVGAIN^{*}$). Alternatively, one can view Eq. (2) as deriving from an asset-based valuation equation, where earnings is a proxy for unrecognized net assets (Barth and Landsman 1995). Both views lead to similar interpretations of the coefficients on IPFV, $REVGAIN^{*}$, and the related interaction variables. Specifically, if investors place similar valuation weights on disclosed and recognized amounts, β ₇ and β ₈ will not differ from 0. Based on prior research, I predict positive coefficients on $TA^{*}$, $NI^{*}$, IPFV, and $REVGAIN^{*}$ and a negative coefficient on TL.

To address the concern that inferences from Eq. (2) are attributable to intertemporally constant omitted variables (Landsman and Magliolo 1988), I also estimate the following pooled cross-sectional equation relating stock returns to current year net income and change in net income allowing for different coefficients on recognized versus disclosed amounts:³²

$$\begin{aligned} RETURN_{i,t} & = \beta_{0} + \beta_{1} NI_{i,t}^{*} + \beta_{2} REVGAIN_{i,t}^{*} + \beta_{3} \Delta NI_{i,t}^{*} + \beta_{4} \Delta REVGAIN_{i,t}^{*} \\ & \quad + \beta_{5} DISC_{i,t} + \beta_{6} (DISC_{i,t} \times REVGAIN_{i,t}^{*} ) + \beta_{7} (DISC_{i,t} \times \Delta REVGAIN_{i,t}^{*} ) \\ & \quad + \beta_{8} MTB_{i,t - 1} + \beta_{9} Log(MVE_{i,t - 1} ) + \beta_{10} InvMillsRatio_{i,t} \\ & \quad + \beta_{p} \sum\nolimits_{p} {YEAR\_FE_{t} } + \beta_{q} \sum\nolimits_{q} {COUNTRY\_FE_{i} } + \varepsilon_{i,t} \\ \end{aligned}$$

(3)

where RETURN is firm i’s year t equity return, measured from 3 months after year-end for year t − 1 to 3 months after year-end for year t, and Δ denotes annual change. This specification resembles those estimated by Easton et al. (1993), Barth and Clinch (1998), and Aboody et al. (1999), with MTB and MVE at the beginning of the period added as controls for the market-to-book and size effects in stock returns (e.g., Fama and French 1992). If revaluation gains provide timely information and the timeliness is similar across recognized and disclosed amounts, β ₆ and β ₇ will not differ from 0. Based on prior research (e.g., Easton and Harris 1991; Aboody et al. 1999), I predict β ₁, β ₂, β ₃, and β ₄ (β ₈ and β ₉) are positive (negative).

To test H4, I consider two investment-property-related identifiable future outcomes—net rental income and cash flows from operations—and I estimate two equations, controlling for differences in reliability. First, I estimate the following pooled cross-sectional equation:

$$\begin{aligned} \Delta NETRENTINC_{i,t + \tau } & = \beta_{0} + \beta_{1} REVGAIN_{i,t}^{*} + \beta_{2} IPFV_{i,t} + \beta_{3} \Delta NETRENTINC_{i,t} \\ & \quad + \beta_{4} DISC_{i,t} + \beta_{5} (DISC_{i,t} \times REVGAIN_{i,t}^{*} ) + \beta_{6} (DISC_{i,t} \times IPFV_{i,t} ) \\ & \quad + \beta_{7} BigFour_{i,t} + \beta_{8} ExtAppraiser_{i,t} + \beta_{9} InvMillsRatio_{i,t} \\ & \quad + \beta_{p} \sum\nolimits_{p} {YEAR\_FE_{t} } + \beta_{q} \sum\nolimits_{q} {COUNTRY\_FE_{i} } + \varepsilon_{i,t} \\ \end{aligned}$$

(4)

where $\Delta NETRENTINC_{i,t + \tau }$ is firm i’s net rental income in year t + τ minus net rental income in year t scaled by market value of equity at the beginning of year t, and τ equals 1 or 2.³³ I include only investment-property-related amounts to predict changes in net rental income because it is a direct outcome of investment property assets. If recognized and disclosed amounts are equally relevant for future financial outcomes, as measured by net rental income, then β ₅ and β ₆ will not differ from zero. Based on prior research, I predict β ₁, β ₂, and β ₃ are positive.

I next estimate the following pooled cross-sectional equation that focuses on changes in cash flows from operations, instead of changes in net rental income:

$$\begin{aligned} \Delta CFO_{i,t + \tau } & = \beta_{0} + \beta_{1} REVGAIN_{i,t}^{*} + \beta_{2} IPFV_{i,t} + \beta_{3} TA_{i,t}^{*} + \beta_{4} \Delta WC_{i,t} + \beta_{5} \Delta CFO_{i,t} \\ & \quad + \beta_{6} DISC_{i,t} + \beta_{7} (DISC_{i,t} \times REVGAIN_{i,t}^{*} ) + \beta_{8} (DISC_{i,t} \times IPFV_{i,t} ) \\ & \quad + \beta_{9} BigFour_{i,t} + \beta_{10} ExtAppraiser_{i,t} + \beta_{11} InvMillsRatio_{i,t} \\ & \quad + \beta_{p} \sum\nolimits_{p} {YEAR\_FE_{t} } + \beta_{q} \sum\nolimits_{q} {COUNTRY\_FE_{i} } + \varepsilon_{i,t} \\ \end{aligned}$$

(5)

Equation (5) resembles Eq. (4) except for the focus on changes in cash from operations, $\Delta CFO_{i,t + \tau }$, and the inclusion of $\Delta WC_{i,t}$, the scaled change in working capital from year t − 1 to year t, and the inclusion of $TA^{*}$. I include change in working capital because of the documented significant association between operating cash flows and lagged working capital accruals (e.g., Dechow 1994). I include $TA^{*}$ to control for potential effects of size. If recognized and disclosed amounts are equally relevant for future financial outcomes, as measured by changes in cash flows from operations, then β ₇ and β ₈ will not differ from zero. Following prior research, I predict β ₁, β ₂, and β ₄ are positive.

To strengthen the argument that the tests of associations of recognized and disclosed investment-property-related amounts with changes in net rental income and cash flows from operations (i.e., tests of H4) are attributed to relevance rather than reliability, I follow prior literature and include controls for an external appraiser, ExtAppraiser, and Big Four auditor, BigFour, which are expected to capture differences in reliability in Eqs. (4) and (5) (e.g., Titman and Trueman 1986; Beatty 1989; Dietrich et al. 2001; Muller and Reidl 2002).³⁴

Finally, to control for unobservable year and country trends, I include year- and country-fixed effects, and to mitigate concerns related to within-firm correlation of residuals (e.g., Gow et al. 2010), I base reported z-values [i.e., Eq. (1)] and t-statistics [i.e., Eqs. (2)–(5)] on standard errors clustered by firm.

5 Data, sample selection, and descriptive statistics

5.1 Data and sample selection

Market and accounting data items that are not related to investment properties are obtained from the Worldscope, and all of the investment-property-related data are hand-collected. I start by hand-collecting information about a firm’s initial (i.e., at the transition to IFRS) and subsequent investment-property-related accounting choice. Next, for firms adopting the recognition regime, I hand-collect the following investment-property-related items: (1) carrying amount (i.e., fair value), (2) revaluation gains/losses, (3) disposal gains/losses (if reported),³⁵ (4) rental revenues and related expenses, and (5) adjustment to equity at the transition to IFRS (i.e., increase in book value of equity that is associated with the first adoption of the fair value model).

For firms reporting under the disclosure regime, I hand-collect the following investment-property-related items: (1) carrying amount, (2) fair value, (3) cost, (4) current period depreciation charges, (5) accumulated depreciation, (6) subtracted accumulated depreciation due to disposal, (7) disposal gains/losses, (8) impairment losses/reversals, (9) rental revenues and related expenses, and (10) adjustment to equity attributable to change in accounting policy (i.e., the increase in book value of equity that is associated with the switch from a disclosure to a recognition regime).³⁶

To increase the power of my tests, that is, (1) to test the hypotheses on a homogenous sample of firms for which investment properties are likely to constitute a large fraction of their total assets and reported net income and (2) to ensure that the sample firms are affected similarly by accounting differences or choices due to transition to IFRS, I focus on firms that are classified as either “Real Estate Investment Services” or “Real Estate Investment Trusts” (REITs), (hereafter, “real estate firms”). To ensure that the accounting choice is between recognition and disclosure of fair value and related revaluation gains, I focus on firms from countries in which the fair value model (or any other type of revaluation model) was not allowed for investment properties under domestic GAAP.³⁷ To facilitate hand-collection, I focus on real estate firms from four large EU economies—France, Germany, Italy, and Spain—from 2005 or “first IFRS year” to 2010. I define “first IFRS year” as the first financial reporting year for which the firm’s consolidated financial statements according to IFRS are publicly available.³⁸

Table 1 presents the sample selection procedure. I begin with an initial list of 214 real estate firms from France, Germany, Italy, and Spain that appear on Worldscope as “Active” in October 2011. Excluding firms that became public after January 1, 2008 (to ensure at least three observations for each firm) and firms that are subsidiaries (because the reporting choice under IAS 40 is unlikely to be independent of the choice made by the parent firm), I get a list of 192 potential firms. After locating annual financial reports of the firms,³⁹ I further exclude firms with no investment properties, firms for which the fair value of investment properties did not exceed 10 % of total assets in any sample year, and firms with fewer than three publicly available annual reports.⁴⁰ This results in a final sample of 86 publicly traded real estate firms with 532 firm-year observations.

Table 1 Sample selection procedure

5.2 Descriptive statistics and univariate comparisons of firms reporting under the recognition regime to those reporting under the disclosure regime

Table 2 provides a country breakdown of sample firms and firm-years and their initial and subsequent recognition versus disclosure choices. The table reveals that France and Germany have the highest representation in the sample, together comprising 80 % of firms and 79 % of firm-years. In addition, the table shows that, at the transition to IFRS, 40 (46) firms adopted the disclosure (recognition) regime and that, of these 40 firms, 12 subsequently switched to the recognition regime, resulting in 198 (334) firm-years that report under the disclosure (recognition) regime.⁴¹

Table 2 Breakdown of the sample by country and the choice of recognition versus disclosure

Table 3 presents distributional statistics as well as univariate comparisons of firm-years reporting under the recognition regime to those reporting under the disclosure regime, across key variables used in the analyses. As is shown in the table, the differences between mean values of many of the variables are not statistically different from zero (e.g., MVE, IP_ACTIVITY, IPFV, $MTB^{*}$, $TA^{*}$, $NI^{*}$, $REVGAIN^{*}$), implying that the two groups of firms are similar in many respects including their size, the materiality of investment-property-related assets and revenues to their total assets and revenues, and the size and changes in fair values of their investment properties. The table also reveals that rental revenues and fair values of investment properties constitute a material fraction of firms’ total revenues (mean RENTREVRATIO of 64.2 and 62.6 % for firms under the recognition and disclosure regimes) and their total assets (mean IP_ASSETRATIO of 72.7 and 71.3 %, for firms under the recognition and disclosure regimes).⁴²

Table 3 Distributional statistics and univariate analyses

Despite the many similarities, Table 3 also reveals some notable differences between firms reporting under the two regimes. First, consistent with hypotheses H1a, H1b, H2a, and H2b, compared to firms adopting the disclosure regime, those adopting the recognition regime have a significantly higher leverage ratio (mean $LEV^{*}$ of 0.507 versus 0.383, t-statistic = 6.58), significantly more dispersed ownership (mean BLOCKHOLD of 0.367 versus 0.601, t-statistic = −4.74), significantly larger income smoothing parameter (mean INCSMOOTH of 0.151 versus −0.342, t-statistic = 8.31), and a significantly larger sum of fair value adjustments to equity and revaluation gains in the year of transition to IFRS or of switching to the recognition regime (mean TRANSGAIN of 0.241 versus 0.075, t-statistic = 3.02).

In addition, there seem to be some differences in the mean percentages of fair values of certain investment property types. In particular, mean percentage of fair value invested in INDUSTRIAL and RESIDENTIAL (RETAIL and OFFICE) assets is significantly larger (smaller) among firms adopting the recognition regime than among those reporting under the disclosure regime.⁴³ Similarly, although a large fraction of the sample firm-years employ a Big Four auditor and an independent appraiser to determine the fair value of investment properties, some differences exist (mean BigFour of 0.594 versus 0.753, t-statistic = −3.54 and mean ExtAppraiser of 0.917 versus 0.861, t-statistic = 1.74). I control for these differences using the selection model I develop as well as using control variables in market valuation and value relevance tests, that is, Eqs. (2)–(5).

Untabulated Pearson and Spearman correlations between variables used in estimating Eq. (1) reveal that the correlation between variables assumed to capture different incentives for accounting choice within the same category (i.e., contractual or asset pricing) are negligible. For example, the Pearson correlation between $LEV^{*}$ and BLOCKHOLD is −0.045 and between INCSMOOTH and TRANSGAIN is 0.017. Pearson (Spearman) correlations between variables used in market valuation tests, that is, Eqs. (2) and (3), and value-relevance tests, that is, Eqs. (4) and (5), are as expected. For example, IPFV and $REVGAIN^{*}$ have high Pearson correlations with MVE (0.484 and 0.304) and low Pearson correlations with $NI^{*}$ (0.057 and 0.098). RETURN has positive and relatively large Pearson correlation with $REVGAIN^{*}$ (0.21) and negative correlations with beginning-of-year market-to-book ratio and market value of equity.

6 Results

6.1 Factors associated with the choice of recognition versus disclosure⁴⁴

Table 4 presents summary statistics from estimating two versions of Eq. (1), that is, the probit model, one with RENTREVRATIO and another with IP_ASSETRATIO, both denoted by IP_ACTIVITY in Eq. (1).⁴⁵ As is shown in the table, both models correctly predict around 82 % of recognition versus disclosure choices with high Chi-squared statistics (114.1 or 113.3), and in both specifications, the variables of interest have significant coefficient estimates with hypothesized signs.

Table 4 Regression analysis of factors associated with the choice of recognition versus disclosure

With respect to contractual incentives, consistent with H1a, Table 4 shows that the coefficient estimate on $LEV^{*}$ is positive and significant (z-value = 4.240 or 4.280), meaning that firms with higher leverage ratios, that is, closer to violating debt covenants that are tied to SFP values, have higher probability of adopting the recognition regime. In addition, in support of H1b, the coefficient estimate on BLOCKHOLD is negative and significant (z-value = −2.140 or −2.071), implying that firms that have more dispersed ownership, that is, impose lower external monitoring on their management, are more likely to adopt the recognition regime.

With respect to asset pricing incentives, Table 4 reveals that, consistent with H2a, the coefficient estimate on INCSMOOTH is positive and significant (z-value = 4.630 or 4.640), indicating that firms with less smoothed profits in the past, relative to other firms in their country of domicile, have higher probability of choosing the recognition regime. Similarly, consistent with H2b, the coefficient estimate on TRANSGAIN is positive and significant at least at the 10 % level test (z-value = 1.840 or 2.043), suggesting that firms with larger sums of fair value adjustments to equity and revaluation gains have higher probability of adopting the recognition regime.

Estimating the explanatory power and the predictive ability of the four experimental variables—$LEV^{*}$, BLOCKHOLD, INCSMOOTH, TRANSGAIN—relative to that of the full model (which includes, among others, controls for composition of investment property portfolio and firm profitability), Table 4 reveals that they are responsible for about 48 % of the total explanatory power (using McFadden R²), and that, using these variables, it is only possible to correctly predict 74 % of the recognition versus disclosure choices (which represents 90 % of the full model’s predictive ability).

Collectively, evidence presented in Tables 3 and 4 indicates that, despite the many similarities, firms choosing the recognition regime have higher leverage (H1a), more dispersed ownership (H1b), more volatile operating profits in the past (H2a), and more investment-property-related gains at the adoption of the recognition regime (H2b) than those choosing the disclosure regime. This implies that, in the context of investment properties, various dimensions of both contractual and asset pricing incentives help explain the recognition versus disclosure choice. In addition, results presented in Table 4 suggest it is appropriate to use Eq. (1) as a selection model to correct for self-selection concerns in subsequent second-stage market valuation and value-relevance tests, that is, tests of H3 and H4, (Heckman 1979).⁴⁶

6.2 Market valuation tests of recognized versus disclosed amounts

Table 5 provides regression summary statistics from estimating four versions of Eq. (2) and of Eq. (3), which are designed to test H3 using specifications in which MVE is the dependent variable (panel A) and in which RETURN is the dependent variable (panel B). The coefficients of interest in panel A of Table 5 are those on the analogous SFP and I/S amounts, that is, IPFV and $REVGAIN^{*}$, and on their interaction with a firm’s choice of recognition versus disclosure, that is, DISC × IPFV and $DISC \times REVGAIN^{*}$. The coefficients of interest in panel B of Table 5 are those on $REVGAIN^{*}$ and $\Delta REVGAIN^{*}$ and on their interaction with a firm’s choice of recognition versus disclosure, that is, $DISC \times REVGAIN^{*}$ and $DISC \times \Delta REVGAIN^{*}$.

Table 5 Market valuation tests of recognized versus disclosed amounts

As is shown in Table 5, panel A, across all versions of Eq. (2), the coefficient estimates on IPFV and $REVGAIN^{*}$ are significantly positive. (For IPFV, they range from 0.458 to 0.498 with t-statistics between 5.17 and 6.32; for $REVGAIN^{*}$, they range from 0.721 to 1.121 with t-statistics between 1.71 and 2.24.) However, the coefficient estimates on DISC × IPFV and $DISC \times REVGAIN^{*}$ are significantly negative. (For DISC × IPFV, they are −0.097 and −0.093 with t-statistics −1.76 and −2.12; for $DISC \times REVGAIN^{*}$, they are −0.843 and −0.811 with t-statistics −2.37 and −2.46.) Although the sums of coefficient estimates on IPFV and DISC × IPFV and on $REVGAIN^{*}$ and $DISC \times REVGAIN^{*}$ are significantly positive, they are significantly lower than the coefficient estimates on IPFV and $REVGAIN^{*}$.

For example, as column (4) in panel A of Table 5 reveals, the sum of coefficient estimates on IPFV and DISC × IPFV is 0.405 (F-statistic for test of equality to zero is 196 and for test of equality to coefficient estimate on IPFV is 8.67), and the sum of coefficient estimates on $REVGAIN^{*}$ and $DISC \times REVGAIN^{*}$ is 0.257 (F-statistic for test of equality to zero is 3.34 and for test of equality to coefficient estimate on $REVGAIN^{*}$ is 28.90). F-statistics for simultaneous tests of (1) equality of both sums to zero is 111 and of (2) equality of the first sum to coefficient estimate on IPFV and of the second sum to coefficient estimate on $REVGAIN^{*}$ is 23.2. This implies that, although equity investors place simultaneously positive valuation weights on analogous SFP and I/S amounts for firms reporting under both the recognition and disclosure regimes, these valuation weights are significantly lower for firms reporting under the disclosure regime.

A similar picture emerges from results pertaining to various versions of Eq. (3), presented in Table 5, panel B. Specifically, the panel reveals that, while coefficient estimates on $REVGAIN^{*}$ and $\Delta REVGAIN^{*}$ are significantly positive (for $REVGAIN^{*}$, they range from 1.015 to 1.378 with t-statistics between 2.98 and 3.13; for $\Delta REVGAIN^{*}$, they range from 0.036 to 0.069 with t-statistics between 0.74 and 1.93), the coefficient estimates on $DISC \times REVGAIN^{*}$ and $DISC \times \Delta REVGAIN^{*}$ are significantly negative (for $DISC \times REVGAIN^{*}$, they are −0.904 and −0.993 with t-statistics −2.34 and −2.31; for $DISC \times \Delta REVGAIN^{*}$, they are −0.032 and −0.036 with t-statistics −2.55 and −2.06). The sums of coefficient estimates on $REVGAIN^{*}$ and $DISC \times REVGAIN^{*}$ and on $\Delta REVGAIN^{*}$ and $DISC \times \Delta REVGAIN^{*}$, which represent valuation weights placed on analogous SFP and I/S amounts for firms reporting under the disclosure regime, are significantly positive. However, they are significantly lower than the coefficient estimates on $REVGAIN^{*}$ and $\Delta REVGAIN^{*}$.

In particular, as is evident from column (4) in panel B of Table 5, the sum of coefficient estimates on $REVGAIN^{*}$ and $DISC \times REVGAIN^{*}$ is 0.385 (F-statistic for test of equality to zero is 4.11 and for test of equality to coefficient estimate on $REVGAIN^{*}$ is 5.18), and the sum of coefficient estimates on $\Delta REVGAIN^{*}$ and $DISC \times \Delta REVGAIN^{*}$ is 0.033 (F-statistic for test of equality to zero is 3.13 and for test of equality to coefficient estimate on $\Delta REVGAIN^{*}$ is 4.43). F-statistics for simultaneous tests of (1) equality of both sums to zero is 3.64 and of (2) equality of the first sum to coefficient estimate on $REVGAIN^{*}$ and of the second sum to coefficient estimate on $\Delta REVGAIN^{*}$ is 9.85.

Taken together, findings presented in Table 5 indicate that, in both specifications (i.e., using either MVE or RETURN as the dependent variable), correcting for self-selection (i.e., including InvMillsRatio) and controlling for year- and country-fixed effects and for other important variables (e.g., $TA^{*}$, TL, $NI^{*}$) that load in the predicted way, investors place smaller valuation weights on disclosed analogous SFP and I/S amounts than on recognized analogous SFP and I/S amounts in determining a firm value. This evidence leads to a rejection of H3.

6.3 Value-relevance tests of recognized versus disclosed amounts

Table 6 presents regression summary statistics from estimating several versions of Eq. (4) and of Eq. (5), which are designed to test H4 using specifications in which the investment-property-related identifiable future outcomes are represented by ∆NETRENTINC _t.t+τ (panel A) and by ∆CFO _t.t+τ (panel B; in both panels, τ = 1 or τ = 2). Similar to market valuation tests, the coefficients of interest in panels A and B of Table 6 are those on the analogous SFP and I/S amounts, that is, $REVGAIN^{*}$ and IPFV, and on their interaction with a firm’s choice of recognition versus disclosure, that is, $DISC \times REVGAIN^{*}$ and DISC × IPFV.

Table 6 Value-relevance tests of recognized versus disclosed amounts

Panel A of Table 6 reveals that, in both 1- and 2-years-ahead specifications (i.e., ∆NETRENTINC _t.t+1 and ∆NETRENTINC _t.t+2 are the dependent variables), while the coefficient estimates on $REVGAIN^{*}$ and IPFV are significantly positive (for $REVGAIN^{*}$, they range from 0.055 to 0.261 with t-statistics between 1.94 and 2.42; for IPFV, they range from 0.024 to 0.033 with t- statistics between 2.57 and 12.47), the coefficient estimates on $DISC \times REVGAIN^{*}$ and DISC × IPFV are statistically indistinguishable from zero. Accordingly, in both specifications, the sums of coefficient estimates on $REVGAIN^{*}$ and $DISC \times REVGAIN^{*}$ and on IPFV and DISC × IPFV are significantly positive and statistically equal to coefficient estimates on $REVGAIN^{*}$ and IPFV.

For example, as column (2) in panel A of Table 6 reveals, the sum of coefficient estimates on $REVGAIN^{*}$ and $DISC \times REVGAIN^{*}$ is 0.056 (F-statistic for test of equality to zero is 4.56 and for test of equality to coefficient estimate on $REVGAIN^{*}$ is 0.32), and the sum of coefficient estimates on IPFV and DISC × IPFV is 0.03 (F-statistic for test of equality to zero is 6.77 and for test of equality to coefficient estimate on IPFV is 0.12). F-statistics for simultaneous tests of (1) equality of both sums to zero is 8.71 and of (2) equality of the first sum to coefficient estimate on $REVGAIN^{*}$ and of the second sum to coefficient estimate on IPFV is 0.21. This implies that analogous SFP and I/S amounts simultaneously have significantly positive associations with identifiable future outcomes and that these associations are statistically the same among firms reporting under both the recognition and disclosure regimes.

Evaluation of results relating to various versions of Eq. (5), presented in Table 6, panel B, offers similar inferences. In both 1- and 2-years-ahead specifications (i.e., ∆CFO _t.t+1 and ∆CFO _t.t+2 are the dependent variables), while the coefficient estimates on $REVGAIN^{*}$ and IPFV are significantly positive, the coefficient estimates on $DISC \times REVGAIN^{*}$ and DISC × IPFV are statistically indistinguishable from zero. As in panel A, Table 6, in both specifications, the sums of coefficient estimates on $REVGAIN^{*}$ and $DISC \times REVGAIN^{*}$ and on IPFV and DISC × IPFV, which represent value-relevance of analogous SFP and I/S amounts for identifiable future investment-property-related outcomes among firms reporting under the disclosure regime, are significantly positive and statistically equal to coefficient estimates on $REVGAIN^{*}$ and IPFV.

Overall, evidence shown in Table 6 indicates that in both specifications (i.e., using either ∆NETRENTINC _t.t+τ or ∆CFO _t.t+τ as dependent variables, τ = 1 or τ = 2), correcting for self-selection (i.e., including InvMillsRatio), controlling for potential differences in reliability (i.e., using BigFour and ExtAppraiser), controlling for year- and country-fixed effects, and for other important variables (e.g., $TA^{*} ,\;\Delta WC^{*}$), disclosed and recognized analogous SFP and I/S amounts are equally relevant for identifiable investment-property-related future outcomes. This finding leads to a rejection of H4.⁴⁷

7 Findings from additional analyses

7.1 Within-firm-changes research design

To draw inferences regarding market valuation and value-relevance of recognized versus disclosed analogous SFP and I/S amounts, I employ cross-sectional specifications [i.e., Eqs. (2)–(5)] and use invMillsRatio to correct for self-selection concerns (e.g., Heckman 1979) and BigFour and ExtAppraiser to control for potential differences in reliability (e.g., Dietrich et al. 2001; Muller and Reidl 2002).

Despite these controls, one concern is that there might be omitted firm characteristics that are not captured by invMillsRatio as well as some differences in reliability that are not addressed by BigFour and ExtAppraiser. To mitigate this concern, I utilize one of the unique features of the study’s setting and conduct market valuation and value-relevance tests using a within-firm-changes research design.⁴⁸ Specifically, in my sample, 12 of the firms that initially adopt the disclosure regime (71 firm-years) switched at some point to the recognition regime. Using this subsample of firms, it is possible to conduct market valuation and value-relevance tests of analogous SFP and I/S amounts, which belong to the same firms and differ in only whether the firms report under the recognition or the disclosure regime. Such tests help to mitigate some self-selection concerns as well as differences in reliability related to omitted firm characteristics which might be constant over time.⁴⁹

Untabulated results show that all my inferences are the same when I use this alternative research design to test H3 and H4. In particular, despite the fact that analyses are based on the same firms, the analogous SFP and I/S amounts are weighted more heavily when they are reported under the recognition regime and their associations with identifiable future outcomes are statistically equivalent under both reporting methods.

7.2 Alternative model specifications

One potential concern with inferences based on results presented in Tables 5 and 6 is that they are based on Eqs. (2)–(5), which include interaction variables for investment-property-related amounts only (i.e., IPFV, $REVGAIN^{*}$, $\Delta REVGAIN^{*}$) and not for other variables such as $TA^{*}$, TL. To address this concern, I estimate Eqs. (2)–(5), using fully interacted specifications.

In addition, to make sure the results presented in Tables 5 and 6 are not driven by unobserved differences between “Real Estate Investment Services” firms and REITs (defined by Worldscope), between the compositions of firms’ investment property portfolios, or both, I estimate Eqs. (2)–(5), controlling for REIT and IP_TYPE (the variables that I originally include in the selection model I develop). Untabulated results show that all of my inferences are the same when I use these alternative specifications to test H3 and H4.

7.3 Alternative proxies for the proximity of violating debt covenants

In investigating factors associated with the choice of recognition versus disclosure, I follow prior literature and use $LEV^{*}$ as a proxy for how close the firm is to violating debt covenants that are tied to SFP amounts. Because a firm with high leverage may cover its debt well, I follow Nini et al.’s (2012) observation that, besides leverage, two additional commonly used measures in debt covenants are CURR_RATIO (current ratio, defined as current assets divided by current liabilities) and INT_COV (interest coverage ratio, defined as the ratio of interest expense to lagged assets). Untabulated results show that all of my inferences are the same when I use these alternative proxies for the proximity of violating debt covenants.

8 Conclusion

The application of International Accounting Standard 40, Investment Property in the European Union, with mandatory adoption of IFRS, effective January 1, 2005, created a unique setting where the recognized and disclosed investment-property-related amounts share a common measurement base i.e., fair value. I use the setting created by first-time adoption of IAS 40 to (1) explore factors associated with a firm’s choice to recognize versus disclose fair values of investment properties and of related revaluation gains, (2) test whether recognized and disclosed amounts are valued equally by equity investors, and (3) determine whether these amounts exhibit equivalent associations with future financial outcomes, such as future changes in net rental income.

I acknowledge the fact that, under IAS 40, firms self-select whether to adopt the recognition or the disclosure regime and take steps to assure that, in market valuation and value-relevance tests, I compare analogous recognized and disclosed amounts. In particular, I develop a selection model, which I use to control for potential self-selection issues, and I construct analogous SFP and I/S investment-property-related amounts that differ only in whether their components are recognized or disclosed.

I find that contractual and asset-pricing incentives help explain the choice. With respect to differences in market valuation, controlling for self-selection concerns, I find strong evidence that equity investors place smaller valuation weights on disclosed than on recognized analogous SFP and I/S amounts. Regarding the value-relevance tests, controlling for both self-selection concerns and potential differences in reliability, I reveal that recognized and disclosed analogous SFP and I/S amounts have statistically equivalent associations with 1- and 2-years-ahead changes in net rental income and cash flows from operations.

Taken together, the evidence suggests that managers are opportunistic in making the recognition versus disclosure choice and that, even when recognized and disclosed amounts share an equivalent measurement base and are equally relevant for future financial outcomes, investors weight disclosed information less heavily in determining a firm value.

Several limitations of my analyses provide opportunities for future research. First, my sample includes real estate firms from four EU countries. Future research can address questions related to recognition versus disclosure in other industries as well as other EU countries. Second, IAS 40 is one of the many financial reporting changes brought about by the adoption of IFRS. Future research might take steps to evaluate the other changes as well as potential differences in implementation of IFRS across the EU countries and examine the potential effect of valuation and value-relevance of recognized versus disclosed amounts. Finally, inferences concerning factors associated with the choice of recognition versus disclosure are partly based on proxies (e.g., debt covenants). Future research might expand this line of research by exploring the debt covenants themselves.

Appendices

Appendix 1: Variable definitions

REC	An indicator variable equals 1 if, at the transition to IFRS, the firm chooses to recognize fair value of investment properties and related revaluation gains; it equals 0 otherwise Similarly, DISC is an indicator variable that equals 1 if a firm-year observation reports under the disclosure regime; it equals 0 otherwise
MVE	3 months after end-of-fiscal-year market value of equity. Accordingly, Log(MVE) denotes the natural logarithm of 3 months after end-of-fiscal-year market value of equity
IPFV	End-of-year fair value of investment properties
IPBV	End-of-year carrying amount (or book value) of investment properties: for firms choosing the recognition regime—the fair value; for firms choosing the disclosure regime—depreciated cost
TA	End-of-year carrying amount (or book value) of total assets
$TA^{*}$	End-of-year carrying amount (or book value) of total assets other than investment properties For firms choosing the recognition regime: $TA^{*} = TA - IPFV$ For firms choosing the disclosure regime: $TA^{*} = TA - IPBV$
IP_ACTIVITY	Represents the materiality of a firm’s investment-property-related assets/revenues with respect to its total assets/revenues. IP_ACTIVITY is measured by either (1) RENTREVRATIO or (2) IP_ASSETRATIO
RENTREVRATIO	The ratio between average rental and average total revenues during the last 2 years
IP_ASSETRATIO	The ratio between the average fair value of investment properties and average total assets adjusted for fair value of investment properties (for firms choosing the disclosure reporting regime) during the last 2 years For firms choosing the recognition regime: $IP\_ASSETRATIO_{i,t} = \frac{{IPFV_{i,t - 1} + IPFV_{i,1} }}{{TA_{i,t - 1} + TA_{i,t} }}$ For firms choosing the disclosure regime: $IP\_ASSETRATIO_{i,t} = \frac{{IPFV_{i,t - 1} + IPFV_{i,1} }}{{TA_{i,t - 1} - IPBV_{i,t - 1} + IPFV_{i,t - 1} + TA_{i,t} - IPBV_{i,t} + IPFV_{i,t} }}$ where i and t are firm and year subscripts
IP_TYPE	Denotes the composition of a firm’s investment property portfolio, and it represents six variables: (1) INDUSTRIAL, (2) RETAIL, (3) OFFICE, (4) RESIDENTIAL, (5) HOTELS, and (6) OTHER. Each variable, respectively, measures the percentage of total fair value of investment properties invested in each of the following investment property types: (1) warehouses and manufacturing spaces, (2) shopping centers and parking lots, (3) office buildings, (4) apartment complexes, (5) hotels, and (6) other investment property assets
BLOCKHOLD	Total percentage of firm’s shares held by large blockholders (i.e., investors holding 5 % or more of the firm’s common stock), as indicated by Worldscope and hand-collected data
INCSMOOTH	Income smoothing parameter, measured as the ratio between standard deviation of the firm’s operating income and standard deviation of its cash flows from operations during 4 years before transition to IFRS or during the last 4 years, excluding investment-property-related amounts, minus its country of domicile real estate firms’ average, to control for country-specific GAAP or local effects
DISPGAIN	Annual gains or losses resulting from disposal of investment property assets
IPDEP_CHRG	Investment-property-related annual depreciation charges (relevant only for firms adopting the disclosure regime)
IPSUBT_DEP	Annual subtracted investment-property-related accumulated depreciation charges due to sale (relevant for firms reporting under the disclosure regime)
IPCOST	End-of-year total cost of investment properties (disclosed by all firms reporting under the disclosure regime and by most firms reporting under the recognition regime)
IMPAIRLOSS	Investment-property-related annual impairment losses. For firms adopting the recognition regime, it is often included in reported revaluation gains
TAXRATE	Corporate tax rate applicable to capital gains in a given year, a given country, and a particular type of firm (i.e., Real Estate Investment and Services rather than REITs)
$LEV^{*}$	The adjusted leverage ratio, calculated differently depending on the reporting regime For firms choosing the recognition reporting regime: $LEV_{i,t}^{*} = \frac{{DEBT_{i,t - 1} + DEBT_{i,t} }}{{TA_{i,t - 1} + TA_{i,t} }}$ For firms choosing the disclosure reporting regime: $LEV_{i,t}^{*} = \frac{{DEBT_{i,t - 1} + DEBT_{i,t} }}{TA_{i,t - 1} + TA_{i,t} + 2 \times (IPFV_{i,t - 1} - IPBV_{i,t - 1} ) + REVGAIN_{i,t}^{*} /(1 - TAXRATE_{i,t}) \,{{-\,(DISPGAIN_{i,t} - IPDEP\_CHRG_{i,t} - IMPAIRLOSS_{i,t} )}} }$ where DEBT denotes the book value of total debt, and i and t are firm and year subscripts
MTB	The ratio between end-of-year market value of equity and end-of-year book value of equity
$MTB^{*}$	The adjusted market-to-book ratio, calculated differently depending on the reporting regime For firms choosing the recognition reporting regime: $MTB_{i,t}^{*} = \frac{{MVE_{i,t - 1} + MVE_{i,1} }}{{BVE_{i,t - 1} + BVE_{i,t} }}$ For firms choosing the disclosure reporting regime: $MTB_{i,t}^{*} = \frac{{MVE_{i,t - 1} + MVE_{i,t} }}{BVE_{i,t - 1} + BVE_{i,t} + 2 \times (IPFV_{i,t - 1} - IPBV_{i,t - 1} ) \times (1 - TAXRATE_{i,t}) + REVGAIN_{i,t}^{*} - (DISPGAIN_{i,t} - IPDEP\_CHRG_{i,t} - IMPAIRLOSS_{i,t} ) {{\times (1 - TAXRATE_{i,t})}}}$ where i and t are firm and year subscripts
REVGAIN	Annual revaluation gains from the re-measurement of investment properties to fair value (relevant for firms reporting under the recognition regime)
NI	Annual net income before discontinued operations
$NI^{*}$	The portion of reported net income that is not sensitive to the choice of recognition versus disclosure of unrealized fair value gains For firms choosing the recognition reporting regime: $NI^{*} = NI - (REVGAIN + DISPGAIN) \times (1 - TAXRATE)$ For firm choosing the disclosure reporting regime: $NI^{*} = NI - (DISPGAIN - IPDEP\_CHRG - IMPAIRLOSS) \times (1 - TAXRATE)$
$REVGAIN^{*}$	Investment property revaluation gains/losses as either recognized or disclosed For firms choosing the recognition reporting regime: $REVGAIN^{*} = (REVGAIN + DISPGAIN) \times (1 - TAXRATE)$ For firms choosing the disclosure reporting regime: $REVGAIN^{*} = (\Delta IPFV - \Delta IPCOST + DISPGAIN - IPSUBT\_DEP) \times (1 - TAXRATE)$ where Δ is the annual change operator
TRANSGAIN	The sum of net-of-tax difference between fair value and carrying amount of investment property at the beginning of the year of transition to IFRS or adoption of the recognition regime and $REVGAIN^{*}$ For firms choosing the recognition regime: $TRANSGAIN_{i,t} = EQUITYADJ_{i,t - 1} + REVGAIN_{i,t}^{*}$ where $EQUITYADJ_{i,t - 1}$ is net of tax difference between fair value and carrying amount of investment property at the beginning of the year, and i and t are firm and year subscripts For firms choosing the disclosure regime: $TRANSGAIN_{i,t} = (IPFV_{i,t - 1} - IPBV_{i,t - 1} ) \times (1 - TAXRATE_{i,t - 1} ) + REVGAIN_{i,t}^{*}$ where i and t are firm and year subscripts
TL	End-of-year total liabilities of the firm as reported in the statement of financial position
ROA	Total net income scaled by average total assets
REIT	An indicator variable that equals 1 if a firm-year is classified as a real estate investment trust (REIT); it equals 0 otherwise
BigFour	An indicator variable that equals 1 if a firm-year’s financial statements are audited by Big Four audit firm; it equals 0 otherwise
ExtAppraiser	An indicator variable that equals 1 if a firm-year’s end-of-year fair value of investment property (either recognized or disclosed) is determined by an external appraiser; it equals 0 otherwise
InvMillsRatio	The inverse Mills ratio, following Heckman (1979), calculated separately for firms reporting under the recognition regime and for firms under the disclosure regimes, using the fitted values obtained from estimating Eq. (1) (i.e., the probit model)
RETURN	Annual stock return adjusted for dividends and stock splits
ΔNETRENTINC	Annual change in NETRENTINC, net rental income NETRENTINC is annual investment-property-related rental revenue minus directly attributed expenses incurred in obtaining these revenues (as disclosed under IAS 40)
ΔCFO	Annual change in CFO, which denotes annual cash flows from operations
ΔWC	Annual change in WC, working capital WC is the difference between current assets and current liabilities
COUNTRY_FE	Represents country-fixed-effect variables
YEAR_FE	Represents year-fixed-effect variables

Appendix 2: Numerical example

2.1 Overview

This appendix uses a simplified example of a fictional firm and consists of two figures. Figure 1 presents the financial reporting effects of the choice of recognition versus disclosure under IAS 40 on amounts reported in the statement of financial position (SFP) and the income statement (I/S). Figure 2 shows how, despite the apparent differences in reported SFP and I/S amounts, the method I develop and discuss in Sect. 4.2 allows one to construct analogous SFP-related amounts (i.e., $TA^{*}$ and IPFV) as well as I/S-related amounts (i.e., $NI^{*}$ and $REVGAIN^{*}$) whose information content is independent of the firm’s choice of recognition versus disclosure. All variables mentioned in this appendix are defined in “Appendix 1”.⁵⁰

2.2 Example details

Consider a firm that was created at the end of 2004 and was involved in the following set of transactions during the period from December 31, 2004 to December 31, 2007:

December 31, 2004:

Received $7000 of initial investment capital (initial issuance of common stock shares) to begin the business.

Received $5000 cash in exchange for a bank loan. Bank loan terms required 5 % interest (compounded annually) and interest-only payments to be made on the last day of each year. Principal is due on December 31, 2014.

Purchased a building for $10,000 in cash. Ten percent of the building’s space was designated for own use (i.e., management offices) and the rest, 90 %, was held for rental. Land accounted for $3000 of the purchase price (i.e., 30 % of the building value). Building’s space is estimated to have useful life of 50 years (with no residual value). The building was depreciated using the straight-line method.

December 31, 2006:

Sold 50 % of building’s space held for rental for $5400 in cash.

December 31, 2007:

Sold 50 % of the remainder of building’s space held for rental for $2250 in cash.

The table below presents building’s fair values (100 % of the building) at the end of 2004–2007 and total rental revenues and related expenses during these years (in $):

	FV of building	Rental revenue	Maintenance expenses
2004	10,000
2005	11,000	900	100
2006	12,000	950	110
2007	10,500	450	50

Starting with January 1, 2005, the firm reports according to IFRS, which distinguishes between investment properties and property, plant, and equipment (PP&E) and accounts for its PP&E using the cost model.

For simplicity, assume there are no taxes and that the firm does not incur any costs in selling the assets (i.e., costs to sell are 0). Also, suppose that all of the firm’s cash inflows and outflows (i.e., rental revenues and maintenance expenses) occur at the end of the year.

Fig. 1

Financial reporting effects of recognition versus disclosure

Fig. 2

Obtaining analogous SFP and I/S amounts