Do fair value adjustments for non-financial assets impact dividend policy choices of firms?

(1)

Do fair value adjustments for non-financial

assets impact dividend policy choices of firms?

Christiaan Veerman 10060332 Master Thesis Accountancy & Control

University of Amsterdam Research field: Financial Accounting

(2)

2 Abstract

From prior literature could be found that regulators are concerned about the distribution of transitory revaluations, although this isn’t predicted by the framework of Lintner (1956). This paper examines the association between unrealised fair value revaluations of investment properties and the dividend policies of firms. The setting is the European Union for the years 2005-2012. From the sample of 638 firm-year observations, 423 observations (66.3%) have chosen to revaluate investment property and 215 firm-year observations (33.7%) had no revaluations. The findings indicate that the unrealised fair value revaluations are not persistent, although some evidence was found that positive fair value revaluations are persistent. The results from the main hypothesis indicate that the unrealised fair value revaluations are not distributed in dividends. So the concerns of regulators aren’t supported. No significant

difference is found between the countries where distribution is allowed and those where it isn’t. For the corporate governance structure this paper found significant evidence that firms with high board average attendance are less likely to cut their dividends. No significant results are found for borrowing capacity and managerial optimism. The main limitation of this paper is the relative small sample size due to data availability.

(3)

3 1. Table of content

1. Table of content 2. Introduction

3. Regulatory framework

4. Theory and hypotheses development

4.1 Earnings persistence and dividend policies 4.2 Hypotheses development

5. Research methodology

5.1 Data and sample description 5.2 Models for main hypotheses 5.3 Models for firm characteristics 6. Emperical findings

6.1 Descriptive statistics 6.2 Multivariate analysis

6.2.1 The persistence of unrealised gains/losses from investment properties 6.2.2 The effect of unrealised gains/losses from investment properties on

dividend policies

6.2.3 The impact of specific firm characteristics on dividend policies 6.3 Robustness tests

7. Discussion and conclusion 8. References

(4)

4 2. Introduction

There is an ongoing discussion between regulators, practitioners and academics on the increasing use of fair values in accounting under the International Financial Reporting Standards (IFRS) framework (Christensen and Nikolaev, 2013). The discussion is about the value relevance and reliability of financial information. Historical costs is seen as a more reliable measure in financial reporting, while the IASB is promoting fair value accounting in their IFRS framework since it‘s expected to deliver more relevant and timely information (Barth, 2007; Barth, et.al. 2001; FASB, 2000; Hitz, 2007). But fair value accounting implies an increase in managerial discretion and the possibility that unrealised, transitory gains are

recognized in net income (Penman, 2007). The inclusion of unrealised gains potentially will introduce noise in the decision making process.

Regulators are concerned about the negative effect that fair value adjustments could have on dividend policies of firms, which could lead firms into financial distress (Enria et.al., 2004). This paper examines the effect of fair value adjustments on dividend policies of firms in European countries subject to the IFRS accounting standards. Prior literature results in contrary theories and empirical findings about this relationship. Lintner’s framework from 1956 predicts that only persistent earnings should have an impact on firms distribution of earnings (Lintner, 1956). Earnings or revenue is defined in IAS 18 as the gross inflow of economic benefits (cash, receivables, other assets), occurring in the course of ordinary activities (such as sales of goods, sales of services, interest, royalties, and dividends) and such inflow increases the equity, not including deposits by shareholders (Deloitte, 2014). Earnings and fair value adjustments consist of both persistent and transitory components. Persistent components of earnings will return in future periods while transitory components are one time items that don’t persist on the long run in earnings. So these adjustment will not influence the dividend policy of a firm. This framework describes the smooth relationship firms seek between earnings and dividends. So dividends are very well related with current earnings. Only when a firm has negative earnings this relationship is not that strong since the firm is inclined to decrease dividends instead of omitting them completely (DeAngelo and DeAngelo, 1990). This is the case because omission has negative signalling effects. According to the Lintner framework (1956) fair value

adjustments could influence dividend policies only when they are persistent. Thus, when managers and investors process the persistence of fair value adjustments effectively, more relevant and timely decisions could be made on dividend policies. But in practice there´re some issues with the framework of Lintner (1956). The assumption of Lintner that firm management

(5)

5

and the investors can effectively assess the persistence of positive fair value adjustments

doesn’t always hold in practice. Jensen (1993) and DeAngelo et. al. (1996) show that the ability of managers to interpret the persistence of current earnings on future earnings is lacking. They show that managers often are too optimistic in estimating the persistence of earnings

components. So they found that managers often misinterpret the persistence and this could cause noise in dividend policies of firms. In addition, Sloan (1996) shows that investors don’t process the persistence of earnings effectively, so they don’t effectively identify the difference between persistent and transitory unrealized gains. This could lead to problems when dealing with transitory earnings components. Those gains could be recognized as persistent fair value adjustments and this could affect dividend policy choices. Firms could payout those transitory unrealized fair value gains by increasing dividend payouts, without knowing those gains aren’t persistent. Enria, et. al. (2004) shows that regulators are concerned that firms will increase dividend payouts because transitory upward adjustments in fair values are identified as persistent. Because of the increase in dividend payouts, they have concerns about the a firm's ability to handle its debt. Increased leverage makes firms more vulnerable to economic shocks (Bernanke, et. al. 1996). When transitory fair values are volatile and interpreted as persistent this could lead to noise in decision making in dividend policies. This could make dividend policies more volatile and could cause problems in real economy (Plantin et. al. 2008, Laux and Leuz, 2009) like problems with debt covenants.

Prior literature that examines the relationship between fair value adjustments and dividend policies resulted in contradicting findings. Goncharov and Van Triest (2011)

examined the effect of positive fair value adjustments on the dividend policies of Russian firms, but they mainly focused on adjustments of financial securities. Lintner’s (1956) framework predicted no effect and regulators were concerned about increased dividend payouts (Enria, 2004), where Goncharov and Van Triest found that firms paid less dividend when they were dealing with upward fair value adjustments. One of their possible explanations was that managers can use upward fair value adjustments as an excuse to cut dividends. Firms cut dividends and justify this with the excuse that those upward fair value adjustments are

transitory. According to Goncharov and Van Triest (2011) this would be the case in firms with weak corporate governance structures. This suggested explanation could be drawn from prior literature as well, as Gugler and Yurtoglu (2003) found that firms with a weak corporate governance structure are more likely to cut dividends. This could be explained by the agency theory and the rent extraction hypothesis. When the corporate governance structure of a firm is weaker the large shareholders are more likely to extract rents from the firm in their self-interest,

(6)

6

because the small shareholders are not that well protected. The second alternative explanation Goncharov and Van Triest (2011) gave is that upward fair value adjustments may be correlated with managerial optimism. As dividend decisions are made after considering investment needs (Brav et. al. 2005), managers may pay relatively lower dividends to finance investments,

leading to the observed negative relationship between dividends and unrealised fair value gains. The IASB promotes fair value accounting in the IFRS to ensure that the information in the financial statements are more timely and more transparent and increasing the decision relevance of accounting information (Barth and Clinch, 1998; Barth et al. 2001; Hitz 2007). But fair value accounting has downsides because fair values are more volatile and not always persistent and the persistence is not always effectively assessed by managers (Jensen, 1993; DeAngelo et. al. 1996) and investors (Sloan, 1996). So it can cause noise in dividend policy choices. This paper examines the following research question: Do fair value adjustments for non-financial assets impact the dividend policy choices of firms? Fair value adjustments can be split up in financial securities (IAS 39) and non-financial assets, like Property, Plant and Equipment (PPE), investment properties (IAS 40) and intangible assets. This paper examines the relationship between fair value adjustments from investment properties and dividend policies. This is interesting because under IFRS it’s allowed to allocate unrealised fair value gains and losses from investment properties in net income (Christensen and Nikolaev, 2013). In addition to the relation between fair value adjustments from investment properties and dividend policies, this relation is further examined under different firm characteristics. These

characteristics are whether or not the law is allowing distribution of fair value adjustments for investment properties, the strength of the corporate governance structure, borrowing capacity and managerial optimism.

The contributions of this paper address scientific, practical and societal aspects. This paper has a scientific contribution by addressing a gap in prior literature. Lintner’s (1956) framework predicts no effect from upward fair value adjustments on dividend policies and regulators are concerned about distributing transitory fair value adjustments. But Goncharov and Van Triest (2011) found contradicting results given a decline in distributed dividends and give two possible explanations for those findings. This paper examines the alternative

explanations of the Goncharov and Van Triest (2011) paper. The impact of managerial optimism and corporate governance structure and borrowing capacity as an additional firm characteristic on the relation between upward fair value adjustments and dividend policy choices are examined. Other circumstances are examined as well, like the applicable law and the different effects of positive and negative revaluations. In addition to addressing this gap in

(7)

7

literature, this paper also contributes by examining the research of Goncharov and Van Triest (2011) for investment properties and in a broader setting, namely in the European Union and in a longer period from 2005-2012. This paper is of importance to practitioners, because the findings could be useful for managers of the firm, investors and regulators. This paper could be useful for the managers of firms and investors, they could enhance their understanding on the persistence between different components of earnings and fair value adjustments. Moreover, they could improve decision making on dividend policies choices. It is good for regulators to know if their concerns about distributing upward fair value adjustments to shareholders are valid. Regulators can assess how different firm characteristics, like if distribution of FV adjustments is allowed, corporate governance structure, borrowing capacity and managerial optimism influence firms dividend policy choices. It also contributes to practice by examining it in the European setting, this could provide practical implications on the fair value accounting discussion in IFRS and to enhance the practical knowledge on the effect of fair value

adjustments on dividend policy choices of firms. This is especially interesting because the member states of the EU have different permitted and required practices for profit distribution of fair value adjustments. This paper provides a societal contribution by addressing the

concerns that regulators have about distributing transitory components of upward fair value adjustments in dividends (Enria et. al. 2004). Distributing those transitory upward fair value adjustments could lead to higher leverage levels and this could make more vulnerable to financial distress. Financial distress impose social costs because it destroys wealth of

shareholders, because the firm can’t focus on the long-term value creation and has to seek for short-term sources of cash (Squire, 2010). So this paper tries to create a better understanding of transitory fair value gains among management, shareholders and regulators and tries to mitigate the societal costs of financial distress.

From the sample of 638 firm-year observations, 423 observations (66.3%) have chosen to revaluate investment property and 215 firm-year observations (33.7%) had no revaluations. The sample is tested on self-selection bias with an Inverse Mills ratio, but no significant coefficients are found. The findings indicate that the unrealised fair value revaluations are not persistent, although some evidence was found that positive fair value revaluations are persistent. The results from the main hypothesis indicate that the unrealised fair value revaluations are not distributed in dividends. So the concerns of regulators aren’t supported. No significant

difference is found between the countries where distribution is allowed and those where it isn’t. For the corporate governance structure this paper found significant evidence that firms with high board average attendance are less likely to cut their dividends. No significant results are

(8)

8

found for borrowing capacity and managerial optimism.

In the following section the regulations regarding the distribution of fair value

adjustments is discussed for the EU member states. In the third section the relation between the persistence of fair value adjustments and dividend policy choices is elaborated and the

hypotheses are developed based on prior literature. Thereafter the models, sample selection and descriptive statistics are presented before the results are discussed. There will be some

sensitivity analyses and robustness test. In the last section the conclusion of this paper is given.

3.1 Regulatory framework

European Union member states have different laws and regulations regarding the distribution of fair value adjustments (KPMG, 2008). Some member states allow the distribution of the income reported under IFRS and in some member states modifications are required and some member states only allow the distribution under local GAAP.

The first possibility is that a firm in a member state can distribute the income reported under IFRS in dividends and thus allowing that the distribution could be affected by transitory unrealised gains and losses from revaluations of investment property. For those firms IAS 40 is the applicable law for distribution fair value adjustments. Investment Property is recognized in the IFRS framework as property held to earn rentals or for capital appreciation or both and is initially recognized at cost (KPMG, 2006). Firms can choose to use the fair value model or a cost model. When using the fair value model changes in fair values should be recognized in profit or loss, this could result in unrealized gains and losses in distributable income as well. The member states that allow distribution of unrealised gains from revaluations of investment properties under IAS 40 are Bulgaria, Cyprus, Czech Republic, Estonia, Finland, Greece, Latvia, Lithuania, Portugal, Slovakia, Slovenia and Spain (from 2008). In Greece it’s

recommended not to distribute unrealised fair value gains, but these recommendations of the Hellenic Commission of Auditing and Accounting Regulation (HCAAR) are not binding (KPMG, 2008).

The other member states that don’t allow the distribution of the income reported under IFRS and firms need to modify the income reported under IFRS for distribution purposes. Those member states don’t allow the distribution of unrealised gains from revaluations of investment property as reported in the IFRS income and require modifications. In Denmark, the Danish Companies Act (DCA) and the Danish Financial Statement Act (DFSA) require Danish

(9)

9

firms to modify the income reported in accordance with IFRS. Modifications are required to filter out unrealised fair value gains from the revaluation of investment properties. Those revaluations are not distributable until the underlying asset is realised (KPMG, 2008). Under Italian GAAP investment property must be stated at cost and depreciated over their useful life, depreciation is not mandatory and revaluating at fair value is not permitted (KPMG, 2010). And when there’re some fair value adjustments or First-Time Adoption (FTA) changes those should be recognized directly in equity in a revaluation reserve (KPMG, 2008). In Malta, under Maltese GAAP some adjustments to the IFRS income have to be made to filter out the non-realised components due to positive fair value adjustments. Those unnon-realised gains are transferred to a non-distributable revaluation reserve (KPMG, 2008). In the Netherlands, investment properties are initially recognized at cost. After initially recognition the investment property can be measured with the cost model or the fair value model. Unlike under IFRS the unrealised gains and losses are under Dutch GAAP, Guidelines on Annual Reporting (GAR) section 213, recognised in a revaluation reserve and not in the profit or loss statement (KPMG, 2006). According to the KPMG (2008) study, in Poland investment properties can be valued at fair value which is dealt with in the Polish Accounting Act (PAA; Article 28.1; 1a). Those unrealised gains and losses should be recognized in a revaluation reserve in equity and are not distributable in dividend (Polish Accounting Act; Article 35.4). Unrealised gains and losses due to revaluations in investment properties should be recognised in a investment revaluation reserve in the United Kingdom. When the gain or loss is of a permanent nature the revaluation can be recognised in profit or loss (EY, 2011). Gains could be recognized in profit or loss when they are realised or readily convertible into cash (KPMG, 2008). Investment properties under Irish GAAP are not depreciated and are included on the balance sheet at their open market value (Grant Thornton, 2013). Gains from revaluations can be distributed only when they are realised. Under Irish GAAP they follow the ICEAW guidelines that are applicable as well in the United Kingdom, this guideline determines realised profit as realisation of the underlying asset or assets that are readily convertible into cash. Unrealised gains on investment properties are reported in a revaluation reserve (KPMG, 2008).

In the other member states the distribution of dividend is required to be based on the net income reported under local GAAP requirements. These countries are Austria, Belgium, France, Germany, Hungary, Luxembourg, Spain, Romania and Sweden (KPMG, 2008). Under

Austrian GAAP in accordance with Austrian Commercial Code, land and buildings are

recognized according to IAS 40 at the cost model minus scheduled amortisation, this schedule is based on actual duration of use (UNIQA, 2009). In Belgium investment property is

(10)

10

accounted for in the same manner as property, plant and equipment. The cost model is commonly used in Belgium for both PPE and investment properties, those properties are carried at cost less accumulated depreciation and only permanent impairment losses are recognized. Revaluations are also permitted for PPE and investment properties. Revaluation gains from investment properties are transferred directly in a revaluation reserve in equity (PWC, 2010b). Under France GAAP (Commercial Code; Art. D 7-1/2/3; L 123-18; D 7-5; D 12), investment properties are accounted for as property, plant and equipment and are only revalued at fair value when all long-term financial instruments and PPE are revalued. Those fair value adjustments are directly transferred into equity, revaluations are not to be made on a regularly basis (KPMG, 2008). In Germany, investment properties are recognized and

accounted for the same as property, plant and equipment using the Cost Model under § 253 I HGB (KPMG, 2008) and German GAAP (revised) (PWC, 2010a) and the Fair Value Model is not permitted. In Hungary investment property is not distinguished from owner-occupied property. The cost model is required on the valuation of these properties. But fair value adjustments of properties can be recognized in an valuation adjustment asset in the balance sheet and an corresponding increase directly in equity and the values are reassessed on a yearly basis. Unrealised gains can’t be placed in the profit and loss statement and thus can’t be

distributed in dividends (KPMG, 2008). In Luxembourg, the measurement basis for investment property is the same as for property, plant and equipment under Lux GAAP, so revaluation is prohibited. Revaluations are thus not possible and revaluation gains of investment properties can’t be distributed in dividend under Lux GAAP (Deloitte, 2007). After December 10th 2010, this changed into a choice between Lux GAAP and the possibility to recognize the revaluations gains and losses of investment properties in profit or loss or in a revaluation reserve (PWC, 2013). According to Biebel and Santella (2008) under Spanish GAAP it wasn’t possible for firms to distribute the IFRS income in dividend up till after December 31, 2007. But after that date they could distribute income reported under the IFRS framework. So unrealised fair value adjustments from investment property revaluations can be distributed in dividend since January 1, 2008. In Romania there is no distinction being made between the assets used by the

enterprise and investment properties. Fair value changes of those properties are permitted or required to have an impact on equity without having any impact on the profit and loss

statement (KPMG, 2008). Under Swedish GAAP; RR 32, firms aren’t allowed to use fair value model for the revaluation of investment properties and are required to use the cost model. Investment properties should be recognized as property, plant and equipment (PPE) and is only allowed to be revalued when the fair value exceeds the carrying value and is considered to be

(11)

11

realised (significant, reliable and permanent of nature), (KPMG, 2005; 2008).

Table 1 Distribution of unrealised gains from investment properties in European Union member states

European Union member state

Distribution of unrealized gains allowed?

Article in law Distribution allowed (D=1), otherwise (D=0)

Bulgaria Yes IFRS (IAS 40) 1

Cyprus Yes IFRS (IAS 40) 1

Czech Republic Yes IFRS (IAS 40) 1

Estonia Yes IFRS (IAS 40) 1

Finland Yes IFRS (IAS 40) 1

Latvia Yes IFRS (IAS 40) 1

Lithuania Yes IFRS (IAS 40) 1

Portugal Yes IFRS (IAS 40) 1

Slovakia Yes IFRS (IAS 40) 1

Slovenia Yes IFRS (IAS 40) 1

Greece Yes, but recommended not to distribute unrealized gains

Greek GAAP; Recommendation HCAAR

1 Spain (2005-2007) No, use cost model Spanish GAAP 0 Spain (2008-2012) Yes IFRS (IAS 40) 1 Denmark No, use a Revaluation Reserve DCA and the DFS 0 Italy No, use cost model Italian GAAP 0 Malta No, use a Revaluation Reserve Maltese GAAP 0 The Netherlands No, use a Revaluation Reserve Dutch Civil Code; GAR 213 0 Poland No, use a Revaluation Reserve PAA; Article 35.4 0 United Kingdom No, use a Revaluation Reserve UK GAAP; SSAP 19 0 Ireland No, use a Revaluation Reserve Irish GAAP 0 Austria No, use cost model Austrian GAAP; UGB 0 Belgium No, cost model or Rev. Res. Belgian GAAP 0 France No, use a Revaluation Reserve French GAAP; Commercial Code 0 Germany No, use cost model German GAAP; § 253 I HGB 0 Hungary No, cost model or Rev. Res. Hungarian GAAP; MKVK 0 Luxembourg

(2002-2009)

No, use cost model Lux GAAP 0 Luxembourg

(2010-2012)

Yes IFRS (IAS 40) 1

Romania No, use a Revaluation Reserve Romanian GAAP; OMFP nr. 3055/2009

0

(12)

12 4 Theory and hypotheses development

4.1 Earnings persistence and dividend policies

In this section the main theoretical relation between earnings persistence and dividend policy choices is described. Lintner’s (1956) framework predicts that distributed dividends of firms have a smooth relationship with earnings. Thus, firms determine their distributed dividends on current earnings and dividend payments of the previous year and create a smooth relationship between those variables. Lintner’s (1956) framework also predicts that firms only will

distribute persistent/realised components of earnings.

The research of Fama and Babiak (1968) support the framework of Lintner (1956). And the paper of DeAngelo and DeAngelo (1990) also found evidence that when earnings are negative, firms are creating a smooth relationship between earnings and dividends. Only they are reluctant to omit dividends, but they just cut divends. They’re reluctant to omit, because of the negative signalling effects, because investors could think they struggle with their liquidity. Earnings consist of different components, namely cash and accrual components. Those earnings components could differ in persistence, persistent earnings components are earnings that will return in future periods while transitory earnings components are unrealised and will not return in future periods (Sloan, 1996). Fair value adjustments are mostly not realised

immediatly before the end of the period, those adjustments could reverse in later periods if they aren’t persistent. So it’s important to assess the persistence of fair value adjustments in the right way. Enria et. al. (2004) states that regulators are concerned that firms don’t assess the

persistence of those adjustments in the right way and distribute transitory components in dividends, which could lead firms into financial distress. This could make dividend policies volatile and could lead to problems in real economy (Plantin et. al. 2008, Laux and Leuz, 2009) like problems with debt covenants.

There’re some practical issues regarding the framework of Lintner (1956). From the research of DeAngelo et. al. (1996) and Jensen (1993) could be concluded that investors can’t effectively assess the persistence of the different components of earnings. Sloan (1996) also finds that investors can’t assess the persistence effectively. So when wrongly assessed the transitory fair value adjustments could be distributed in dividends. This led to the concerns of regulators (Enria et. al., 2004). In addition, from the framework of Lintner (1956) the following hypothesis could be extracted, higher dividends are paid by firms that have high and more

(13)

13

stable earnings. This hypothesis was supported by other research papers of Jagannathan et. al. (2000) and Skinner (2004).

So there’re several research papers that found support for the smooth relationship between earnings and dividend payout ratio, as suggested by Lintner (1956). But there’re some critiques that are concerned about the ability of managers and investors to assess the

persistence of earnings components in an effective way (Enria et. al., 2004; DeAngelo et. al., 1996; Jensen, 1993; Sloan, 1996).

This paper examines the effect of fair value adjustments of investment properties on the changes in dividends. Under IFRS investment property revaluations could be distributed even when they are unrealised, it is thus interesting to examine the persistence of fair value gains, to assess if this could cause financial distress. IAS 40 indentifies investment property as property held to earn rentals or for capital appreciation or both (KPMG, 2006). Contractual negotiated rentals for a fixed period could be simply calculated and discounted for future periods. Wirtz (2013) found that unrealised gains from investment properties align the book value of equity with the market value of equity and concluded that this could be an indication of faithfull representation of unrealised gains and an accurate indication for future cash flows. This is preliminary evidence that unrealised gains from fair value adjustments of investment property are persistent of nature. This paper also examined the persistence of those fair value

adjustments before the main research question could be assessed, this led to the following hypothesis 0.

Hypothesis 0: The unrealised gains and losses from fair value adjustments of investment property are persistent

4.2 Hypotheses development

The IASB is promoting the importance of fair value accounting in accounting, and

implemented fair value accounting in the IFRS accounting framework (Barth and Clinch, 1998; Barth et al. 2001; Hitz 2007; FASB, 2000). This implementation has to ensure that financial statements are value relevant for investors and other users of these financial statements. The information becomes more value relevant, because the valuations are better reflecting the current value of balance sheet items and are more transparant. This makes the decision-making process of investors more accurate. The implementation of fair values in IFRS caused that

(14)

14

financial statements are less reliant on the historical cost accounting figures. Fair value accounting has not only positive effects, Laux and Leuz (2009) state that the Financial

Accounting Standards Board (FASB) and the International Accounting Standards Board (IASB) are dealing with a tradeoff between relevance and reliability of financial statements. With the use of fair values the relevance of the financial information for investors increased. But the measurement of those fair values could cause some noise in financial statements. Because market prices could include noise and lead to a reduction in reliability (Shleifer et. al., 1992; 1997). Assets are priced at the market value of identical assets. But there’re not always identical assets or there’s no deep market for those assets.

The main relation that is examined in this paper is the relation between upward fair value adjustments of investment properties and the dividend policy choices. Lintner’s (1956) framework suggest that firms seek to create a smooth relationship between earnings and dividend payouts, this is supported by the papers of Fama and Babiak (1968) and DeAngelo and DeAngelo (1990), and that only persistent earnings components are distributed. Regulators are concerned that those dividend payments include transitory components (Enria et. al., 2004). In prior literature this effect is examined by Goncharov and Van Triest (2011). They found contradicting results from what they expected, they found that when fair value adjustments of financial securities were present, firms were more likely to cut their dividend. From Lintner’s (1956) framework and from regulators concerns (Enria, 2004) could be expected no effect from upward fair value adjustments on the dividend payouts or increases in dividend payouts

according to regulators concerns. Regulators expecting distribution of transitory components of earnings. This feared distribution of transitory components could be explained by the lack of ability to effectively process the persistence of earnings components. Prior studies from

DeAngelo et. al. (1996) and Jensen (1993) show that managers can’t interpret the persistence of earnings and are often to optimistic in assessing the persistent of earnings components. In addition Sloan (1996) shows that investors can fully process the difference between persistent and transitory unrealized gains. For this reason regulators are concerned that transitory

unrealized gains are distributed in dividends and firms could become financial distressed, because those unrealized fair value gains aren’t persistent.

So this paper examines if these contradicting findings hold in a broader setting, namely

Europe, in another and longer period of time, from 2005-2012. So based on the framework from Lintner (1956) could be expected that unrealised income from fair value adjustments do not impact the dividend policy choices, this led to the development of the following hypothesis.

(15)

15 Hypothesis 1: Unrealised income from fair value adjustments from investment property has no

impact on dividend policy

When the persistence of unrealised gains is examined, it’s important to examine the impact of differences in regulatory framework amongst the European Union. As stated before in the regulatory framework section the applicable law, IFRS or local GAAP, in the different member states of the European Union differs on the topic of distribution of unrealised fair value gains in dividend.

In a couple of member states the distribution of unrealised fair value gains reported in profit and loss under IFRS is allowed. These countries are Bulgaria, Cyprus, Czech Republic, Estonia, Finland, Greece, Latvia, Lithuania, Portugal, Slovakia, Slovenia and Spain (from 2007). (KPMG, 2008).

In the other member states the IFRS income should be modified or the income reported under local GAAP should be used for distribution purposes. In the following countries the distribution of unrealised fair value gains isn’t allowed because modifications should be made to the reported IFRS income. It’s permitted to use the fair value model and required to use the cost model or the unrealised fair value adjustments should be transferred into equity in a revaluation reserve. In some member states the profit distribution isn’t allowed to be based on the IFRS income, but firms are required to use the income reported under local GAAP, where unrealised fair value gains are non distributable. These countries are Austria, Belgium, France, Germany, Hungary, Luxembourg, Spain, Romania and Sweden (KPMG, 2008). In Spain on January 1st 2008 and in Luxembourg on December 11, 2010, the requirements changed to the IFRS regulation where unrealised fair value adjustments are permitted to be distributed in dividend.

These differences in legal framework in the different member states are expected to lead to differences in distribution of unrealised fair value adjustments in dividends. Firms that are allowed to distribute their IFRS income and thus unrealised fair value adjustments in dividend, the unrealised fair value adjustments are expected to have a significant higher impact on the dividend policies. This could lead to increasing leverage which makes firms more vulnerable for economic shocks (Bernanke, et. al. 1996). Those transitory fair value adjustments of investment properties are volatile and could cause noise into decision making of dividend policy choices. This could make dividend policies more volatile and could cause problems in real economy (Plantin et. al. 2008, Laux and Leuz, 2009) like problems with debt covenants. This increased volatility and vulnerability for economic shocks could cause a increase in

(16)

16

financial distressed firms in countries that comply with IFRS distribution regulations. This led to the following hypothesis.

Hypothesis 2: When distribution is allowed, unrealised fair value adjustments from investment properties have a significant impact on dividend policies of firms.

After examining the persistence of fair value adjustments and the examination of the main hypothesis, the results are examined for different firm characteristics. Those characteristics are expected to have an impact on the relationship between unrealised fair value adjustments of investment properties and the dividend policy choices firms make. From the research of Goncharov and Van Triest (2011), this paper propose two possible firm characteristics that are considered to influence the dividend policies of firms. From Lintner’s (1956) framework can be expected that transitory earnings have no effect on dividend policy choices, where regulators are concerned about the distribution of transitory unrealised fair value adjustments. But on the contrary, Goncharov and Van Triest (2011) found that firms in Russia were cutting dividends.

Their first alternative firm characteristic that could have led to this result is the presence of a weak corporate governance structure at the firm. Goncharov and Van Triest (2011) argue that investors and managers can’t effectively assess the persistence of upward fair value adjustments, which is in line with prior literature (DeAngelo et. al., 1996; Jensen, 1993; Sloan, 1996) and use this as an excuse to cut dividends. Firm management could argue that those fair value adjustments are transitory of nature and are thus not distributable. The effect of corporate governance structures on dividend policies is previous examined in Germany by Gugler and Yurtoglu (2003) and Gugler (2003). They found a direct positive effect of corporate

governance structures on dividend policy choices, when the corporate governance structure is stronger the dividends are higher. This is also found by Farinha (2003). Gugler and Yurtoglu (2003) conclude that strong corporate governance structures are very important in the

protection of minority shareholders in case of dividend distribution. The alternative explanation from Goncharov and Van Triest (2011) could be found in prior literature as well, as Gugler and Yurtoglu (2003) found that firms with a weak corporate governance structure are more likely to cut dividends. This could be explained by the agency theory and the rent extraction hypothesis. With weak corporate governance structure, large shareholders are more likely to extract rents from the firm in their self-interest. With a weak structure, the small shareholders are not that well protected against rent extraction. The strength of the corporate governance structure of a firm could be assessed and measured by the following variables: independent directors,

(17)

17

independence of committees, board size, CEO duality/CEO role, board meetings, auditor reputation, competence of audit committee members and audit committee meetings (Khanchel, 2007). Those variables are measuring the independence, integrity, competence of the executive and supervisory boards and the quantity of audit committee and board meetings. In this paper the strength of the corporate governance structure is examined as alternative explanation in impacting the dividend policy choices when unrealised fair value adjustments are present. Corporate governance is assessed by board diversity, CEO duality, board size and a corporate governance score. So this paper examines if firms use fair value adjustments as an excuse to cut dividends to for example extract the money from the firm for their self-interest when a firm has weaker corporate governance in place.

Hypothesis 3: The strength of a firms corporate governance structure has a positive effect on the distribution of fair value adjustments from investment properties

Goncharov and Van Triest (2011) alter another explanation that could possibly explain their contradicting findings. They state that people in general (De Bondt & Thaler, 1995) and managers in specific (Jensen, 1993) are often to optimistic in making investment decisions.

Overoptimism lead to higher investment needs, which likely lead to lower dividends (Ben-David et. al., 2007). This could further explained by the research of Brav et. al. (2005), who state that dividend payout decisions are made after considering investment needs, leading to the observed negative relationship between dividends and unrealised fair value gains.

On the contrary there’s a theory that state that this overinvestment isn’t in the best interest of the firm and how it could be prevented. The free cash flow hypothesis predicts that firms with high cash flows could better distribute a portion of this free cash, because the chance to invest it in negative net present value (npv) projects declines (Gugler and Yurtoglu, 2003; Jensen, 1986). This hypothesis is especially useful for firms that deal with overinvestment. On the otherhand Goncharov and Van Triest (2011) state that could be expected that optimistic managers are to optimistic in assessing the persistence of fair value adjustments and thus distributing transitory fair value adjustments.

To measure managerial optimism, this paper uses a proxy found in prior research of Campbell et. al. (2011) and Malmendier & Tate (2005). The proxy they use is measured by the investment decisions a firms management makes. This paper uses the capital expenditures over sales over total assets to assess managerial optimism. This paper also uses this measure because it also allows to measure optimism relative to members of a peer group.

(18)

18

From the theories of De Bondt and Thaler (1995), Jensen (1993), Ben-David et. al. (2007) and Brav et. al. (2005) could be expected that firms with high managerial optimistism in investment decisions are more likely to distribute lower amounts of dividends, on the contrary from Goncharov and Van Triest (2011) could be expected that management is optimistic in assessing the persistence of earnings components and thus paying higher dividends. To test this expectation the following hypothesis was developed.

Hypothesis 4: Managerial optimism has no impact on the distribution of income derived from fair value adjustments from investment properties

Another firm characteristic that possibly could influence the association between fair value adjustments and distributed dividends is the borrowing capacity of a firm. From the paper of Leary and Roberts (2010) and Fama and French (2002) could be found that borrowing capacity or debt capacity is associated with investment and dividend decisions. From the research of Chae et. al. (2009) could be found that firms deal with both agency costs between shareholders and the firm’s management and financial constraints between shareholders and external lenders. More severe agency costs between shareholders and the firm’s management could be measured by the cost of capital and the financial constraints are caused by higher cost of debt (Chae et. al., 2009). They argue when agency problems are more severe than the financial constraints, firms will pay more dividends with more efficient corporate governance to mitigate those agency problems. When external financing constraints are bigger than the agency problems, they found that those firms decrease their distributed dividends and even if the agency costs will increase due to this decrease in dividends. Distributing high levels of dividends could lead to higher cost of debt (Shivakumar, 2013). When creditors have weak rights, managers are likely to sign debt covenants that restrict dividend distributions to reduce future external finance costs (Brockman and Unlu, 2009). This could imply that firms with high cost of debt are less likely to distribute high levels of dividend.

A proxy that is used in this paper to assess the borrowing capacity is the corporate leverage ratio of a firm. When firms have high leverage (high debt to equity ratio), it is likely that the firm has restrictive debt covenants to mitigate the potential conflicts between

shareholder and debtholders (Jensen and Meckling, 1976). Leverage and the debt covenants decrease flexibility in investment decisions. Farinha (2003) state that this could lead firms in default.

(19)

19

value accounting, managers gain discretion in revaluating assets. Prior research suggest that firms that are highly leveraged and dealing with debt covenants are more likely to revalue assets (Brown et. al. 1992; Whittred and Chan 1992; Cotter and Zimmer 1995). With the revaluations, a firms management can signal liquidation values of those assets to lenders, and possibly reduce cost of debt. But on the contrary, shareholders want to receive higher dividends due to this fair value adjustments in the financial statements. This demand could lead to higher dividend payout ratios and thus increase the agency costs of debt, because concerns about financial distress increases. This could lead to higher cost of debt (Shivakumar, 2013). These theories are supported by Rozeff (1982), who states that higher dividend payouts can help to overcome agency problems with shareholders and management, while it could increase agency problems between external lenders and shareholders on the longer term leading to higher cost of debt. This paper created the following hypothesis to examine the effect of high leverage on the association between fair value adjustments and dividend policies. This research paper examines the effect of borrowing capacity with cost of debt and corporate leverage as proxies. This led to the following null-hypothesis.

Hypothesis 5: Borrowing capacity has no effect on the distribution of income derived from fair value adjustments of investment properties

5. Research methodology

5.1 Data and sample description

For the examinination of the effect of fair value adjustments on dividend policy choices of firms, this paper uses a sample of firms from the European Union which are subject to IFRS accounting standards and local GAAP. The IASB is promoting fair value accounting in the IFRS accounting framework, the use of fair value accounting is important for this research paper. In this research, data on European listed firms is gathered for the period 2005-2012 for this sample. First data is gathered on dividend payments and fair value adjustments. Secondly, data on the different firm characteristics is needed, like data on different proxies for the corporate governance structure of the firms, like board diversity, CEO duality, board size, a corporate governance score and board meetings attendance average. This paper used capital expenditures over sales to assess managerial optimism, and corporate leverage to assess

(20)

20

borrowing capacity. The data for this research is gathered from different databases, the data on fair value adjustments is gathered from Datastream and data on firm characteristics are

gathered from Datastream Worldscope. Financial institutions and utilities (ICB Codes between 7000-8999) are excluded from the sample, because of the extraordinary composition of their assets. All the variables that are used in the research are scaled with total assets to make the comparison between firms possible. The variables are winsorized at 1% and 99% to filter out outliers from the sample.

5.2 Models for the main hypotheses

Firstly, it’s critical for this research paper to assess the persistence of the fair value adjustments of investment properties. In this way could be assessed if the fair value adjustments should be distributed according to the Lintner (1956) framework. This paper examines the persistence of these components by controlling the effect of fair value adjustments for investment properties scaled over total assets (

UNR_INV_PR_REV

it) for multiple years (t-1, t-2, t-3, t-4) and the

return on assets before fair value adjustments from investment properties (

ROA_BFV

it-1; it-2; it-3; it-4) of the years t-1, t-2, t-3 and t-4, on the return on assets before fair value adjustments of year

t (

ROA_BFV

it). In this way the effect of fair value adjustments in previous years on current

income could be measured. From prior research could be found how to assess the persistence of earnings components (Sloan, 1996) and in specific for fair value adjustments (Goncharov and Van Triest, 2011) in a regression model. The following regression is developed to examine the predictability of fair value adjustments from investment properties on the future return on assets before fair value adjustments (

ROA_BFV

it). The effects are additionally tested with

control variables like firmsize (

SIZE

it) which is the natural logaritm of total assets (

TA

it), debt

(

DEBT

it) the proxy for leverage which is total debt over total assets, cash over total assets

(

CASH

it) and sales growth over total assets (

GROWTH

it) and some dummy variables like

country (

ΣDCOUNTRY

), firmindustry (

ΣDINDUSTRY

) and a year dummy (

ΣDYEAR

(21)

21

+

α

12

GROWTH

it

+

Σ

DCOUNTRY +

Σ

DINDUSTRY +

Σ

DYEAR +

Ɛ

it(1)

If the fair value revaluation of investment properties predict the future return on assets before fair value adjustments, the coefficients of these variables

α

drift over years in dividends. The effect is controlled on by different control variables.

ΔDIV

it

=

Σ

DINDUSTRY +

Σ

DYEAR +

+

Σ

DCOUNTRY +

Σ

DINDUSTRY +

Σ

DYEAR +

Ɛ

it (3)

This equation could be further adjusted to examine the difference between upward and

(22)

22

(

FVREV

- ). This is helpful to examine the findings of DeAngelo and DeAngelo (1990), who found that when firms deal with losses they are not willing to ommit dividend payments but only cut them so the expected smooth relationship between earnings and dividends is not that smooth when the firm is making losses. This paper examines the effect of upward and

downward fair value adjustments on dividends and examine if the results also hold for fair value adjustments.

DDIFF

it

=

β

+

β

9

GROWTH

it

+

Σ

DCOUNTRY +

Σ

DINDUSTRY

+

Σ

DYEAR +

Ɛ

it (4)

In the European Union, the research area of this paper, there’re different laws in place dealing with the ability of the firms to distribute unrealised gains from investment properties in dividend. In this paper a countrydummy is created, for member states where distribution of unrealised fair value gains is (not) allowed UNR_INV_PR_REV*ddistr_allowed is 1(0) and

UNR_INV_PR_REV *ddistr_not_allowed is 0(1). These countries are Bulgaria, Cyprus, Czech

Republic, Estonia, Finland, Latvia, Lithuania, Portugal, Slovakia, Slovenia, Greece and Spain (2008-2012) and Luxembourg (2010-2012). The countries where distribution is not allowed are Denmark, Italy, Malta, The Netherlands, Poland, The United Kingdom, Ireland, Hungary, Austria, Belgium, France, Germany, Romania, Sweden, Spain (2005-2007) and Luxembourg (2005-2009). It’s expected that in countries where it’s allowed to distribute unrealised gains, the unrealised gains of investment properties will impact the change in dividends over assets more than in countries where distribution isn’t allowed.

DDIFF

it

= β

0 +

β

1

Σ

DCOUNTRY +

Σ

DINDUSTRY +

Σ

DYEAR

(23)

23 5.3 Models for firm characteristics

To examine the moderating impact of the different firm characteristics this paper has taken into account these variables for those characteristics in the equation of the main hypothesis

(equation 3). The firm characteristics that are examined are the corporate governance structure, managerial optimism and borrowing capacity. From prior literature could be expected that these characteristics have an impact on the association between fair value adjustments of investment property and the distribution of dividends.

The first firm characteristic under examination is the corporate governance structure of firms. When a firm has a strong(weak) corporate governance structure relative to their peer, the dummy variable DStrongCG is 1 (0) and DWeakCG is 0 (1). From the papers of Gugler and Yurtoglu (2003), Farinha (2003) could be expected that dividend distribution is higher for firms with strong corporate governance structures. The strenght of the corporate governance structure is assessed by four proxies, board diversity, CEO duality, board size and a corporate

governance score. The hypothesis is stated as a null hypothesis, so when β1 and β2 are zero, the

corporate governance structure does not have a significant impact on the association between fair value revaluations of investment properties and dividend policies. This is tested by the following equation.

DDIFF

it

=

DCOUNTRY +

Σ

DINDUSTRY +

Σ

DYEAR +

Ɛ

it (6)

The second firm characteristic that is examined is managerial optimism. Expected is that optimistic management is investing more and thus is not willing or able to pay high amounts of dividends. On the otherside, management could be to optimistic in assessing the persistence of a fair value revaluation and distribute unrealised gains in dividends. So this paper uses the capital expenditures over sales as a proxy for the assessment of managerial optimism. A firms management is qualified as overoptimistic(underoptimistic) when its capital expenditures are higher(lower) than the mean of the firms in the sample. This measure is used because it allows to measure optimism relative to members of a peer group. When the firm’s management is overoptimistic (underoptimistic) the DOverOPT is 1 (0) and DUnderOPT is 0 (1). The null

(24)

24

hypothesis will be supported when

β

1 and

β

2 are zero, this means that managerial optimism has

no significant impact on the association between fair value adjustments and the distribution of dividends. This is tested by the following regression model.

DDIFF

it

= β

0 +

β

1

DOverOptUNR_INV_PR_REV*

it

+ β

2

DUnderOPT*

+

Σ

DINDUSTRY +

Σ

DYEAR +

Ɛ

it (7)

The third firm characteristic that is expected to have an impact on the first hypothesis is a firms borrowing capacity. In this paper borrowing capacity is measured with the proxy corporate leverage. When firms have a higher (lower) debt to equity ratio relative than their peers,

DHighLEV is 1 (0) and DLowLEV is 0 (1). The expectation is that firms with a higher debt to

it

+

β

2

DLowLEV*

(25)

25 6. Empirical findings

6.1 Descriptive statistics

For the sample firms that are used in this paper, the following financial characteristics are identified. These characteristics are needed to examine the hypothesis of this paper. First of all 66.3% (0.663) of the sample of 638 observations (missing values excluded) made fair values adjustments for their investment properties. The mean of the unrealised fair value adjustments scaled over total assets is (-0.0003) and a median of (0). When splitting those adjustments up into positive and negative adjustments, the mean of the 236 positive adjustment is 0.015 and of the 187 negative adjustments the mean is -0.019. The return on assets before unrealised fair value adjustments (

ROA_BFV

it) for an average firm has a mean of (0.008) and the median is

(0.035). The mean of the scaled debt (

DEBT

it) which is an indicator for leverage, for the

sample is (0.243) and has a median of (0.226). The mean of salesgrowth over total assets

(

GROWTH

it) is (0.028) and a median of (0.036). Cash (

CASH

it) has a mean of (0.129) and a

median of (0.075) and firmsize (

SIZE

it) has a mean of (12.474) and a median of (12.368). For

(26)

26

Table 3 Descriptive statistics Statistic

Variable

Obs Mean Median Std. dev. Min Max

DRevaluation 638 0.663 1 0.473 0 1 UNR_INV_PR_REVit 638 -0.0003 0 0.029 -0.153 0.140 UNR_INV_PR_REV+ 236 0.015 0 0.027 0.00001 0.140 UNR_INV_PR_REV- 187 -0.019 0 0.035 -0.153 -0.000004 UNR_INV_PR_REV* ddistr_allowed 638 0.0004 0 0.077 -0.160 1.920 UNR_INV_PR_REV* ddistr_not_allowed 638 -0.0007 0 0.071 -1.167 0.387 UNR_INV_PR_REV* DHighLev 638 0.0003 0 0.011 -0.091 0.140 UNR_INV_PR_REV* DLowLev 638 -0.001 0 0.026 -0.153 0.140 UNR_INV_PR_REV* DOverOpt 638 0.00004 0 0.013 -0.153 0.140 UNR_INV_PR_REV* DUnderOpt 638 -0.0003 0 0.025 -0.153 0.140 DDIFFit 1914 0.001 0 0.015 -0.064 0.077 DIVt-1 1939 0.012 0.005 0.021 0 0.129 ROAit 2314 0.021 0.040 0.134 -0.691 0.335 ROA_BFVit 578 0.008 0.035 0.146 -0.691 0.287 CapExp/Salesit 2181 0.00008 0.00001 0.0003 0 0.002 GROWTHit 1960 0.028 0.036 0.277 -1.291 0.832 CASHit 2252 0.129 0.075 0.151 0.00002 0.782 SIZEit 2552 12.474 12.368 2.073 1.609 17.964 DEBTit 2249 0.243 0.226 0.199 0 0.980

(27)

27

For the sample of this research the correlation matrix with the interactions between the variables is presented in table 3. The decision to revaluate the fair value of an investment property is significantly correlated with the leverage position of a firm (0.121) and the amount of cash a firm has (-0.179).

Table 4 Spearman correlation matrix

Variable DRev DDIFFit UNR _REVit

ROA_BFV DIVt-1 ROAit SIZE GROWTH DEBT CASH

DRev 1 0.028 0.082 * -0.034 0.101 ** -0.023 0.018 -0.060 0.121 *** -0.179 *** DDIFFit 0.028 1 0.078 * 0.313 *** _-0.025 0.320 *** 0.133 *** 0.118 *** -0.151 *** 0.139 *** UNR_ REVit 0.082 * 0.078 * 1 0.085 * 0.037 0.282 *** 0.073 0.140 *** 0.044 0.038 ROA_ BFVit -0.034 0.313 *** 0.085 * 1 0.263 *** 0.945 *** 0.133 *** 0.395 *** -0.389 *** 0.219 *** DIVit-1 0.101 ** 0.025 0.037 0.263 *** 1 0.256 *** 0.210 *** 0.024 -0.147 *** -0.052 ROAit -0.023 0.312 *** 0.282 *** 0.945 *** 0.256 *** 1 0.158 *** 0.390 *** -0.357 *** 0.222 *** SIZE 0.018 0.133 *** _0.073 0.133 *** 0.210 *** 0.158 *** ₁ 0.134 *** 0.188 *** _-0.006 GROWTH -0.060 0.118 *** 0.140 *** 0.395 *** _0.024 0.391 *** 0.134 *** ₁ -0.192 *** 0.161 *** DEBT 0.121 *** -0.151 *** 0.044 -0.389 *** -0.147 *** -0.357 *** 0.188 *** -0.192 *** 1 -0.380 *** CASH -0.179 *** 0.139 *** _0.038 0.219 *** _-0.052 0.222 *** _-0.006 0.161 *** -0.380 *** ₁

Note: Significant at 1%, 5% and 10% indicated when ***, ** and * respectively. Variables winsorized at 1% and 99%. Definitions of variables in Appendix A.

(28)

28 6.2 Multivariate analysis

6.2.1 The persistence of unrealised gains/losses from investment properties

The hypothesis 0 is tested by examining the persistence of fair value revaluations of investment properties and the results are presented in the table 4. Hypothesis 0 tests the persistence of unrealised fair value revaluations from year t, t-1, t-2, t-3 and t-4 and the persistence of return on assets before fair value revaluations from the years t, t-1, t-2, t-3 and t-4 as well. The model included the additional control variables, firmsize, growth, debt, cash and controldummies for year, country and industry. Because of limited data on unrealised gains and losses from investment property for 5 consecutive years (only 37 firm-year observations), this paper also examines the persistence of unrealised fair value adjustments for a maximum of only three years, two years and one year backwards.

The results of these models indicate that the return on assets before fair value

revaluations of investment property (ROA_BFV) are persistent in the models with only some exceptions. So ROA_BFVit is determined by the return on assets before fair value revaluations

of investment property of the previous four years. In the model with t-4 included ROA_BFVt-1 (-0.729), ROA_BFVt-3 (-1.077) and ROA_BFVt-4 (-1.488) are significantly persistent. The main findings of this model indicate that the unrealised gains and losses from investment property aren’t persistent and don’t predict future return on assets effectively. Only

UNR_INV_PR_REVt-2 (0.757) in the first model is significant but the other figures aren’t, so

the findings implicate that unrealised gains and losses from investment property aren’t persistent. These findings contradict to the preliminary evidence of Wirtz (2013), who found that unrealised fair value gains and losses map future cash flows effectively. Of the control variables debt and growth are significant in all models, size is significant in the first three models and cash is only in the first model, this could be due to the limited amount of firms in that model. The finding that the unrealised fair value revaluations of investment properties are transitory implies that those revaluations shouldn’t affect the dividend policies of firms.

(29)

29

Table 5 The persistence of gains/losses of investment properties

Dependent variable ROA_BFVit

Independent variables Coefficients

ROA_BFVt-1 _-0.729** _-0.419** _-0.377*** _0.157*** ROA_BFVt-2 -0.257 -0.222 _-0.248** ROA_BFVt-3 _-1.077** -0.259 ROA_BFVt-4 _-1.488* UNR_INV_PR_REVt-1 -0.084 0.496 0.128 -0.222 UNR_INV_PR_REVt-2 _0.757* 0.075 -0.070 UNR_INV_PR_REVt-3 0.132 0.082 UNR_INV_PR_REVt-4 -0.195 SIZEit _0.151*** _0.030** _0.013* 0.004 GROWTHit _0.125** _0.166*** _0.132*** _0.086*** DEBTit _-0.741*** _-0.539*** _-0.410*** _-0.145*** CASHit _0.489*** 0.036 -0.064 -0.027 Intercept _-1.400** 0.058 _-1.309*** 0.141

Year, country and industry dummies

Yes Yes Yes Yes

Observations 37 81 152 292

R

2

0.9933 0.8891 0.8413 0.7382 F-Score (P>F) 23.91 (0.0011) 6.56 (0.0000) 8.04 (0.0000) 7.44 (0.0000)

Note: Significant at 1%, 5% and 10% indicated when ***, ** and * respectively. Variables winsorized at 1% and 99%. Definitions of variables in Appendix A.

6.2.2 The effect of unrealised gains/losses from investment properties on dividend policies

After examining the persistence of the unrealised gains and losses from investment property, the effect of those gains and losses on a firms dividend policy is examined. And additional tests are included in table 6 to test for the difference between positive and negative fair value

adjustments and to examine the difference between countries where it’s allowed to distribute the unrealised gains from investment property and the countries where distribution isn’t allowed.

From the empirical findings could be found that unrealised gains and losses

(UNR_INV_PR_REVit) don’t have a significant impact (0.008) on the dividend policy choices

of firms. This result supports the framework of Lintner (1956), which states that transitory earnings shouldn’t affect the dividend distribution. And these findings disproves the concerns