1 *Corresponding author Oliviero Roggi University of Florence Via delle pandette, Firenze Alessandro Giannozzi* University of Florence Via delle pandette, Firenze Abstract The paper aims to explore the role of the fair value hierarchy as useful information tool in estimating the liquidity risk of an asset. We investigate the existence of a relationship between the three levels of fair value and the investors reaction (313 financial and non financial companies listed on the Eurostoxx Index) during an event of liquidity expansion or liquidity contraction. The usefulness to investors of liquidity risk information is analyzed by tracing investors reactions to 103 events during the financial crisis ( ) using fixed effects model and PLS regression. Our findings demonstrate that investors firmspecific reaction to the crisis events is influenced by the threelevel of fair value hierarchy. It is also demonstrated that investors react according to the level of liquidity risk in both financial and nonfinancial firms. During liquidityconstraining events, investors negative reaction is stronger for the most illiquid assets and liabilities (Level 3) than for the liquid ones. During liquidityexpanding events, investors react more positively to the most illiquid assets (Level 3) than to the liquid ones (Level 1 and 2). Keyword: Liquidity Risk, Fair value disclosure, stock returns JEL CODES: G18, M40, G30
2 1. Introduction The time frame has been the scene of major economic upheavals which have profoundly marked the financial world. Events with high economic impact have occurred, such as the collapse of Lehman Brothers, which experts considered highly unlikely in the financial sector. These events have led to high volatility in the financial markets which has caused losses and the spreading a fear of a depression amongst investors of comparable to that of For the above mentioned reasons it is interesting to deepen the understanding of the causes of this financial crisis, focusing on the risk of corporate liquidity. In particular, it seems that there could be a relationship between this type of risk and the information provided by the companies to the markets on the toxic assets held in their balance sheet. The value of financial instruments, as is known, is measured according to the fair value principle. The international standard setters (IASB and FASB) have agreed that the evaluation of such balance sheet items can be achieved through the use of different methodologies and in accordance with a liquidity criterion of the evaluated financial instruments. As is known, the first form of assessment used for financial instruments is the price at which it is listed in a liquid and active market (the so called marktomarket or 1 st level). Securities assessed in this manner are classified in the first level of Fair Value Hierarchy and according to standard setters their price represents the nearest concept to that of fair value. If market prices are not available, the nearest approximation of a security s fair value is the price of a similar financial instrument listed in a liquid and active market or the price of a recent and similar transaction between knowledgeable and willing parties (2 level). In the event that such information is unavailable, a security s fair value can be estimated by using financial and statistical models of common acceptance (the so called marktomodel or 3 rd level). This last technique may require the imposition of unrealistic assumptions causing the generation of errors and / or information asymmetry between the firm and the financial markets. By defining liquidity risk as the inability to raise new funds (funding liquidity) or to liquidate assets on the market at an "expected price (asset liquidity risk) (Brunnermeier, 2007), Level 1 of the Fair Value Hierarchy appears to be the most liquid since it is made up of assets/liabilities whose values are based on their market price. In contrast, the value of
3 securities included in Level 3 of the Fair Value Hierarchy is calculated through internal models, resulting in assets/liabilities with a higher liquidity risk.the higher the level 3 amounts are, the greater the assets liquidity risk will be and therefore the higher the total liquidity risk of the company linked to the uncertainty of disinvesting securities at a predictable price (Brunnermeier 2007). Assuming the existence of the link described above and the presence of a contraction of liquidity on the financial markets (as happened during the 2008 financial crisis) investors tend to penalize illiquid companies (i.e. ones with large amounts of Level 3). In fact, low liquidity levels on the financial markets lessen the possibility of disinvesting assets at a predictable price, causing serious consequences for companies. The increased uncertainty should generate a negative impact on company s value. In an efficient financial market this would be reflected in stock prices. Therefore, a high level of liquidity risk may cause a decline in corporate stock prices. In particular, there may be situations in which the assumptions on which the models are built result particularly unrealistic. This is what happened during the recent financial crisis. Activities and / or liabilities may become illiquid, i.e. investors may not trust most of the securities assessments and may decide to no longer include theme in their portfolios. The main goal of this paper is to explore the role of the fair value hierarchy as an information tool in calculating the liquidity risk of a security/company. In particular, the existence of a relationship between the three levels of fair value and the reaction of investors when a liquidityconstraining or liquidityexpanding event occurs is investigated. This will allow the understanding of whether fair value hierarchy properly informed investors of firms'liquidity risk or whether, ultimately, its introduction by the international standard setter has been useful for investors. Lev & Zhou (2009) have contributed to the debate on fair value accounting by analyzing the separation of financial assets and liabilities into three liquidity levels. In particular, they investigated investors reaction to crisis events during the last months of 2008 (the peak of the financial crisis) based on the fair value disclosures of 3,929 U.S. financial and nonfinancial companies under SFAS 157. They demonstrated a complex reaction to the liquidity risk information conveyed by the three fair value levels. Starting from Lev & Zhou s (2009) hypotheses, investors reaction to crisis events in European firms under IAS 39 and IFRS 7 are analyzed. Moreover, this analysis is conducted between February 17 th, 2008 and June 22 nd, In addition, a non
4 parametric methodology (PLS regression) is applied to test the robustness of the models The paper is structured as follows. In Section 2, a survey of the most relevant fair value literature is provided. Sections 3 and 4 illustrate our research design and the data sample. Sections 5 and 6 are dedicated to the measurement of investors reaction to crisis events. In particular, the group of crisis events and the cumulative abnormal returns (CARs) are introduced. In Section 7, our hypotheses are stated. Section 8 provides the results of the empirical analysis. In Section 9, the final conclusions are stated. 2. A Summary on Fair Value Literature Starting in the early nineties, extensive research has been carried out on the use of market prices for asset and liability pricing. First of all it must be stated that researchers have focused primarily on the valuerelevance of fair value. For valuerelevance we intend the ability of fair value data to incrementally explain the behavior of the stock prices. Mary E. Barth 1 (1994) supports the existence of a strong relationship between these two variables and believes that banks stock prices changes can be measured by the use of the securities at market value. This important result was confirmed by Barth, Landsman and Wahlen 2 (1996). They demonstrate that when banks net profits are measured using Fair value accounting, they result more volatile than those calculated with historical cost asset valuations. Barth, Landsman and Wahlen (1996) also demonstrate that banks violate regulatory capital requirements more frequently when using Fair Value Accounting than when using Historical Cost Accounting. In 1996, Barth, Beaver and Landsman 3 noted 1 Barth, M. E. (1994), Fair value accounting: evidence from investment securities and the market valuation of banks, The Accounting Review, Vol. 69, January, pp Barth, M. E. and Landsman W. R. and Wahlen J. M. (1995), Fair Value Accounting: Effects on Banks'Earnings Volatility, Regulatory Capital, and Value of Contractual Cash Flows, Journal of Banking and Finance, Vol. 19, pp Barth, M. E. and Beaver W. H. and Landsman W. R. (1996), Valuerelevance of banks fair value disclosures under SFAS No. 107, Accounting review, Vol. 1, pp
5 that bank stock prices are partly explained by the difference between estimated fair value under SFAS 107 and the current book value. Unlike previous researches, Barth, Beaver and Landsman (1996) noted for the first time that the additional explanation required by SFAS 107 significantly affected bank stock prices. By the contrary in the same year, Eccher, Ramesh and Thiagarajan 4 (1996), performed an analysis in which the historical cost seemed to have a higher explanatory power over bank stock prices, both in absolute and in relative terms. Previously, Beaver and Landsman 5 (1983), Beaver and Ryan 6 (1985) and Bernard and Ruland 7 (1987) had found that the estimated fair value calculated for certain types of assets according to SFAS 33, had no incremental explanatory power over bank stock prices than had their book value. In addition, other studies such as Murdoch 8 (1986), In 1996 Cornett, Rezaee andtehranian 9, investigating the impact on stock prices of 23 fair value accounting (FVA) statements issued by financial institutions demonstrated how the increases (decreases) caused by the likelihood of a new emission of FVA standards are followed by negative (positive) abnormal returns on bank stock prices. During the following years numerous studies were conducted on the valuerelevance of Accounting standards. Holthausen and Watts 10 (2001) are critical of all research conducted thus far, stating that these studies offer little support to the standard setter, since they fail to create a generalized descriptive theory. 4 Eccher, E. A. and Ramesh K. and Thiagarajan S. R. (1996), Fair value disclosures by bank holding companies, Journal of Accounting and Economics, Vol. 22, pp Beaver, W. H. and Landsman W. R. (1983), Incremental information content of Statement 33 disclosures, FASB: Stamford, Connecticut. 6 Beaver, W. H. and Ryan S. (1985), How well do Statement No. 33 earnings explain stock returns?, Journal of Financial Analysts, Vol. 41, September/October, pp Bernard, V. L. and Ruland R. (1987), The incremental information content of historical cost and current cost income numbers: time series analyses for , The Accounting Review, Vol. 62, October, pp Murdoch, B. (1986), The information content of FAS 33 returns on equity, The Accounting Review, Vol. 61, No. April, pp Cornett, M. M. and Rezaee Z. and Tehranian H. (1996), An investigation of capital market reactions to pronouncements on fair value accounting, Journal of Accounting and Economics, Vol. 22, pp Holthausen, R. W. and Watts R. L. (2001), The relevance of the valuerelevance literature for financial accounting standard setting, Journal of Accounting and Economics Vol. 18, pp. 375.
6 Barlev and Haddad 11 (2003) state the need to evaluate financial reports based on the benefit investors gain by the reduction of agency costs and the improved efficiency in corporate control that these reports are able to offer. In fact, they argue that Historical Cost Accounting offers ample room for balance sheet manipulation, leading to an incorrect display of the current situation and so reinforcing the relevance of Fair Value Accounting.. Fair Value Accounting draws attention to equity and its variations and, therefore, managers should be asked to better monitor the "health" of equity itself, its maintenance and the profits it generates. With the application of FVA, Barlev and Barlev and Haddad (2003) suggest that managers create an additional financial report, which is to indicate the operations conducted by the firm in order to maintain equity. Historical Cost Accounting induces managers to create hidden reserves of capital to cover, if necessary, administrative errors, increasing agency costs. Morevoer, historical Cost Accounting distorts the financial ratios on which creditors concentrate their attention, such as the long term debt coverage ratio and the interest coverage ratio. Barlev and Haddad (2003) claim that FVA has another important feature: it stimulates management to better understand and study the market, making them face a global, open and competitive environment, and therefore inducing them to be more competent. Freixas and Tsomocos 12 (2004), historical cost allows the distribution of a higher amount of bank dividends over time. According to these scholars, in fact, the bank's role as institutional intervenor of intertemporal smoothing may worsen with the use of fair value accounting. Generally, it is believed that a sufficiently high level of criminal manager prosecution incentives managers to report accurate and correct values in balance sheets. If moral hazard and information asymmetry are very high, the current value (fair value) could thus increase market discipline, inducing managers to make the right decisions. According to these scholars, if this method of measurement is correctly used it could prevent systemic crises since information on financial issues is obtained prior than with Historical Cost Accounting. Fair Value Accounting creates lower bankruptcy costs for the environment because it promotes the bankruptcy of Distress companies, preventing them from prolonging their existence and redistributing their Barlev B. and Haddad J. R. (2003), Fair value accounting and the management of the firm, Critical Perspectives on Acccouting Vol. 14, No.1, Freixas, X. and Tsomocos D. (2004), Book vs. fair value accounting in banking, and intertemporal smoothing, Oxford Financial Research Centre working paper.
7 resources amongst healthy companies. This will also prevent banks from betting on the survival of Distress companies and investing in their equity and supporting high risks. Tsomocos and Freixas (2004) observe, however, that the use of fair value increases the volatility of bank profits and lacks accuracy since it relies on subjective judgments for the assessment of illiquid items. Banks insure themselves against unforeseen contingencies through their dividend policy, building up reserves in times of economic prosperity and using them in times of crisis, in order to equalize the amount of dividends paid to shareholders over time. By using fair value, banks might be tempted to shift their focus on stakeholders shortterm interests, even though regulation could prevent this from happening through the imposition of stricter rules. According to Freixas and Tsomocos (2004), the adoption of fair value measurements induces bank managers who expect temporary adverse shocks in bank stock prices to act in a conservative way, such as not investing in risky assets, reducing interest rates on deposits or choosing not to distribute dividends by reducing intertemporal smoothing. Landsman (2006) identifies key issues that standard setters should seriously consider. In particular a key issue is related to how to minimize management's manipulation on economic and statistical models used to assess financial instruments classified in the 3rd level of Fair Value Hierarchy. In addition, Landsman (2006) clarifies that the institutional differences in the implementation of Fair Value Accounting will play an important role towards an effective and efficient use of the above mentioned evaluation system. Landsman also emphasizes the need to assess the incremental explanation that the disclosure required by the standard setter offers shareholders.. Allen and Carletti 13 (2007) demonstrate that there is a possibility that Fair Value Accounting can lead to a contagion effect between banks and insurance and that in times of crisis this contagion is not desirable, since it could then lead nondistress banks to be insolvent. This is because markets can be illiquid even when prices are continuous, since in some cases a further increase in the demand or in the offer can cause a significant price change. Allen and Carletti (2007) argue that during an economic crisis, relatively low prices could lead asset managers to aim towards achieving liquidity objectives rather than to reflect on future cash flows that can be generated from such assets. In this case, Historical Cost Accounting is certainly more 13 Allen, F. and Carletti E. (2007), Marktomarket accounting and liquidity pricing, Journal of Accounting and Economics Vol. 45, No. 1, pp
8 efficient. Therefore, when there is an economic crisis and the evaluation system is not adjusted for illiquidity, the only way to mitigate a contagion effect is to temporarily suspend the application of Fair Value Accounting. In their subsequent study, Allen and Carletti 14 (2008) add that Fair Value Accounting should be used particularly when asset prices collapse due to fundamental reasons, i.e. the falling expectations on future cash flows which are reflected in stock prices. Allen and Carletti (2008) also clarify that there are more damaging effects when prices collapse and the historical cost accounting method is being used. According to these authors, standard setters should rightly apply Fair Value Accounting even in periods of illiquidity providing some corrections for Fair Value Accountings aforementioned shortcomings. Allen and Carletti (2008) propose to supplement market valuations with asset value estimated by financial models and by Historical Cost Accounting. This in order to inform investors on the cash flows generated by the associated securities and on the ability of banks to reclassify their assets at historical cost (i.e. from Available For Sale to Held To Maturity). The incremental disclosure requested by Allen and Carletti (2008) would prevent bank managers moral hazard, which would then lead them to reclassify assets even in non illiquid periods in order to park risky and volatile (but profitable) assets in bank balance sheets, basically implementing arbitrage regulation. While some sustain that the historical cost leads to some forms of inefficiencies, others believe that the market value could inject artificial risks in the economic system, degrading the value of the information included in prices and causing suboptimal choices. Plantin, Sapra and Shin 15 (2008) found that the damage caused by the market value reaches its maximum value when this method is used for the assessment of longterm illiquid and senior instruments which form a large portion of banks assets. In reviewing previous literature, the abovementioned authors pointed out the inadequacy of pricing certain asset categories below market value, particularly, for the asset categories assessed on OTC markets. The IMF study 16 of 2008 states that standard setters are following the correct direction with the implementation of Fair Value 14 Allen, F. and Carletti E. (2008), Should financial institutions mark to market?, Banque de France Financial Stability Review, October. 15 Plantin, G. and Sapra H. and Shin H. S. (2008), Markingtomarket: panacea or Pandora s box?, Journal of Accounting Research, Vol. 46, No. 21, pp International Monetary Fund (2008), Global financial stability report. Chapter 3: Fair value accounting and procyclicality, October.
9 Accounting, despite the difficulties caused by the measurement and the volatility that it entails. This is explained by the better assessment of the company's financial situation (especially for financial firms), which is obtained through the use of market prices, even though, according to the IMF (2008) this accounting method, must be improved. For the IMF (2008), Fair Value Accounting allows investors to have greater transparency on the effects of the economic volatility on company s perfomances, exacerbating cyclical movements. According to this institution, if high profits result in poor incentives for management in times of economic prosperity, it is also true that insecure evaluations could damage corporate funding in times of crisis, resulting in lower levels of available credit. All this to emphasize the need to apply specific crisis covenants and limitations to short term gains during periods of economic prosperity. The IMF (2008) also points out that the balance sheet of a highly capitalized company is more resistant to market fluctuations, especially when there is a large amount of assets assessed at market value rather than when are liabilities valued with the same method. Similarly, the cyclical volatility of corporate financial statements is exacerbated when, with the use of market prices as accounting method, there is a severe lack of liquidity in corporate balance sheets. The IMF (2008) therefore suggests that, in addition to having sufficient capitalization and liquidity, companies should counterbalance their total assets at fair value with a similar set of liabilities, in order to alleviate the procyclicality of corporate balance sheets. Laux and Leuze 17 (2009) affirm that countercyclical capital requirements are a better solution to the difficulties brought forth by Fair Value Accounting. According to these authors, we must investigate the interrelationships between Fair Value Accounting and financial crises: for example, implementation problems of this accounting method and, particularly, the risks of litigation, which may have played a key role in the recent financial crisis should be better explored. Laux and Leuze (2009), in fact, suggest that the current crisis may have been exacerbated by transparency issues on the financial markets (especially banks) rather than by the overreaction to Fair Value Accounting: in times of economic prosperity banks are unable to increase their leverage ratio both with Fair Value Accounting and Historical Cost Accounting. This may contrast the critics of Fair Value Accounting, who see a high distortion for securities held to maturity (HTM) Laux, C. and Leuz C. (2009), The crisis of fair value accounting: making sense of the recent debate, Accounting, Organizations and Society, Vol. 34, pp
10 and consider this method as scarcely credible and reliable if obtained with the use of models. By using Historical Cost Accounting, managers have an information advantage over their controllers. This can lead to management issues, since it is difficult to give managers flexibility without sustaining any risk of free riding. Laux and Leuze (2009) argue, therefore, that it is correct to give more flexibility to business managers only if the two parties (the controller and the agent) have the same set of information. Otherwise, more stringent accounting standards or possible contagion effects on financial companies are the price to pay for correct market signals to investors. According to Penman 18 (2006), the FAS 157 considers fair value as an exit value, raising two main questions: the first consists of the pertinence of fair value in capturing value for shareholders and the second relies on the possibility of applying it to the aggregates of assets and liabilities. In his paper, Penman (2006) distinguishes what the advantages and disadvantages of using the fair value are. In the event that shareholder value is created only by the exposure to market prices, Fair Value Accounting leads to favor the shareholders. The disadvantages, according to this author, reside in the possible mismatch between asset and liability evaluations (i.e. when different types of security assessments are used), in the possible introduction of financial bubbles in corporate accounts, in the total replacement of information that can be derived from the historical price (the price, in fact, also depends on the historical price and not only on the information gathered with the market price) with that gathered from the market. In his opinion, the primary issue is not to determine which between the historical cost and the market price constitutes the best evaluation system, since these two types of accounting methods function differently. The historical cost, in fact, according to Penman (2006) is able to better predict a company s future earnings when compared to the market price. It also concentrates investors attention on the income statement rather than on equity as happens with the use of market price. With Historical Cost Accounting the value of a company depends on its profitability and not on the subsequent equity changes. Historical cost accounting also does not include the present value of projected revenues, but exclusively recognizes the benefits of adding value to 18 Penman, S. (2006), Financial Reporting Quality: Is Fair Value a Plus or a Minus?, paper for presentation at the Information for Better Markets Conference Institute of Chartered Accountants in England and Wales, December 18 th 19 th.
11 traded goods and transactions. Penman also argues that Historical Cost Accounting should be used in its natural framework, and only here can it be correctly judged. 3. Data sample In the data sample reference is made to the European context as it represents the largest geographical area which applies the accounting standards issued by the IASB. The companies chosen for the sample are those listed in the Eurostoxx index. The Stoxx indexes have a high level of liquidity. They are widely used in financial practice as "underlying" financial instruments such as ETFs, futures, options and structured products and as a benchmark for measuring risk and corporate performance. The Eurostoxx index consists of a variable number of small, medium and largely capitalized companies located in 12 Eurozone countries (Austria, Belgium, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal and Spain). The sample used for the analysis consisted of 313 companies listed in the Eurostoxx index. The observation period is the 2008, 2009 and 2010 fiscal years. Because the 2008 economic crisis started in the financial sector and subsequently spilled over to the real sector, we study the two sectors separately. Two subsamples are built based on the dataset, both for financial and nonfinancial firms since financial data differs substantially between these two categories of firms, as can be seen from the descriptive statistics shown in the Appendix. In fact, there is a relevant difference between the values of financial firms balance sheet items measured at fair value and the values of those that are not. Also a much stronger leverage is noticeable in all three periods for financial firms. The latter achieve a lower ROA than nonfinancial firms, while the values of operating activities vary in the three years considered. Further, the values at fair value of financial firms weighed significantly on the total (55.1% and 28.9% for assets to liabilities) than weighed those made of nonfinancial firms. In subsequent years a drastic drop of these percentages for both subsamples is evident, due to the financial crisis. In 2007 the EBIT of non financial enterprises is higher than that of financial enterprises, who were profiting greatly due to the precrisis situation on the credit markets. In the 2008 financial statements of nonfinancial companies a higher average value of the statistics is noticeable, as is shown in the Appendix, since these
12 firms were slightly more affected by the financial crisis. In 2009, thanks to the reclassification and stabilization of the financial markets, the financial companies are back to recording a higher average EBIT. Considering the medians of the various hierarchical levels of fair value, it is noticeable that level 3 Assets has higher relevance for financial firms than that of the nonfinancial companies.. The sum of the percentages of the medians of levels 2 and 3 of the financial firms is always higher than when compared to nonfinancial corporations. 4. Research design According to Fama, Fisher, Jensen e Roll (1969) and Ball e Brown (1968), our analysis takes the form of a typical event study 19. Here below a summary of our research design: 1. Sample selection and data collection (stock market prices, Fair value assets and liabilities, ect ) 2. Identification of liquiditycontracting and liquidityexpanding events in the period and aggregation in event groups 3. Estimation of the abnormal returns (dependent variable) after each crisis event. 4. Check of the statistical significance of each event by ANOVA (ANalysis Of VAriance) 5. Regression analysis using fixed effects model and Partial Least Square (PLS) as robustness check, separately for financial and nonfinancial companies. To test our hypotheses an OLS regression was carried out for the group of financial companies, for that of nonfinancial companies and for the full sample of companies: CAR = α + β fair _ value _ assets _ level1+ β fair _ value _ assets _ level β fair _ value _ assets _ level β n x n 19 MacKinlay, A. C. (1997), Event studies in economics and finance, Journal of Economic Literature, Vol. 35, No. 1, pp. 139; Binder, J. J. (1998), The event study methodology since 1969, Review of Quantitative Finance and Accounting, Vol. 11, pp ; Fama, E. F. and Fisher L. and Jensen M. C. and Roll R. (1969), The adjustment of stock prices to new information, International Economic Review, Vol. 10, No. 1, pp. 121; Ball, R. and Brown P. (1968), An empirical evaluation of accounting income numbers, Journal of Accounting Research, Vol. 6, No. 2, pp ; Brown, S. J. and Warner J. B. (1985), Using daily stock returns: the case of event studies, Journal of Financial Economics, Vol. 14, No. 1, pp
13 For each event group three regressions are run: the aforementioned regression, the aforementioned regression with fixed effect correction and, finally, a regression which compounds the possibility to reclassify assets/liabilities at their historical cost introduced by the IASB in midoctober Each regression is run with and without the control variables. 5. Crisis events In the period from February 2 nd, 2008 to June 22 nd, 20 we identified 103 crisis events. Appendix (Table 13) provides a brief description of each event. Furthermore, according to Lev & Zhou (2009) a twoday event window for stock return computation is used. In particular, it is assumed that the effects of each individual event terminate within a two day period (the same day and the next). The mean cumulative event returns (MCRs) are calculated to measure the stock market reaction to the 103 crisis events. In general, the direction of investors reaction in both financial and Nonfinancial firms was identical negative to alarming events, like the Fannie and Freddie takeover or Lehman s implosion. Nevertheless certain events were apparently perceived as beneficial to financial firms yet harmful to nonfinancial firms and vice versa. In order to identify the statistical significance of the above described events, we apply a widely used techniques in social sciences, the ANalysis Of Variance (ANOVA), This approach allow us to eliminate nonstatistically significant events which could adversely affect the results of the analysis. Four different oneway ANOVAs (monofactorial) were conducted by using the following samples: Fair Value and Non Fair Value samples for financial companies Fair Value and Non Fair Value samples for nonfinancial companies Fair Value and Non Fair Value samples for all the companies Fair Value samples of financial firms and Fair Value samples of nonfinancial corporations. The crisis events that did not report any statistical significance are excluded from the analysis. After this process, 54 events were considered in the analysis. Then, we
14 classified the 54 crisis events into three groups according to their nature: Rescue; Distress; Capital injections. The first group consists of events in which financial institutions have been financially helped or bailed out. The second group classifies all financial adverse events that occurred during the period (bank troubles, fraud, ect). The third group includes events in which injections of capital were operated by Central Banks and by the American and European governments. This group includes interest rate cuts from the major Central Banks, liquidity injections, legislative actions and policy announcements, the TARP program, ext. 6. Measuring investors' reaction to crisis events The investors reaction to fair value hierarchy was measured through the concept of mean cumulative abnormal returns (CARs). The use of abnormal returns as the dependent variable is strongly supported by the main academic literature. In this case, the abnormal returns were calculated as follows: (1) Where CARs ij = the mean cumulative returns at the event j for the ith firm; T = the number of trading days covered by the event; R it = the rate of return at day t of the ith firm R Et = the return of the Eurostoxx index at the day t.
15 7. Our Hypotheses Nonfinancial companies The fair value assets of nonfinancial companies pertain generally to noncore businesses and they are retained in balance sheets exclusively for liquidity and hedging purposes. These assets are highly liquid and mainly classified in levels 1 and 2 of the fair value hierarchy. When liquidityrestrictive events occur, investors tend to require liquid securities, causing a consequent increase in the value of the assets held by nonfinancial companies, according to the phenomenon better known as flight to quality 20. By the contrary, the fair valued liabilities of nonfinancial companies consist mainly of relatively illiquid instruments such as mortgagebacked securities, whose values tend to decrease in constraining liquidity events. Investors reactions to fairvalued liabilities of nonfinancial firms are more complex. For liquidityconstraining events, the value of liabilities will decrease, due to heightened default risk of firms, benefiting shareholders. On the other hand, in a liquidity crunch, firms will be less able to rollover (refinancing risk) their shortterm liabilities, thereby damaging investors. Thus, investors overall reaction in liquidityconstraining events to the liabilities of nonfinancial firms primarily depends on the maturity of the liabilities (Lev & Zhou 1999). If the firm s fairvalued liabilities are primarily shortterm, investors will react negatively (rollover concerns), whereas if most liabilities are longterm and of Levels 1 or 2 quality, investors will react positively (credit risk or default concerns). Due to this ambiguous behavior, we will not consider the impact of fairvalued liabilities of nonfinancial firms. Thus: Proposition 1: when liquidity constraining events occur, investors will react 21 positively to the levels 1 and 2 fair value assets and will not react (or only mildly and negatively) to Level 3 fair value assets 20 Lev, B. and Zhou N. (2009), Unintended Consequence: Fair Value Accounting Informs on Liquidity Risk, forthcoming; Acharya V.V. and Richardson M. (2009), Causes of the financial crisis, Critical Review Vol. 21, Nos. 2 3, pp We have to remember that we assess the investors reaction over the abnormal returns on companies s share prices.
16 According to this statement, the stock market prices of nonfinancial companies should behave in the opposite way when positive events occur, i.e. they would diminish in expansive liquidityevents. Proposition 1bis: when expansive liquidityevents occur, investors will react negatively to levels 1 and 2 of fair value assets and do not react (or only mildly and positively) to Level 3 assets. Financial companies In financial firms, there is a greater concentration of financial instruments classified in levels 2 and 3 (such as CDOs, MortgageBacked Securities ) of the fair value hierarchy than the latter, as shown in the descriptive statistics in the Appendix. Level 3 assets are more illiquid and present higher levels of information asymmetry than level 1 and level 2 assets. Liquidity constraining events would have, therefore, a negative effect on the stock prices of financial firms. More illiquid the assets of financial firms, the greater investors negative reaction is. The following hypothesis is reached: Proposition 2: when liquidityconstraining events occur, investors will react negatively to fair value assets and more strongly to those classified in level 3. Clearly, the opposite is true for positive events. Proposition 2bis: when liquidityexpanding events occur, financial company investors react positively to fair value assets and more strongly to those classified in level 3. Financial companies liabilities are less illiquid than their assets. Further, there is a large government insurance on customers deposits. Because of these two considerations, the following statement must be considered: Proposition 3: when positive and negative liquidity events occur, investors do not react to the level 1 and 2 fair value liabilities and react mildly to level 3 liabilities..
17 8. The relevance of Fair Value Hierarchy on European Companies stock returns 8.1. The behavior of financial firms The results of the empirical analysis for financial companies are shown in the appendix (tables 17, 20 and 23). Rescue group The hypothesis supporting the influence on stock prices of financial companies fair value assets during rescue events (assumption 2bis), appears to not be confirmed. In fact a negative and highly significant (99%) pattern 22 in is reported in L3A for the rescue group. This is confirmed by the result obtained by the control regressions which show a large number of significant coefficients which present the same pattern. In particular, the OLS calculated with the use of a dummy finds significant coefficients for L1L and L3L with an Fstatistic significant at 99%. This is the opposite of what one might expect, given that the bank bailouts should be expansive liquidity events. The OLS regression thus confirm the negative signs of the coefficients and leads us to state that the bank bailouts are events with specific characteristics. These features do not allow us to assimilate them conceptually to the other expandingliquidity events. A possible explanation for the aforementioned results may reside in a negative perception of the investor's "rescue risk/risk of nationalization" which is strongly felt in banks with substantial exposure to Level 3 assets. In fact, the stock prices of these banks could "include the fact of being the next to be saved and have to undergo governmental interference in their management. Attention must be brought to the fact that the financial sector tends to be somewhat opaque and that a greater control by the savior state is not the as good as a bank might expect. This is confirmed by the fact that in the rescue group there are no significant values for nonfinancial firms. Level 3 Liabilities shows a negative coefficient in each of the performed regressions. As mentioned above, this result could be due to investors perception of a greater level of 22 We remember that negative pattern is defined as a decrease from level 1 to level 3 of the values of the coefficients and, similarly, positive pattern is a growth between the two values.
18 "risk of nationalization" for banks with a liability of Level 3, often without insurance. In the regression with the control variables the leverage coefficient is negative and highly significant. Thus, the investors negative reaction, in the presence of bank bailouts, is stronger for heavily indebted institutions. This supports the belief of the existence of a "rescue risk" described above. Distress group The events that make up this group are all liquidityconstraining and hence, as mentioned above, they have negative coefficients. Furthermore, lower levels in the fair value hierarchy (levels 2 and 3) should correspond to lower values in the OLS coefficients calculated above. However, the OLS regression shows positive coefficients for levels 1 and 3 and negative sign for the second level of assets. Clearly, hypothesis 2 bis is not verified and this could be given by the fact that the Distress group reported the lower number of observations than any of the other event groups. In fact statistical significance is not detected with the analysis carried out on this set of events. We can ultimately recognize that there is a negative pattern in the levels of fair value (i.e., going from level 1 to level 3) and this confirms that equity prices have been hampered by the reaction of investors to companies liquidity risk. The three coefficients of liabilities in the OLS regression (L1L, L2L and L3L) drive us to confirm hypothesis 3. In fact, despite the different sign of the coefficients of L1L (which is negative) and L2L (which is positive), the coefficients show absolute values that are much lower than L3L, which as expected, has negative sign. Capital injections group The coefficients generated from the capital injections group confirm hypothesis 2bis. With the OLS regression a positive pattern between the asset levels show positive values for levels 2 and 3. In particular, level 3 has a positive coefficient higher than the level 1 and is significant at 90%. This is confirmed in the regression with control variables. Morevoer, hypothesis 3 is not confirmed. The rescue group and the capital injections group are the most significant (in the dummyregression of the capital
19 injections group L2L and L3L are significant at 95% while in the rescue group L1L and 3 are significant at 90 % and 95% respectively).this leads to the rejection of the idea that financial companies stock prices are not influenced by the levels of fair value liabilities The behavior of nonfinancial companies The results of the empirical analysis for nonfinancial companies are shown in the appendix (tables 18, 21 and 24). Rescue group The bank bailouts do not seem to have an impact on nonfinancial firms. In fact, there were no significant factors except in the case of a level of liabilities that seems to have some negative influence on stock prices. Distress group There were no significant coefficients in any of the performed regressions. However, the coefficient of L3A is mostly negative and higher than that of L1A, partly supporting proposition 1. investors reaction to liquidity constraining events is negative and the stronger it is, the higher the level of the fair value hierarchy (i.e. level 3). The "flight to quality" phenomenon in nonfinancial companies is not confirmed because the coefficient of L1L is negative in the ordinary OLS and positive in that with fixed effects for the distress group. We can notice a negative pattern (between L1A and L3A) for both assets and liabilities in this group. Not being able to benefit from the comparison with the dummy regression coefficients, we can only compare ordinary OLS coefficients with OLS fixed effects coefficients. Even here we can see that investors are not indifferent to the assets assessed at fair value because the coefficients of L3A are relevant and negative. In the regression with control variables, the coefficients of L1A and L2A, although not statistically significant, are positive and lead us to hypothesize a positive reaction of investors to such assets, as proposed in Proposition 1. As for liabilities, the signs of the regression coefficients lead to the belief that investors have not experienced a refinancing risk for nonfinancial firms.
20 Capital injections group The dummyregression and the regression with fixed effects support hypothesis 1bis due to the negative sign of the coefficients of L1A and L2A. When liquidityexpanding events occur, the fair value coefficients show an increasing trend from level 1 to level 3 of fair value assets. We cannot investigate liabilities due to their aforementioned ambiguous behavior when events which restrict global liquidity occur. Similarly to Lev and Zhou (2009), both the pattern and the positive coefficients of L2L and L3L for the rescue group suggest that the positive reaction resulting from improved credit risk (default) may have prevailed over the negative risk of refinancing (rollover). This is not followed by a similar pattern of the levels of liabilities of the capital injections group, which also contains expansionary liquidity events Robustness checks Regressions with control variables We replicate all the regression analysis reported above using control variables. We believe important to determine what happens if control variables are included into the analysis to anchor the strong negative trends in stock prices during the financial crisis to some key financial information. This analysis is intended to strengthen what has been previously found, although a lower statistical significance should be expected as some information is captured by the control regressors. According to the previous literature, the control variables used are: 1. EBIT; 2. ROA; 3. NET CASH FLOW FROM OPERATING ACTIVITIES; 4. LEVERAGE By analyzing the results reported in the appendix (table 20 and 21), we can confirm the previous considerations illustrated in section 8.2.