A review of the pre-adoption market reaction to the application of IFRS 9 across European Industries

(1)

A review of the pre-adoption market

reaction to the application of IFRS 9 across

European Industries

Rick Weening Rijksuniversiteit Groningen p.weening@student.rug.nl 06- 46441246 June 13, 2016

(2)

As of 01-01-2018, IFRS 9 will be the new accounting standard used to account for financial instruments. This study assesses the general opinion of investors regarding the application of the new standard. I do this by observing the market reaction across varying industries within Europe in response to pre-adoption news events about the new accounting standard over a period ranging from 15 July 2009 till 1 January 2015.

Between 1995 and 2012, the total amount of financial derivatives has increased by a

staggering 1700%. (Abdel-khalik & Chen, 2015). This immense increase went hand in hand with the development of new products and instruments, which caused the derivative market to become more divers on the one hand, but more complex on the other hand.

This greater diversity in products led to an increase in the number of firms that adopted a derivative programme. Because of the greater variety in products it was made easier to apply it to their firm specific needs. Nowadays derivatives and other financial instruments are used by a great deal of companies, varying across all sorts of industries. A vast amount of research has already been done on how and why those companies use derivatives. (e.g. Nance et al. 1993; Geczy et al., 1997; Guay, 1998; Bartram et al., 2009; Bartram et al., 2011 ). The general accepted outcome of these articles is that most of the companies use financial derivatives to hedge certain risks. However, some companies also use them to speculate on certain expectations.

If derivatives would be mostly used to mitigate risks, how then could it be that they’ve played such a crucial role in the recent economic downturn starting in 2008? One of the main

reasons was the sheer complexity of the financial derivatives and other financial instruments that were used at the time. Gorton (2009) shows how this complexity caused asymmetric information amongst users of the derivatives. This eventually led to a situation where risk was spread in such an opaque way, that it eventually led to a banking panic.

But not only the complexity of derivatives and other financial instruments itself, but also the way they were accounted for, contributed to the escalation that lead to the crisis. The applicable accounting standards for financial instruments in the period preceding the financial crisis were IAS 39: Financial Instruments: Recognition and Measurement in Europe and FAS 133 in the U.S. Both have been at the centre of heated debates over the recent years.

(4)

Therefore, in response to the pressure following the 2008 crisis, the IASB announced that the used accounting standard to account for financial instruments was going to be revised. This eventually led to the announcement of IFRS 9.

Next to the perceived benefits of implementing the new standard to account for financial instruments, it might also come with its potential costs and disadvantages. Therefore the question remains: what do the users of the financial statements think of this new

standard? Do they perceive the change to be positive or negative? Some of the most important users of a firms financial statements are its shareholders and potential investors. And while the new standard might lead to more transparency and understanding with regards to the usage of financial instruments within a company, the implementation of the new standard might also incur significant costs which could lower the company’s profits. Since the shareholders’ (and potential investors’) general opinion with regards to the

advantages and disadvantages of the new standard will be reflected in the firms’ share price, I will assess this opinion by observing movements in the share prices.

This study will add to the existing literature in the field of accounting and finance in various ways. First of all I expands on previous studies done by Armstrong et al. (2010) and Onali and Ginesti (2014). They’ve looked at the cross country market reaction with regards to the compulsory implementation of IFRS and the reaction to the possible implementation of IFRS 9, respectively. This study will add to their results by examining a cross-industry view, instead of a cross-country one. This approach is potentially relevant, since different industries might have different motives to use derivatives, use derivatives and other financial

instruments on a different scale and may foresee different implications with regards to the implementation of IFRS 9.

Also, this study covers an extended time period when compared to the previously mentioned studies. This might be particularly interesting because, the previous mandatory implication date of IFRS 9 was set to be the 1st_{of January, 2015. This implication date was}

later postponed to 1-1-2018. This event hasn’t been accounted for in the previous studies. Those two aspects will increase our understanding of the possible impact of the

implementation of a new accounting rule such as IFRS 9.

Furthermore, this study adds to existing literature done by Panaanen et al. (2012) and Fiechter (2011). Panaanen et al. (2012) expect that the amendment of IAS 39 adversely influences market pricing of banks’ accounting numbers. Fiechter (2011) on the other hand finds that return on equity (ROE) increases significantly because of the amendment of IAS 39, which suggests that there should be an increase in market prices of banks. Results from this study

(5)

could underpin/undermine the expectations from those studies based on the market reaction following the studied events. Also this effect can be studied across sectors, other than banks.

Moreover, this study can add to the existing literature within the decision usefulness approach to accounting statements. This theory addresses the preparation of financial accounting information and how this information complements the nature and type of

information that investors need to make sound investment decisions. This is also described by Richardson et al. (2010). They state that investors use information from the financial

statements to forecast future earnings on the reporting entity, estimate the risk of these earnings and make an assessment of the intrinsic value of the firm. Investors can use the increased amount of information from the financial statements to improve those estimations. The perceived usefulness of IFRS 9 for investors can be derived from the market reaction following the events identified in this study. We then can also compare the usefulness and impact of IFRS 9 across different industries. IFRS 9 should lead to more transparent and more useful financial statements, therefor also decrease the information asymmetry experienced by investors.

2. Literature review

IAS and IFRS can be determined as standards that are mostly “principle-based” instead of “rules-based”. “Principle-based” standards refer to fundamental understandings that inform transactions and economic events. (Carmona and Trombetta, 2008). This means that not every issue is separately focused on. Instead, the underlying principles should help firms to make the right accounting choices. Such an approach is way more flexible than a “rule-based” approach and is particularly useful in complex situations. Given the complex nature and varying usage of financial instruments and derivatives in particular, a principle based approach would also be helpful when accounting for such products. Surprisingly the accounting standard used in Europe prior to 2008 (IAS 39), was said to be fairly rule-based instead of principle based. Also, it was being too complicated and had to many exceptions (Triana, 2007; Lopes, 2007). It led to confusion amongst stakeholders, instead of what it intended: clarification with regards to reporting on financial derivatives. Therefore in 2008 IASB launched “project IFRS 9” to replace IAS 39. Eventually this should result in a completely new standard regarding how to account for financial derivatives. Clearly, replacing IAS 39 with IFRS 9, might have its advantages. However, prior research (EY, 2013; Deloitte 2014) has also shown that the adoption of IFRS 9 will have a significant

(6)

impact on banks and other institutions linked to the financial sector. According to EY, the changes will have a significant impact on, not only financial statement presentation, but also on their information systems and processes. They also state that it may even influence

business decisions, such as what financial products to offer.

2.1 Implemenation steps

The replacement of IAS 39 can be divided into three phases. Thos phases are: Classification and Measurement, Impairment, and Hedge Accounting. The first phase, Classification and Measurement, defines how financial assets in general should be classified and how they are measured on a continuous basis. The classification and measurement of financial assets also determines the way they should be accounted for. Furthermore, the requirements for hedge accounting and impairment are based on the classification. figure 1 shows how financial assets were classified under IAS 39 (Fiechter, 2011). Basically financial assets could be classified into one of four groups: Loans and Receivables (LAR), Held-to-maturity (HTM) financial assets, Available-for-Sale (AFS) financial assets, and financial assets Held-for-trading. The type of classification then stated which accounting method should be applied to the financial asset. Loans and Receivables, and Held-to-maturity financial assets are

recognized at amortized costs. Available- for-sale financial assets are reported at fair value with adjustments being made by using the other comprehensive income account. Financial assets Held-for-trading are measured at Fair Value where changes in this value go through the profit and loss account (FVTPL).

This last method can also be applied in some certain cases, when a company prefers to report certain financial instruments at fair value, instead of using one of the other accounting

(7)

IAS 39 was largely rule-based, difficult to apply and was also rather complex when it came to the reclassification of financial assets. Opposed to IAS 39, IFRS 9 applies a different

classification approach to all financial assets. This approach is more principle based, consists of only one impairment model and classifications, as well as reclassifications are based on the company’s business model and the nature of the cash flows of the financial asset. Pounder (2009) describes this new classification approach. If, in accordance with the entity’s business model, the financial asset has the objective to be sold with a gain prior to their maturity, this financial asset should be measured at fair value through other comprehensive income, if its cash flows describe a SPPI (Solely Payment of Principal and Interest). If the financial asset is meant to collect all its contractual cash flows and its cash flows also resemble SPPI, then the financial asset should be measured at amortized cost (with a fair value option under limited circumstances). If the financial asset doesn’t resemble the payment of a principle and interest then it should still be measured at fair value, but this time changes in value will be accounted for through the profit and loss. By using this method of classification and measurement, the use of fair value accounting (FVA) will be increased, compared to IAS 39.

Figure 1 Classification of financial assets under IAS 39 (Fiechter, 2011).

Under IAS 39 financial assets could be classified into one of the four categories: Loans and Receivables, Held-to-Maturity assets, Available-for-Sale assets, and assets measured at fair value with adjustments going through the profit and loss account.

(8)

The second phase regards the impairment of financial assets. In their project summary with regards to IFRS 9, the IASB states the following: “As part of IFRS 9 the IASB has introduced a new, expected loss impairment model that will require more timely recognition of expected credit losses. Specifically, the new standard requires entities to account for expected credit losses from when financial instruments are first recognised and it lowers the threshold for recognition of full lifetime expected losses. … in addition, under IFRS 9 the same impairment model is applied to all financial instruments that are subject to impairment accounting, removing a major source of current complexity.” Basically what this means, is that they will replace the current “incurred loss approach” with an “expected loss approach”. As a result, financial instruments will represent their fair value when reported and the expected credit losses on financial instruments have to be recognized and updated at all times to reflect changes in their values. This change in the way of addressing impairment results from the fact that, during the financial crisis, the delayed recognition of (credit) losses on loans was used to postpone losses, and in that way applied a sort of earnings management. The new “expected loss approach” has been the centre of a lot of recent discussions. Opponents of the new loss approach argue that it will lead to an enormous increase of the loan loss provisions amongst banks and other financials. This consequence is also identified by Deloitte

(Deloitte, 2014) stating that, under IFRS 9, impairments will be recognized much earlier. According to research done by analysts at Barclays, the new loss approach could lead to an addition of approximately €61.4bn to the loan loss provisions for Europe’s leading banks. Moreover, the new approach could lead to “banks to overestimate losses during severe downturns, increasing earnings volatility”. This also fits the results by Song (2014), who argues that such a method might lead to increased market volatility due to the “procyclical

(9)

nature” of fair value accounting. That is, “highs” and “lows” of the market will be magnified by using FVA. She explains this by the fact that, under FVA, write ups permit companies to increase their leverage during market “highs”. During market “lows” on the other hand, write downs under FVA magnify losses and increase financial stress.

The third phase within the project of replacing IAS 39 concerns hedge accounting. The main criticism with regards to Hedge Accounting under IAS 39 was that a company’s risk

management activities weren’t reflected enough in their financial statements (IASB, 2014). IFRS 9 aims to better reflect those activities and will therefore help investors to better understand the effect of hedging activities on the financial statements and future cash flows. They try to achieve this by replacing the distinction that IAS 39 made between financial and non-financial items (financial items could be hedged, non-financial items could not), by a more principal based approach under IFRS 9. IFRS 9 solely looks at whether a risk

component can be identified and measured. Therefore more entities will be enabled to use hedge accounting and better reflect their risk management activities (IASB, 2014).

2.2 Fair value accounting

The replacement of IAS 39 with IFRS 9 will increase the amount of FVA when it comes to accounting for financial instruments. A lot of research has already been done with respect to fair value accounting. For example, Barth et al. (2001). examine how share prices reflect historical costs and fair values, and use their evidence to inform standard setters to what degree they thinks fair value accounting should be used when setting new standards. In their article they describe several studies that have been done with regards to FVA. They consistently find that investors perceive fair value estimates more value relevant than historical cost amounts. Value relevance is described by Barth et al. (2001) as: Whether particular accounting amounts reflect information that is used by investors in valuing firms’ equity. Hitz (2007) also studied the decision usefulness of FVA. He did this by using two different approaches, the valuation perspective and the information perspective. According to the view of the valuation perspective accounting should directly report on essential

information needed by investors to make proper valuations. This view shows similarities with the definition given by Barth et al. (2001). The information approach on the other hand,

(10)

assumes a broader definition. Under this view, useful information is defined as a signal which influences beliefs about, in this case, stocks. Hitz (2007) finds that the decision usefulness of FVA is supported under both approaches. Based on this studies we may imply that fair value accounting will better inform investors, so that they can make better investment decisions. IFRS 9 aims, amongst other things, to extend the use of fair value accounting (FVA) for financial instruments (Palea, 2014). The associated advantages of this extension of the use of FVA is also recognized by the Securities and Exchange Commission (SEC). They observe two major advantages of using FVA to report on financial instruments. First of all it would reduce the amount of transactions which were solely done to exploit opportunities created by the mixed system (partially historical cost, partially fair values) to generate managerial benefits. Secondly, FVA of financial instruments would lead to a reduction in the complexity with regards to financial reporting. But as stated before, Song (2014) on the other hand, shows that FVA might negatively affect the value relevance of accounting reports due to the increased market volatility caused by FVA.

2.3 Efficient markets

But how is this regulatory change perceived by different stakeholders, and investors in

particular? Do they believe the advantages of replacing IAS 39 with IFRS 9 (and therefore an increase of the amount of FVA when it comes to accounting for financial instruments)

outweigh its disadvantages? Or do they believe the opposite? A widely used method to estimate the general attitude of investors towards some sort of change or event is by examining stock prices and stock returns directly following the change. The theory behind this method lies in the “efficient market hypothesis” as developed by Fama (1970). He states that a market is efficient when market prices always “fully reflect” all available information. He makes a distinction between 3 forms of market efficiency. The weak form of market efficiency examines the forecast power of past returns, the semi-efficient market examines how quickly security prices reflect public information and, lastly, the strong form of efficient markets observes whether any investors have private information that is not incorporated in the market prices. Since I will be examining the response of investors with regards to news items (public information) related to the application of IFRS 9, I further assume the market to

(11)

be semi-efficient. This method of identifying events and testing their stock market reaction is also adopted by Armstrong et al. (2011) and Onali and Ginesti (2014). In other words, by using this method I assume all new information with regards to the application of IFRS 9 and the investors opinion regarding this new information will be directly reflected in the stock prices. The assumption of semi-efficient markets is also widely used in previous research e.g. Ball (1972).

2.4 Information usefulness

However, the degree to which new information has an impact on the current stock prices doesn’t solely rely on how quickly investors can interpret new information but also on to what extent the new information is useful to them. Over the years a lot of research has been conducted with regards to the usefulness of accounting information. Holthausen & Watts (2001) give an overview of the research that is done over the past years with regards to this topic, and present a critical assessment of the research specifically done on value relevance in the context of standard setting. Kothari (2001) examines the relationship between capital markets and financial statements. He states that, in an efficient capital market, switching from one accounting method to another without direct cash flow effect, a signaling effect, or incentive consequence should not affect security prices. But, according to Kothari (2001), when assuming the market to be semi-efficient instead of efficient, the opposite is true. 3. Hypothesis development

As shown in the previous section, there has already been done a great deal of research with regards to the perceived usefulness of applying IFRS and IFRS 9 in particular. One of the widely used arguments in favor of IFRS, is that it would lead to greater transparency, less complexity and a lowering of asymmetric information. This might increase the usefulness of the financial statements for investors. Also, as mentioned before, IFRS 9 will enable more entities to apply hedge accounting and therefore better reflect their risk management activities. On the other hand one could state that the application of IFRS 9 could lead to greater earnings volatility, due to the ”pro-cyclical nature” of FVA, making it harder for investors to make sound estimations about future earnings.

It seems clear that applying IFRS 9 might have potential benefits, but also some downsides as well. The objective of this study will be to examine the general opinion of investors with regards to the replacement of IAS 39 with IFRS 9 across different industries within Europe.

(12)

That way we may obtain a better understanding of the (potential) impact of the new accounting standard. I will do this by looking at the market reaction across different industries as a reaction to news events with regards to the application of IFRS 9.

Taking the information aggregation hypothesis (Hitz, 2007), where the consensus expectations of investors are reflected directly in the market, into account, this research objective can be translated into the following hypothesis:

H1: New information on the content and (mandatory) application of IFRS 9 will result in significant market reactions in each identified market sector.

As stated before, the impact on banks and other financial institutions is expected to be of a significant (negative) impact. Therefore, I also expect to find evidence for the following two hypothesis.

H2: a significant downward market reaction for the financial sector in specific can be observed when news with regards to the content and application of IFRS 9 is published.

H3: The market reaction for the financial sector differs significantly from that of the other sectors.

4. Methodology 4.1 Events

I examine the investor perceptions with regards to the application of IFRS 9 by observing equity return reaction to the relevant identified events. Following Ognali & Ginesti (2014) an event is identified as an “official announcements and initiatives by the IASB and European Financial Reporting Advisory Group (EFRAG)”. The reason for this is, just as mentioned by Ognali & Ginesti, that these organizations have the strongest influence on the public debate with regards to regulatory accounting changes. The identified events are depicted in

Appendix A and are collected over the time period ranging from 15 July 2009 (the day after the IASB proposed to improve accounting with regards to financial instruments) until 1 January 2015 (the application date of IFRS 9 as proposed by the IASB on 4-6-2011). The period ranging from 15 July 2009 until 31 December 2012, overlaps with the time period used by Ognali & Ginesti (2014). Since we both use the same method to identify and assess

(13)

our events, I identify the same events compared to their paper during this period. Initially 31 events were identified in total.

I further assess those events by looking at their overall media coverage1_{, thus creating a proxy}

for their relevance in the public debate with regards to the application of IFRS 9. I also checked whether there were any confounding events on the event dates that might prejudice the results2_{. I only identify 2 other events that may have had a significant impact on stock}

returns that day. The first confounding event took place simultaneously with the publicatio n of the revised proposal for loan-loss-provisioning by the IASB on 7-3-2013. On the same day a remark with regards to the interest rate by Mario Draghi (President of the European Central Bank) led to a short rally for the euro. But the same article also states that the announcement was already accounted for by investors. Therefore, I believe the impact to be minimal. The other confounding event occurred on 15-4-2013, when a bombing took place at the Boston Marathon. This event may have led to some turmoil on stock exchanges all around the globe. However, I don’t expect it to have any implications with regards to my results.

I also determine whether the event will increase (decrease) the likelihood of IFRS 9 to become effective in line with earlier announcements. In this I follow the method used by Onali & Ginesti (2014), where the returns used when calculating the abnormal returns related to an event decreasing the likelihood of adoption of IFRS 9 are multiplied by -1.

Since the development and applicatio n of IFRS 9 is a process that took several years, I will look at the overall market reactions corresponding to all identified events (instead of single events) when interpreting the results. In calculating this overall market reaction I excluded events that haven’t had significant media coverage. The calculated results are based on the three-day abnormal returns for weighted industry portfolios consisting of all listed firms across Europe for the 31 identified events.

4.2 Industry returns

1_{Following the same method as Ognali & Ginesti (2014) I check the following news sources: Financial Times, Bloomberg, Reuters, Wall Street Journal,} CFO.com. I only exclude IFRS.com from my sample, since this is the official site of the IASB. Clearly, all their announcements are stated there, making it non -applicable when checking for overall media coverage. To search for related articles/publications I use the keywords: “ IFRS 9”, “ IAS 39”, “ IAS 39 refor m”, “ Adjustment IAS 39”, “ IASB accounting reform”, “ accounting for financial instruments”, “ derivatives accounting”.

2_{I checked for confounding events by looking at news items released on the same day, or a day prior to the event date. For thi s I used the website} www.marketwatch.com , since it gives a good overview of general financial news, as well as stock specific news. General event s that might have a relevant impact are, e.g. ECB announcements, Job reports, etc.

(14)

To construct the weighted industry portfolios I first retrieved all listed firms in Europe. The initial sample of listed firms in Europe comprises 9206 firms. The number of listed

companies was retrieved from Orbis, by selecting all listed public firms within the regions Western- Europe and Scandinavia. Thereafter I excluded companies from Cyprus, Iceland, Malta and Turkey, San Marino, Liechtenstein, Gibraltar, Monaco and Andorra, companies who had no corresponding ISIN-code and companies who were no longer active as public companies as of 20th_{of april 2016, leaving 5393 firms in the sample. I grouped firms in}

industry portfolio’s based on their NACE rev. 2 code, which is used to classify economic activity within Europe. The NACE rev. 2 distinguishes 21 different classes (see Table 1). An overview of the distribution per industry is given in Table 23_.

(15)

Table 1 NACE classes

This table presents the different NACE rev. 2 codes and their corresponding industry classes used to classify economic activity within Europe.

Class Division

Code

Description

A 01-03 agriculture, forestry and fishing

B 05-09 mining and quarrying

C 10-33 manufacturing

D 35 electricity, gas, steam and air conditioning supply

E 36-39 water supply, sewerage, waste management and remediation activities

F 41-43 construction

G 45-47 wholesale and retail trade; repair of motor vehicles and motorcycles

H 49-53 transportation and storage

I 55-56 accommodation and food service activities

J 58-63 information and communication

K 64-66 financial and insurance activities

L 68 real estate activities

M 69-75 professional, scientific and technical activities

N 77-82 administrative and support service activities

O 84 public administration and defence; compulsory social security

P 85 education

Q 86-88 human health and social work activities

R 90-93 arts, entertainment and recreation

S 94-96 other service activities

T 97-98

activities of households as employers; undifferentiated goods- and services-producing activities of households for own use

(16)

Table 2 Firms per industry

This table gives an overview of the distribution of firms retrieved from Orbis over the various NACE rev. 2 industry classes.

Class Description Number of firms

A agriculture, forestry and fishing ₅₁

B mining and quarrying 231

C manufacturing ₁₆₈₄

D electricity, gas, steam and air conditioning supplyelectricity, gas, steam and air conditioning supply

104 E water supply, sewerage, waste management

and remediation activities ₃₁

F construction ₁₃₀

G wholesale and retail trade; repair of motor

vehicles and motorcycles 335

H transportation and storage 141

I accommodation and food service activities

65

J information and communication ₅₈₈

K financial and insurance activities 1009

L real estate activities 267

M professional, scientific and technical activities 384

N administrative and support service activities 145

O public administration and defence; compulsory

social security ₄

P education 8

Q human health and social work activities

58

R arts, entertainment and recreation ₆₃

S other service activities 107

T activities of households as employers;

undifferentiated goods- and services-producing

activities of households for own use -

U activities of extraterritorial organisations and

bodies -

Total 5393

Next, I calculated each firms weight in their respective industry portfolio by dividing the firms market value by the total market value of the portfolio. The firms’ market value was calculated by first multiplying the daily closing price of the stock with the number of outstanding shares (over an interval ranging from 1-1-2009 till 1-10-2014) and thereafter calculating the average of these calculations. Both shares outstanding, as well as the daily closing price were retrieved from the WRDS Compustat IQ database by using each firms’ individual ISIN-code. The company weights were used to calculate the market weighted portfolio returns. These returns were calculated by multiplying each individual firms’ log-normal daily return with its respective weight within the portfolio and then summing those results together.

(17)

To estimate the market return, the iShares MSCI Europe UCITS ETF was used as a proxy for the European market. Historical data regarding the ETF were retrieved from the Ishares website4_{. I then modified the data to fit the time frame that is examined in this study.}

The risk free rate was proxied by using the yield on 10-year german government bonds. Those were retrieved from Bloomberg in accordance with the same time frame.

4.3 Capital Asset Pricing Model

The returns for each industry portfolio to each of the corresponding events are estimated by making use of the Capital Asset Pricing Model where the excess weighted lognormal industry return, 𝑅_𝑖,𝑡, over the risk free rate, 𝑟_𝑓, is calculated as:

𝑅_𝑖,𝑡− 𝑟_𝑓 = 𝛽_𝑖,𝑚(𝑅_𝑚,𝑡 − 𝑟_𝑓)

Where 𝑅_𝑚,𝑡 is the return on the market, i denotes the portfolio and t denotes the date. The expected portfolio returns, 𝐸(𝑅_{𝑖,𝑒,𝑡}) during the day of the event, as well as the day prior and following the event are then estimated by multiplying the estimated beta for that event (calculated over a 200 day period ending 10 days before the event), 𝛽_{𝑖,𝑒,𝑚}, with the excess market return, (𝑅_𝑚,𝑡− 𝑟_𝑓) for these days, plus the risk free rate for the same days:

𝐸(𝑅_𝑖,𝑒) = 𝛽_{𝑖,𝑒,𝑚}(𝑅_𝑚,𝑡 − 𝑟_𝑓) + 𝑟_𝑓

Where 𝐸(𝑅_{𝑖,𝑒,𝑡}) depicts the estimated portfolio return for that day and e stands for the corresponding event we are observing.

These expected portfolio returns are then used to calculate the abnormal returns within the estimation period. Abnormal returns, 𝐴𝑅_𝑖,𝑒, are calculated as

𝐴𝑅_{𝑖 ,𝑒,𝑡} = 𝑅_𝑖,𝑡 − 𝐸(𝑅_{𝑖,𝑒,𝑡}) Where 𝑅_𝑖,𝑡 stands for the actual portfolio return on that day.

4_{Retrieved on 11-5-2016 from https://www.ishares.com/uk/individual/en/products/251860/ishares} -msci-europe-ucits-etf-inc-fund?siteEntryPassthrough=true&locale=en_GB&userType=individual

(18)

Then, since I am making use of a multiday event window [-1,1], Cumulative abnormal returns (CARs) from day 𝜏1 to 𝜏2, which depicts the event window, for each portfolio are calculated as:

𝐶𝐴𝑅_{𝑖,𝑒,𝜏1,𝜏2}= ∑ 𝐴𝑅_{𝑖 ,𝑒,𝑡}

𝜏2 𝑡= 𝜏1

Since we are interested in the market reaction corresponding to all identified significant events, instead of single events, I thereafter calculated the Cumulative average abnormal return (CAAR) for each portfolio over all significant events. The Cumulative Average Abnormal Return, 𝐶𝐴𝐴𝑅_𝑖, is calculated as

𝐶𝐴𝐴𝑅_𝑖 = ∑ 𝐶𝐴𝑅𝑖,𝑒,𝜏1,𝜏2 3 5. Data

5.1 Return characteristics

The average and standard deviation of returns for the industry portfolios, as well as the risk free rate and the market return, are listed in Table 3. The averages, as well as the standard deviations, are calculated over the period ranging from 01-01-2009 till 01-10-2014. Table 3 Portfolio returns 01-01-2009/01-10-2014

This table states the mean return as well as the standard deviation for each industry portfolio, the market return and the risk free rate, over a period ranging from 01-01-2009 till 01-10-2014.

Portfolio Mean Stand. Dev.

A 0,0015 0,0296 B 0,0001 0,0135 C 0,0003 0,0094 D -0,0000 0,0284 E -0,0000 0,0110 F 0,0003 0,0129 G 0,0001 0,0141 H 0,0004 0,0102 I 0,0004 0,0110 J 0,0002 0,0113 K 0,0019 0,0499 L 0,0002 0,0093

(19)

Table 3 continued

Portfolio Mean Stand. Dev.

M 0,0003 0,0087 N -0,0001 0,0170 O -0,0002 0,0098 P 0,0001 0,0149 Q 0,0001 0,0071 R 0,0003 0,0099 S 0,0004 0,0117 Market Return 0,0018 0,0176

Risk free rate 0,0224 0,0078

I also performed a Durbin-Watson test over the complete time period, to test whether the industry portfolios were affected by autocorrelation. The Durbin-Watson test-statistic ranges between 0 and 4 where a test statistic of 2 indicates no autocorrelation whatsoever. A value lower than 2 indicates negative autocorrelation, while a value higher than 2 indicates positive autocorrelation. The results are depicted in table 4.

Table 4 Durbin Watson test statistics

This table gives an overview of the Durbin Watson test statistic per portfolio to control for autocorrelation.

Portfolio Durbin Watson test

statistic A 1,977996 B 1,883774 C 2,338913 D 2,50178 E 1,808597 F 1,774764 G 1,93909 H 1,910674 I 1,703355 J 2,358086 K 1,09024 L 2,006621 M 1,567569 N 1,900835 O 1,940526 P 1,804866 Q 1,793838 R 1,914422 S 2,657891 Max. 2,657891 Min. 1,09024

(20)

As can be seen, generally the portfolios aren’t really affected by autocorrelation. Only portfolio K, Financial and Insurance activities, could be considered to be moderately

affected by negative autocorrelation, since this test statistics differs from 2 the most (Durbin-Watson test statistic: 1,09024).

5.2 Beta estimation

To estimate the expected returns, I first estimated factor betas for each industry portfolio for each event, over a 200-day period, ending 10 days before the event took place. Table 5 shows the average betas and average adjusted R-squared for all events. Those are calculated as the simple average of the beta and adjusted R-squared for all events, respectively.

Table 5 Average estimated betas per industry

The table presents the average estimated betas per industry, calculated over the 31 identified events. C depicts the constant in an OLS regression. The beta as depicted in the CAPM formula for each industry per event was estimated by running an OLS regression with the portfolio returns as dependent variable and MRF (market return – risk free rate) as the independent variable. The average beta per industry was then calculated by summing all the individual betas per industry per event and dividing it by the number of identified events. The average adjusted R-squared was calculated in a similar fashion. That is, all the individual calculated

Portfolio C MRF Adj. R-squared A -0,0136 0,3377 0,1409 B -0,0110 0,5327 0,4554 C -0,0191 0,1208 0,2055 D -0,0137 0,4146 0,3115 E -0,0152 0,3232 0,2803 F -0,0120 0,4812 0,3827 G -0,0142 0,3912 0,2821 H -0,0148 0,3196 0,3513 I -0,0131 0,4043 0,3941 J -0,0147 0,3468 0,4149 K -0,0106 0,5047 0,3358 L -0,0162 0,2644 0,3315 M -0,0147 0,3450 0,4500 N -0,0147 0,3583 0,3690 O -0,0205 0,0801 0,0623 P -0,0212 0,0229 0,0168 Q -0,0184 0,1636 0,1947 R -0,0172 0,2052 0,1588 S -0,0172 0,2095 0,3185 Average -0,0154 0,3066 0,2872

Looking at table 5 we can state the average explanatory power, as indicated by the average Adjusted R-squared measure, over all events and all portfolios is 0,2872. That is, on average 28.72% of the variability in the portfolio returns can be explained by the model. However,

(21)

this average also includes portfolio O and P, for which the Adjused R-squared measure is especially low. A likely explaination for this is that those portfolios only consist of four and 8 firms, respectively. Excluding those portfolios when calculating the Average Adjusted R-squared would increase the models explanatory power to 0,3163 (31,63%). This, however is still roughly 1/3 of the variability. We should bear this in mind when interpreting the results and drawing conclusions based on those results.

5.3 Normality

After calculating the abnormal returns (ARs) over the estimation window for each portfolio and each event I also tested the abnormal returns for normality. This was done by running a Jarque-Bera test and by testing for skewness and kurtosis. Appendix B gives an overview of the tests. Based on these results there cannot be stated that the abnormal returns are normally distributed. However, since the number of observations included in the tests is sufficiently large (N>50) the central limit theorem can be applied. This theory states that the mean of a sample of data having any (non-normal) distribution converges upon a normal distribution as the sample size becomes large enough. Nevertheless, the non-normality issue is something to bear in mind when interpreting the results.

6. Results

6.1 Overall market reaction

Table 6 presents the Cumulative abnormal average returns for all 31 events. The cumulative abnormal average return is calculated by taking the simple average from the cumulative abnormal returns per industry.

Table 6 Cumulative Average Abnormal Returns per event

This table presents the events that had significant media coverage, the Cumulative average abnormal return for all industry portfolios and its corresponding p-value. The p-value was calculated by running a two-sided t-test, assuming H0: CAAR = 0. The significance level used was 0.05

N Event date Event Description CAAR P-value

1 12-11-2009 IASB issues IFRS 9 (completing the first phase – Classification and Measurement)

-0.0771 (0.0000)*

4 16-7-2010 EFRAG releases the comment letter on the IASB exposure draft – fair value option for financial liabilities

-0.0802 (0.0000)*

6 28-10-2010 IASB issues additions to IFRS 9 for financial liability accounting, completing the

classification and measurement phase

(22)

7 9-12-2010 IASB releases the exposure draft on accounting for hedging activities

-0.0567 (0.0000)*

8 13-1-2011 IASB and FASB publish a joint proposal approach on credit impairment of loans and other financial assets managed in an open portfolio

-0.0429 (0.0000)*

9 31-1-2011 IASB and FASB publish common proposal for accounting for impairment of financial assets such as loans managed in an open portfolio

-0.0659 (0.0000)*

10 4-3-2011 EFRAG recommends that IASB and FASB agree on a joint timetable to finalize

accounting standard for financial instruments

-0.0682 (0.0000)*

12 8-4-2011 EFRAG releases the final comment letter to IASB in response to supplementary document financial instruments: impairment issued on 31 January 2011

0.0799 (0.0000)*

14 4-8-2011 IASB proposes adjustments to effective date of IFRS 9 from January 1, 2013 to January 1, 2015

0.1174 (0.0000)*

16 16-12-2011 IASB releases amendments to IFRS 9 that defer the mandatory effective date from 1 January 2013 to 1 January 2015

-0.0320 (0.0000)*

17 27-1-2012 IASB and FASB inform on the joint intention to reduce differences in classification and measurement models for financial instruments

-0.0173 (0.0000)*

19 7-9-2012 IASB releases draft of forthcoming general hedge accounting requirements that will be added to IFRS 9

-0.0144 (0.0000)*

20 28-11-2012 IASB releases proposal for limited changes to IFRS 9 classification and measurement

requirements

-0.0161 (0.0000)*

21 28-2-2013 IASB publishes proposals for amendments to IAS 39 and IFRS 9

-0.0283 (0.0000)*

22 7-3-2013 IASB publishes revised proposals for loan-loss-provisioning

-0.0230 (0.0000)*

23 22-3-2013 EFRAG's releases its final comment letter on the transition from IAS 39 to IFRS 9 for macro-hedging practices

0.0199 (0.0000)*

25 15-4-2013 EFRAG releases its final comment letter on the IASB's proposal for amendments to IAS 39 and IFRS 9

0.0311 (0.0000)*

26 27-6-2013 IASB provides relief for novation of derivatives

-0.0465 (0.0000)*

27 19-11-2013 IASB completes important steps in reform of financial instruments accounting

(23)

28 17-4-2014 IASB publishes discussion paper on accounting for macro hedging

-0.0363 (0.0000)*

30 24-7-2014 IASB completes reform of financial instruments accounting

-0.0306 (0.0000)*

31 22-8-2014 Membership of Impairment Transition Group Confirmed

-0.0139 (0.0000)*

As can be seen in Table 6 the overall average market reaction (as measured by the simple average of all portfolios) is significant for each of the identified events. Also, what can be noted is that all CAARs have a negative sign, except for the events related to an event that would decrease the likelihood of the adoption of IFRS 9.

6.2 Industry reactions

Table 7 presents the cumulative abnormal average returns for each portfolio calculated over the 22 events that have had a significant amount of media attention. The cumulative abnormal average return is calculated by taking the simple average from the cumulative abnormal returns per industry.

Table 7 CAAR per industry

This table presents the CAAR per portfolio, calculated as the simple average over all events, and its corresponding p-value. The p-value was calculated by running a two-sided t-test, assuming H0: CAAR = 0. A significance level of 95% was used in running these tests.

Portfolio CAAR P-value

A -0.0191 0.1387 B -0.0132 (0.0041)* C -0.0319 (0.0000)* D -0.0169 (0.0343)* E -0.0192 (0.0001)* F -0.0131 (0.0062)* G -0.0210 (0.0040)* H -0.0200 (0.0000)* I -0.0178 (0.0001)* J -0.0200 (0.0000)* K -0.0135 0.0870 L -0.0224 (0.0000)* M -0.0205 (0.0000)* N -0.0218 (0.0013)* O -0.0324 (0.0000)* P -0.0571 (0.0000)* Q -0.0271 (0.0000)*

(24)

As can be seen in Table 7, the CAAR is significant on a 5% level for all portfolios except Portfolio A, Agriculture, forestry and fishing, and portfolio K, Financial and insurance activities.

6.3 Comparing portfolios

As stated in the hypothesis, I expect the reaction from portfolio K, Financial and Insurance activities, to differ significantly from the other portfolios. The results of these tests are depicted in table 8.

Table 8 CAAR portfolio K compared to other portfolios

This table presents the difference between the CAAR of portfolio Z* (where N ranges from A …. S) and the CAAR of portfolio K calculated over all events. The difference is calculated by simply subtracting the CAAR from portfolio K of the CAAR from portfolio Z*. The p-value is calculated by comparing the two portfolios as two independent groups where H0: difference = 0, making use of pooled variance (except for portfolio P. There unequal variance was used, as the F-test showed that variance differed significantly).

Portfolio K Compared to Difference P-value A -0.0113 0.3662 B 0.0006 0.9592 C -0.0159 0.1704 D -0.0047 0.6781 E -0.0066 0.5964 R -0.0234 (0.0000)* S -0.0268 (0.0000)*

(25)

F -0.0009 0.9366 G -0.0078 0.4923 H -0.0079 0.5043 I -0.0046 0.6806 J -0.0085 0.4601 L -0.0099 0.4229 M -0.0076 0.5162 N -0.0083 0.4729 O -0.0175 0.1403 P -0.0340 0.0000 Q -0.0127 0.2945 R -0.0103 0.4098 S -0.0110 0.3461

As can be seen in Table 8, the CAAR for portfolio K only differs significantly from the CAAR of Portfolio P.

6.4 Robustness tests

Since the amount of variability in the dependent variable is also determined by the fit to the data of the used model, I also ran robustness tests. That is, I repeated the same methodology of calculating abnormal returns, but now I’ll be using different models instead. For this, I used one simplified model, the mean return model, and one more sophisticated mode, an adoption of the Fama-French 3 Factor model.

6.4.1 The mean return model

When using the mean return model, Abnormal returns are calculated by taking the actual return of the portfolio on time t and subtracting the mean return of the portfolio. More formally put abnormal returns (𝐴𝑅_𝑖,𝑡) are calculated as:

𝐴𝑅_𝑖,𝑡 = 𝑅_𝑖,𝑡 − 𝑅̅_𝑖

Where 𝑅̅_𝑖 is the mean return for portfolio i, and i depicts the portfolio.

The CAR and CAAR are calculated in a similar fashion as under the method used with the CAPM formula. Table 9 gives an overview of the CAAR per event and its corresponding p-value when using the mean return method instead of the CAPM method.

Table 9 Cumulative Average Abnormal Returns per event using Mean return method

(26)

corresponding p-value. The p-value was calculated by running a two-sided t-test, assuming H0: CAAR = 0. The significance level used was 0.05

Table 10 Cumulative average abnormal return per industry using the Mean return method

This table presents the CAAR per portfolio, calculated as the simple average over all events, using the mean return method, as well as its corresponding p-value. The p-value was calculated by running a two-sided t-test, assuming H0: CAAR = 0. A significance level of 95% was used in running these tests

Event CAAR P-value

1 0,0087 0,0000 4 -0,0163 0,0000 6 0,0036 0,0024 7 0,0009 0,4295 8 -0,0010 0,3968 9 0,0030 0,0100 10 0,0005 0,6369 12 0,0071 0,0000 14 0,0860 0,0000 16 0,0079 0,0000 17 -0,0049 0,0011 19 0,0075 0,0000 20 0,0097 0,0000 21 0,0017 0,0550 22 0,0031 0,0003 23 0,0004 0,6157 25 0,0227 0,0000 26 -0,0027 0,0002 27 -0,0136 0,0000 28 0,0192 0,0000 30 -0,0044 0,0000 31 0,0083 0,0000

(27)

A 0,0075 0,5588 B 0,0117 0,0410 C 0,0029 0,3765 D 0,0113 0,1680 E 0,0118 0,0151 F 0,0115 0,0429 G 0,0062 0,3867 H 0,0091 0,0327 I 0,0087 0,0676 J 0,0095 0,0268 K 0,0070 0,4053 L 0,0083 0,0262 M 0,0085 0,0241 N 0,0077 0,2336 O _0,0050 _0,2666 P -0,0193 0,0054 Q 0,0062 0,0597 R 0,0086 0,0512 S 0,0052 0,1567

6.4.2. The Fama-French 3 factor model

The returns for each industry portfolio to each of the corresponding events are estimated by making use of the of the Fama and French 3-factor model:

𝑅_𝑖,𝑒 = 𝛼_𝑖+ 𝛽_{𝑖 ,𝑚}(𝑅_𝑚,𝑒− 𝑟_𝑓) + 𝛽_{𝑖,𝑆𝑀𝐵}𝑆𝑀𝐵_𝑟,𝑒+ 𝛽_{𝑖,𝐻𝑀𝐿}𝐻𝑀𝐿_𝑟,𝑒+ 𝜀_{𝑖 ,𝑒} Where i denotes the firm and e denotes the event.

The factor (𝑅_𝑚,𝑒 − 𝑟_𝑓) represents the market risk premium, which is the return on the market minus the risk free rate.

The SMB (Small minus Big) portfolio is used to represent the returns that are related to the size of the companies within a portfolio. It consists of the simple average of the three small size portfolios (S/L, S/M, S/H) minus the simple average of the three big size portfolios (B/L, B/M, B/H).

The HML (High minus Low) portfolio is used as a representation of the returns related to the book to market ratio. The book to market ratio is used as a proxy to represent value. It is

(28)

calculated by taking the simple average of the two high BE/ME portfolios (SH, BH) and then subtract the simple average of the two low BE/ME portfolios (S/L, B/L).

The average market value was used as the firms market value in calculating the Fama and French factors SMB (Small-minus-Big) and HML (High- minus-Low). To calculate these factors I retrieved each firms individual book equity value. Since this value can’t be retrieved from Compustat IQ, I used Datastream instead. In Datastream the same ISIN -codes were used as before in the Compustat database. After merging the data of those two databases, 845 companies were excluded when calculating the SMB and HML factors, due to the fact that there was no corresponding data availab le, book-equity value was negative or 0.

Table 11 Number of firms per Fama-French portfolio

This table gives an overview of the number of firms per Fama-French portfolio. The first letter of the portfolio represents the size of the portfolio were S represents a company with a market value lower than the median market value of the sample and B represents a company with a market value higher than the median market value. The second letter represents the book -to-market ratio, were the top 30% is considered high (H), the bottom 30% is considered low (L), and the remaining companies are considered medium (M).

Portfolio Number of firms

S/L 557 S/M 864 S/H 853 B/L 807 B/M 955 B/H 511 Total 4547

Next, I calculated the average market value and book-equity value for each company over a period ranging from 01-01-2009 till 01-10-2014. Thereafter, I used the median of the market values as a cut-off point to determine whether a company could be considered to be big (market value is greater than the median) or small (market value is smaller than the median). Then, the companies where divided over 3 portfolios (High, Medium, Low) based on their book-to-market ratio. This ratio was calculated by dividing the average book-equity value by its average market value. companies with a book-to-market ratio ranging in the top 30% of the total were included in the high (H) portfolio, companies ranging in the bottom 30% were included in the low (L) portfolio. All the remaining firms were included in the Medium portfolio. Combining the size portfolios (based on market value) with the value portfolios (based on book-to-market ratio) resulted in 6 portfolios. For these portfolios the daily total

(29)

returns were calculated by summing all weighted daily returns of each individual firm in the portfolio.

Thereafter the SMB factor was calculated by using the following formulas.:

𝑅𝑆𝑀𝐵= (

𝑅_𝑆𝐿+ 𝑅_𝑆𝑀+ 𝑅_𝑆𝐻

3 ) − (

𝑅_𝐵𝐿 + 𝑅_𝐵𝑀 + 𝑅_𝐵𝐻

3 )

Were 𝑅_𝑆𝑀𝐵 represents the return on the SMB factor, and 𝑅_𝑆𝐿, 𝑅_𝑆𝑀, 𝑅_𝑆𝐻, 𝑅_𝐵𝐿, 𝑅_𝐵𝑀 and 𝑅_𝐵𝐻 represent the return on the corresponding portfolio.

And the HML factor was calculated by using: 𝑅_𝐻𝑀𝐿 = (𝑅𝑠ℎ + 𝑅𝐵𝐻

2 ) − (

𝑅_𝑆𝐿+ 𝑅_𝐻𝐿

2 )

were 𝑅𝐻𝑀𝐿 represents the return on the HML factor and 𝑅𝑠ℎ, 𝑅𝐵𝐻, 𝑅𝑆𝐿 and 𝑅𝐻𝐿 represent the

return on their respective portfolio.

To test to what degree the dependent variables are influenced by each other, I estimated the correlation between the dependent variables. These results are depicted in Table 12. Table 12 Correlation between independent variables

This table shows the correlation between the independent variables MRF, HML and SMB.

MRF HML SMB

MRF 1

HML -0,1017 1

SMB -0,2704 0,9004 1

As can be seen, the correlation between the independent variables is fairly low. Therefore, the SMB portfolio seems to measure a premium for size relatively free of the HML effects. The same goes for the HML portfolio with regards to the size premium.

To estimate the expected returns using the Fama-French 3 factor model, I first estimated factor betas for each industry portfolio for each event, over a 200-day period, ending 10 days before the event took place. Table 13 shows the average betas as well as the average adjusted R-squared for all events, those are calculated as the simple average of all betas and adjusted R-squared for each event.

Table 13 Average Beta Fama French 3 Factor model

This table presents the average beta for the MRF, SMB and HML factor for each industry, as well as its constant (C) and Adjusted R-squared value. The MRF factor is calculated as the return on the market portfolio minus the risk free rate. SMB stands for the return on the Small minus Big portfolio which is calculated by subtracting the return of the Fama French

(30)

portfolio consisting of large firms from the Fama French portfolio consisting of small firms. The HML factor stands for the High minus Low portfolio, where the return of the portfolio consisting of companies with a low Market to book ratio is subtracted from the return on the portfolio consisting of companies with a high market to book ratio. The Adjusted R-Squared depicts the average amount of variability in the portfolio returns that is explained by the model.

Portfolio C MRF SMB HML Adj. R squared A 0.0041 0.1550 -6.9245 3.2294 0.1939 B 0.0072 0.3823 -5.3455 2.7897 0.5614 C 0.0016 0.0717 -1.6879 0.5861 0.3493 D 0.0045 0.2677 -6.0583 3.4788 0.3837 E 0.0047 0.2478 -3.1025 2.0233 0.3102 F 0.0061 0.3250 -5.2361 2.5671 0.4907 G 0.0039 0.2473 -3.0688 -0.6566 0.3531 H 0.0041 0.1978 -3.8870 1.6330 0.4417 I 0.0057 0.2799 -4.0697 1.7827 0.4653 J 0.0046 0.2452 -3.3175 1.4356 0.5066 K 0.0073 0.3485 -5.4251 2.7832 0.4101 L 0.0033 0.1682 -3.0692 1.4180 0.4041 M 0.0044 0.2281 -3.9188 1.8804 0.5889 N 0.0045 0.2458 -3.4613 1.5878 0.4477 O 0.0010 0.0693 -0.1814 0.0290 0.0144 P -0.0005 -0.0285 -1.8378 0.8072 0.0080 Q 0.0016 0.0866 -2.1294 0.8218 0.2066 R 0.0025 0.1127 -2.6895 0.9436 0.1675 S 0.0031 0.1427 -2.2563 1.1623 0.3556 Average 0.0039 0.1996 -3.5614 1.5948 0.3505

As can be seen in Table 13, on average, the model predicts only 35.05% of the variability in portfolio returns. This result, however may be misleading, since the low values for the Adjusted R-squared for portfolio O and P greatly reduce the average. Excluding those two from the average, would lead to an average Adjusted R. Squared value of 0,3904 (39,04%) for all events. Thus, for all portfolios (when excluding portfolio O and P) the model would, on average, explain 39,04% of the variability in the portfolio returns.

Table 14 represents that CAAR per event for all portfolios, calculated using the Fama-French 3 factor model. Table 15 presents the CAAR per industry calculated over all the events using the Fama-French method.

Table 14 Cumulative Average Abnormal Returns per event using the Fama-French 3 factor model

(31)

and its corresponding p-value. The p-value was calculated by running a two-sided t-test, assuming H0: CAAR = 0. The significance level used was 0.05

Event CAAR P-value

1 0,0010 0,0087 4 -0,0018 0,0138 6 -0,0006 0,4138 7 -0,0004 0,5791 8 0,0053 0,0000 9 0,0086 0,0000 10 0,0033 0,0000 12 -0,0101 0,8933 14 0,0278 0,0000 16 -0,0031 0,0000 17 0,0008 0,2477 19 0,0038 0,0000 20 0,0057 0,0000 21 -0,0052 0,0000 22 0,0034 0,0000 23 -0,0032 0,0000 25 -0,0048 0,8942 26 -0,0029 0,0000 27 -0,0090 0,0000 28 0,0127 0,0000 30 0,0029 0,0000 31 0,0565 0,0000

(32)

Table 15 Cumulative average abnormal return per industry using the Fama French 3 factor model

This table presents the CAAR per portfolio, calculated as the simple average over all events, using the Fama French 3 factor model, as well as its corresponding p-value. The p-value was calculated by running a two-sided t-test, assuming H0: CAAR = 0. A significance level of 95% was used in running these tests

A 0,0128 0,8182 B 0,0047 0,9306 C 0,0038 0,9441 D 0,0062 0,9094 E 0,0071 0,8961 F 0,0053 0,9226 G 0,0039 0,9427 H 0,0056 0,9176 I 0,0046 0,9335 J 0,0058 0,9148 K 0,0020 0,9708 L 0,0058 0,9158 M 0,0051 0,9249 N 0,0014 0,9800 O 0,0045 0,9343 P -0,0151 0,7832 Q 0,0056 0,9179 R 0,0064 0,9060 S 0,0026 0,9627 7. Discussion

As shown in Table 6., the general market reaction is significant on a 5% level for all events. Therefore, one could argue that generally investors do expect the identified events to have an effect on their shareholder wealth. Also, when comparing the results of the initial test with those from the robustness test, most of the events are still perceived to have a significant impact on a 5% level. For those methods 17 out of 22 are significant on a 5% level. Looking at the two events only event 7 is being insignificant in both robustness tests.

But since we’re interested in the general investors opinion for each individual industry, rather than the overall market reaction, as a response to pre-adoption news events with regards to IFRS 9 we should focus on these results instead. When looking at the abnormal returns and their significance when using the CAPM-method, the overall reaction is also significant on a 5% level, except for Portfolio A, Agriculture, forestry and fishing, and portfolio K, Financial and insurance activities. Based on these results there can be stated that hypothesis one holds for all sectors except for the Agriculture, forestry and fishing sector and the Financial and

(33)

insurance activities sector. This also means that hypothesis two does not hold, since the abnormal returns for the Financial and insurance activities sector are not significant. However, the predicted negative sign does match the actual CAAR that was presented in Table 7. The CAAR for the Financial and insurance activities sector was -0.0135 or -1.35%. A possible explanation could be that both industries make us of financial instruments, and derivatives in specific, on a wide scale. Agriculture, forestry and fishing companies often use derivatives to hedge their sales so that they don’t have to deal with market risk on the global commodities markets, whereas Financial and Insurance companies use derivatives and other financial instruments on an enormous scale for all sorts of goals. Therefore, it might take longer for such companies (and investors analyzing such companies) to make an analysis of the impact of IFRS 9 on these companies. This argument can be supported by the fact that several advisory reports on the application of IFRS 9 (e.g. Barclays’ re-visioning

provisioning; PwC’s IFRS 9 Hedging in Practice report) only emerged from January 2015 and onward. This, unfortunately falls outside the time frame examined in this study. A field for future would be to assess the impact of such reports on portfolio returns.

I also tested whether the CAAR of the Financial and insurance activities sector differed significantly from the other sectors. Table 8 shows that this isn’t the case except for the Education Sector. In this case there was a difference of 3.4%. Therefore, hypothesis three should be rejected except for the case where the Financial and insurance activities sector is compared to the Education sector. One possible explanation for this significance could be the fact that, in general, educational companies use financial instruments and derivatives at a smaller scale than companies in the Financial and insurance activities sector. But it could also be that the significant difference was caused by sheer coincidence, since the industry portfolio P, Education, only consists of 4 companies (Table 2) and the model that is used to calculate the corresponding CAAR only explains 1.68% in the returns of portfolio P. Therefore I suggest to interpret this result with caution.

All the above applies when using the CAPM-method. I also ran two robustness check to test whether the results also hold when using different models. Table 16 gives a summary of the P-values under different models for each industry’s CAAR calculated over all events. As can be seen in Table 16, all the CAARs per industry are not significant when using the mean-return model and the Fama-French 3 factor model, respectively.

(34)

Table 16 P-values for the CAAR per industry under different methods

This table summarizes the P-values for the CAARs per industry under different industries. The P-values for the CAPM method, the Mean-return model and the Fama-French 3 factor method are retrieved from Table 7, Table 10 and Table 15 respectively. Portfolio Method CAPM Mean-return model Fama-French 3 Factor A 0.1387 0,5588 0,8182 B (0.0041)* 0,0410 0,9306 C (0.0000)* 0,3765 0,9441 D (0.0343)* 0,1680 0,9094 E (0.0001)* 0,0151 0,8961 F (0.0062)* 0,0429 0,9226 G (0.0040)* 0,3867 0,9427 H (0.0000)* 0,0327 0,9176 I (0.0001)* 0,0676 0,9335 J (0.0000)* 0,0268 0,9148 K 0.0870 0,4053 0,9708 L (0.0000)* 0,0262 0,9158 M (0.0000)* 0,0241 0,9249 N (0.0013)* 0,2336 0,9800 O (0.0000)* 0,2666 0,9343 P (0.0000)* 0,0054 0,7832 Q (0.0000)* 0,0597 0,9179 R (0.0000)* 0,0512 0,9060 S (0.0000)* 0,1567 0,9627

Also, as can be seen in Table 9, only 7 industries respond significantly (on a 5% level) to the news items overall, when using the mean return method. Using the Fama French 3 Factor model, none of the industries shows a significant response to the news items. Looking at the significant (and insignificant) results under the CAPM method we could therefore also argue that the results are caused by the model that is used. The fact that the model on average only explains 28.72% of the variability in the returns, supports this explanation.

Assuming that the significant results are caused by the model, rather than actual response by investors to the news items all hypothesis should be rejected. Based on those results we could state that it take longer for all industries to analyze the impact of IFRS 9, or that investors in general don’t find the information conveyed in the news items very useful.

A review of the pre-adoption market reaction to the application of IFRS 9 across European Industries