The effect of Solvency II ratio reporting on common stock prices of European life and multiline insurance companies after controlling for IFRS reporting

(1)

University of Amsterdam

Amsterdam Business School

Master in International Finance

Master Thesis

The Effect of Solvency II Ratio Reporting on Common Stock Prices of European

Life and Multiline Insurance Companies After Controlling for IFRS Reporting

Submitted by

Cornelis van der Hulst

11081902

Under the supervision of

Dr. Razvan Vlahu

(2)

2

Acknowledgements

I wish to express my sincere gratitude to my supervisors, Dr. Razvan Vlahu and Henk-Jan Nanninga, for providing their help and expertise on the road to the completion of this thesis.

(3)

3

Abstract

This study uses an event study methodology to determine the value relevance of Solvency II ratio announcements, after controlling for IFRS based reporting. In other words, it examines whether changes in Solvency II ratio announced by insurance companies result in abnormal common stock returns. It focuses on the Solvency II ratio changes announced between 2015 FY and 2016 FY by 22 European life and multiline insurance companies listed across six countries, namely: the United Kingdom, the Netherlands, Belgium, France, Germany, and Italy. In order to prevent mixing up the common stock returns of positive and negative changes in Solvency II ratio in one dataset, this study creates two separate datasets, one for positive and one for negative changes. The empirical results of the multiple linear regression analysis, indicate that there is significant evidence of the value relevance of the change Solvency II ratio, in the case of the positive dataset. Whereas the weaker evidence in terms of the negative dataset is determined to be the result of the fact that in absolute terms the negative changes in Solvency II ratio, during the sample period, are smaller than the positive changes. The finding of this study confirms that the Solvency II regulatory framework, effective as of 1 January 2016, is increasing transparency on the capital adequacy of insurance companies. In addition, it provides guidance to insurance companies on how to deal with the hedging of economic exposures that are different under IFRS and Solvency II.

(4)

4

1. Introduction

Increased market volatility during periods of financial distress can be expected to have a significant impact on the performance of the insurance sector given its material financial leverage. However, looking at the relative performance of European insurers compared to the STOXX Europe 600 (SXXP) index and their implied cost of equity during both the global financial crisis (2008-2009) and the European sovereign debt crisis (2011-2012) it becomes apparent that the impact on insurers of those events has been rather extreme. More specifically, during the global financial crisis and the European sovereign debt crisis the implied cost of equity for the insurance sector rose to levels of 20% and 14%, respectively, up from normal levels of around 9% (see Figure 1). According to the European Insurance and Occupational Pensions Authority (‘EIOPA’)1_{(2017) the lack of robust risk and governance} management under Solvency I made it difficult for the national supervisors to obtain adequate information. The volatility in the implied cost of equity signals that investors experienced a similar lack of transparency regarding the financial condition of insurance companies. Furthermore, Foroughi and Crean (2017) stress the specific importance of transparency on cash flows and capital adequacy towards the next crisis.

Figure 1: Relative share price performance of European Insurers vs STOXX Europe 600 (SXXP) index (2006 = 100) and implied cost of equity

Source: Foroughi and Crean (2017)

When looking into the Solvency I framework this lack of transparency towards equity analysts and investors appears to have been to a large extent the result of limited consistency across the 14 EU directives and the 28 national supervisory regimes that were in place. Conversely, the Solvency II supervisory framework, effective as of 1 January 2016, is designed as a common regime to be applied

1 EIOPA is at the heart of inurance and occupational pensions’ supervision in the European Union. It is part of the European System of Financial Supervision (ESFS) which includes also the European Banking Authority (EBA), the European Securities and Markets Authority (ESMA) and the European Systemic Risk Board (ESRB).

(6)

6

by all 28 member states of the European Union2_{thus resulting in more homogenous reporting.} Furthermore, it is a risk-based supervisory regime that holds great potential value in terms of increased transparency on capital adequacy. According to Duverne & Le Douit (2009) this risk-based focus is based on a fair value/economic balance sheet that is closely aligned to the position of the European industry on how the business is managed. Article 75 of the Solvency II directive, by the European Commission (2009), covering the Valuation of assets and liabilities, outlines the economic balance sheet approach by stating that Member States must ensure that insurance and reinsurance undertakings value assets and liabilities as follows:

a)! assets shall be valued at the amount for which they could be exchanged between knowledgeable willing parties in an arm’s length transaction;

b)! liabilities shall be valued at the amount for which they could be transferred, or settled, between knowledgeable willing parties in an arm’s length transaction.

In addition, Article 9 (covering the Valuation methodology – general principles) and 10 (covering the

Valuation methodology – valuation hierarchy) included in the Solvency II regulation, by the European

Commission (2015), provide further guidelines on this approach.

However, the increased transparency and consistency of the new supervisory regime comes at the disadvantage of having less consistency between the regulatory (economic) balance sheet and the current (local) accounting balance sheet. In section 2.1, the current IFRS insurance accounting framework as well as the recently announced IFRS 17 standard are discussed in detail. In practice the differences between the regulatory and accounting balance sheet results in significant differences in the estimated equity levels available to insurers. Figure 2 shows a break-down of this difference into its main factors for a sample of 17 insurance companies at the end of December 2016 relative to their IFRS equity that is set to a scale of 100 percent. For example, the fact that goodwill and intangible assets are valued at zero under Solvency II while they are reported at amortized cost under IFRS results in a relative average decrease of 34.1% in available equity. The differences in valuation of the other key factors reported (i.e. MV adjustments to assets, Liability valuation adjustments, other smaller asset/liability adjustments, and subordinated debt inclusion) are due to the fact that under Solvency II they are reported at fair value while under IFRS they are typically valued at amortized cost.

(7)

7

Figure 2: Conversion of IFRS equity to Solvency II eligible own funds

Source: Foroughi and Crean (2017)

On average the Eligible Own Funds (‘EOF’) under Solvency II are reported to be 23.8% higher than reported IFRS equity. This significant difference combined with the aforementioned fact that the Solvency II regulatory framework is based on an economic balance sheet, and thus provides a valuation based on fair value, raises the question whether announcements of Solvency II ratios (= Eligible Own Funds / Solvency Capital Requirement3_{) are value relevant to investors complementary to IFRS} reporting.

A recent event involving Aegon N.V. (one of the largest insurers worldwide with over 30 million customers) provided a clear indication that Solvency II ratio announcements, independent of any other news, can result in large stock price reactions. As noted by Hensen (2017), when on the 24th of March 2017, it adjusted its Solvency II ratio as of 31 December 2016 downwards to 157% for the group and to 135% for its Dutch subsidiary from the previously communicated 159% and 141%, respectively, its share price decreased with more than 5%.

In accordance with the aforementioned the following research question was derived:

Does Solvency II ratio reporting by European life and multiline insurance companies provide value relevant information to investors (measured using abnormal common stock returns) when controlling for IFRS based reporting?

3 Solvency Capital Requirement is set at a level to ensure that insurers and reinsurers can meet their obligations to policy holders and beneficiaries over the following 12 months with a 99.5% probability.

(8)

8 1.1 Scientific relevance

The scientific relevance of this study is closely related to the fact that it is the first empirical study to investigate the value relevance of Solvency II reporting since its recent introduction, effective 1 January 2016. Furthermore, it covers the sample period between 2015 FY and 2016 FY4_{which covers all reliable} Solvency II ratio announcements by the included listed European life and multiline insurers to date. When, considering a broader range of research on the value relevance of ‘realistic reporting’ Horton (2007) established the relevance of in-force long-term business assets reported by insurers as part of voluntary embedded value reporting.

1.2 Research design

Given that Insurers announce their Solvency II ratios on their financial reporting dates5 this study will directly control for relevant reported IFRS based figures. This study will examine the change in Solvency II ratio compared to the most recent ratio that the respective insurer communicated to the market.6_{Hence, the prior Solvency II ratio thus serves as a proxy for the current market expectation.} The rationale for doing so is based on the assumption that the semi-strong Efficient Market Hypothesis (EMH), as introduced by Fama (1970), holds meaning that all publicly available information is reflected in the market prices and that an updated Solvency II ratio contains private information that is new to the market on the announcement date.

In order to prevent mixing up the common stock returns of positive and negative changes in Solvency II ratio in one dataset this study creates two separate datasets, one for positive and one for negative changes. This is closely aligned with the methodology followed by (Holthausen and Leftwich (1986), Goh and Ederington (1998), and Kim and Nabar (2003)) that investigate potential private data contained in announcements of credit rating changes and their effect on common stock returns.

In terms of structure, this research can be separated into three components. First, based on an event study methodology introduced in section 3.1, the Abnormal Returns (ARs) on the common stock of the listed European life and multiline insurers are determined on the Solvency II ratio reporting dates. Second, for both the positive (sub-question 1.1) and negative dataset (sub-question 1.2) the existence of significant ARs will be investigated using a one-sample t-test. Third, using multiple linear regression analysis, the effect of a change in solvency II ratio on ARs will be estimated while controlling for the IFRS based control variables (sub-question 1.3 and 1.4). The control variables have been selected based on related literature and can be allocated to either of four categories, namely: operational efficiency, profitability,

4 The potential Solvency II ratio announcements are thus related to the financial figures as of 2015FY, 2016Q1, 2016Q2, 2016Q3, 2016FY.! 5 Alongside their quarterly, semi-annual or full year IFRS results.

6 For example: %"#ℎ%&'(")&"*+,-(&./"00"1%2)+ = 456789:;"<<"=>?@5"ABCDEFG456789:;"<<"=>?@5"ABCDEH 456789:;"<<"=>?@5"ABCDEH

(9)

9

growth, and coverage/leverage ratios. The control variables and the expected sign of their coefficient are introduced in section 2.2. More specifically, first a regression only including the change in Solvency II ratio as independent variable is estimated. Second, four regressions containing the change in Solvency II ratio variable and the independent variable(s) belonging to either of the four categories of control variables are estimated. Third, a full regression including all independent variables and best-fit regression based on the Akaike Information Criterion (‘AIC’) are estimated. The respective sub-questions are presented below:

Sub-question 1.1: Are there significantly positive abnormal common stock returns associated with

increases in Solvency II ratio as announced by European life and multiline insurance companies?

Sub-question 1.2: Are there significantly negative abnormal common stock returns associated with

decreases in Solvency II ratio as announced by European life and multiline insurance companies?

Sub-question 1.3: When controlling for IFRS reporting does a positive change in Solvency II ratio

reported by European life and multiline insurance companies result in significantly positive abnormal common stock returns?

Sub-question 1.4: When controlling for IFRS reporting does a negative change in Solvency II ratio

reported by European life and multiline insurance companies result in significantly negative abnormal common stock returns?

1.3 Overview of results

First of all, it is important to mention that as expected by Gabriel Bernardino, Chairman of EIOPA, the Solvency II regulation represents a change in metrics that has resulted in a relatively volatile Solvency II ratio. For this study, the average change in Solvency II ratio is 5.19% and -4.69% for the positive and negative datasets, respectively, as presented in Table 6 and 7. Hence, if the Solvency II ratio is indeed an important metric for investors, the announcements of such changes have the potential to lead to ARs. For the first part of the analysis, the result of the t-test for the positive dataset shows significantly positive ARs. Conversely, for the negative dataset, the negative ARs are insignificant at a 10% level. Hence, the empirical evidence indicates that the financial markets react mainly to increases in the Solvency II ratio. However, a potential explanation for the asymmetric nature of these results is that, in absolute terms, the positive dataset contains larger Solvency II ratio changes, as presented in Table 8. In support of this explanation additional t-tests were performed for two subsets of both the positive and negative dataset based on whether the Solvency II ratio change was smaller or larger than one standard deviation. For the subset of negative Solvency II ratio changes larger than one standard deviation, the ARs are more negative and significant at a 10% level. Furthermore, for the subset of positive Solvency II ratio changes

(10)

10

larger than one standard deviation, the ARs are more positive and significant. Hence, this study finds significant evidence that financial markets react strongly to large changes in Solvency II ratio in either direction. This finding is in line with the view expressed by David Prowse, senior director at Fitch Ratings, that investors can be expected to respond to highly volatile Solvency II metrics, as noted by Tanner (2016).

For the second part of the analysis, the estimation result of the best fit multiple linear regression for the positive dataset indicates that, the change in Solvency II ratio, change in insurance investment yield and growth in net premiums earned variables significantly explain part of the variance in the dependent variable, AR. The positive coefficient of the change in Solvency II ratio variable signals that, after controlling for IFRS based control variables, investors perceive an increase in Solvency II ratio as a signal of an improved capital position of the insurer. As a result, it is a relevant metric in terms of the capital management of insurance companies. Furthermore, the positive coefficient of the insurance investment yield variable indicates that the realized investment yield by insurers is one of the main concerns of investors in the current low interest rate environment. Lastly, the positive coefficient of the net premiums earned shows that investors reward the strong growth in net premiums earned during the sample period, as further elaborated on in section 5.3.

In addition, the estimation result of the best fit multiple linear regression model, for the negative dataset, indicates that the coefficient of the change in Solvency II ratio, is insignificant at a 10% level. Hence, for this dataset there is no empirical evidence that the Solvency II ratio is a relevant metric in terms of the capital management of insurance companies. Furthermore, the insurance investment yield, policy reserves/total assets, and policy reserves/equity variables do significantly explain part of the variance in the dependent variable, AR. The positive coefficient of the insurance investment yield reconfirms the aforementioned importance of the realized investment income to investors in the current environment. Furthermore, the positive coefficient of the policy reserves/total assets variable contradicts with the expectation expressed in section 2.2. Lastly, the negative coefficient of the policy reserves/equity variable support the perspective that a one percent increase in insurance contract related obligations relative to equity is perceived by investors as a signal of a deterioration in the capital position.

2. Theoretical framework

2.1 Literature Review

First, this section elaborates on the Solvency II regulatory framework. Thereafter, on the IFRS accounting framework. As aforementioned, the introduction of the Solvency II framework has resulted in less consistency between the regulatory (economic) balance sheet and the current (local) accounting balance sheet. Hence, it is important to discuss both the current insurance accounting framework and whether developments (i.e. the recently published IFRS 17 standard) will improve the consistency going forward.

(11)

11 Solvency II regulatory framework

While the Solvency II framework is based on three interconnected pillars namely, pillar I – capital requirements, pillar II – risk management and system of governance, and pillar III – external reporting, this section focuses on the most relevant aspects of pillar I. Some details related to pillar I and a brief overview of the scope of pillar II and III are presented in appendix B and C, respectively. In addition, it is important to discuss the long-term guarantees (‘LTG’) measures. LTG measures are allowed and used to ensure an appropriate treatment of insurance products7_{that include LTGs, because of their significant} impact on the Solvency Capital Requirement (‘SCR’) ratio and widespread usage. More specifically, this section focuses on the extrapolation of the risk-free rate because of its mandatory application. In addition, using the example of interest rate risk, it briefly discusses some approaches followed by insurance companies, in this early stage of applying the new regulatory framework, to hedge their risk exposures. Other (voluntary) LTG measures, namely the: Volatility Adjustment (‘VA’), Matching Adjustment (‘MA’), and Transitional on Technical Provisions (‘TTP’), are covered in appendix D.

Pillar I – Capital requirements

As aforementioned in the introduction of this paper, under the Solvency II regulatory framework assets and liabilities are valued at the amount for which they could be exchanged between knowledgeable willing parties in an arm’s length transaction. The assets of an insurer consist mainly of the investment portfolio financed by the insurance premiums that it has received. Conversely, the liabilities consist mainly of technical provisions set up for the outstanding insurance obligations that, especially in the case of life insurance companies, can have a long duration. The difference between the assets and liabilities, previously identified as the Eligible Own Funds, should cover the capital requirements applicable under Solvency II.

Under Solvency II there are two capital requirements namely the, (1) Solvency Capital Requirement, and; (2) Minimum Capital Requirement (‘MCR’). The SCR is equal to the 99.5% one-year Value at Risk (VaR). In other words, with 99.5% confidence the worst annual loss will not exceed the SCR. In case an insurance company is incompliant with the SCR it has to take measures to ensure that it will be compliant within the following 6 months. It should be noted that the SCR ratio is equal to the previously defined Solvency II ratio (= Eligible own funds / Solvency Capital Requirement). Hence, a SCR ratio of 100% or higher implies that the insurer complies with the SCR. An overview of the risk modules on the basis of which the SCR is calculated with the standard formula as well as an overview of the approximate SCR charges by asset class is provided in Appendix B, for illustration purposes. The MCR corresponds to the absolute minimum level of security that is required under Solvency II and is usually

7 For example, traditional life insurance (e.g. with profit contracts ,saving products, annuities), unit-linked policies with guaranteed investment yield or capital protection, and variable annuities.

(12)

12

between 25% and 45% of the SCR. According to EIOPA (2016), in case an insurance company is incompliant with the MCR its authorization will be withdrawn unless the MCR is covered again within 3 months.

For illustration purposes the simplified balance sheet of an insurer displayed below (figure 3) provides overview of the capital requirements and the EOF.

Figure 3: Simplified balance sheet for an insurer

In relation to the scope of pillar II and III presented in Appendix C the public disclosure, by insurers, by means of the annual Solvency and Financial Condition Report (‘SFCR’)8_{is expected to provide valuable} information on their solvency position for investors. More specifically, it will provide detailed information regarding their solvency and financial condition, activities and results, operations, risk profile, the principles used to value their assets, technical provisions and other liabilities, and capital management.

Long-term guarantee measures – the extrapolation of the risk-free rate

Under Solvency II there is a specific yield curve determined for each currency that insurers use to discount their insurance obligations. For the relatively short durations this yield curve is based on actual swap rates, however for longer durations when the markets become less liquid the curve is extrapolated towards the Ultimate Forward Rate (‘UFR’). The UFR has three main elements, namely (1) the ultimate forward rate itself currently still 4.2%9_{to which the long-dated interest rates are assumed to converge;} (2) the Last Liquid Point (LLP) in case of the EURO this is 20 years being the point from which the

8 The first SFCRs will be published starting 20 May 2017 concerning the 2016 financial year.

9 On the 5th of April 2017 the EIOPA published an updated methodology for calculating the UFR on the basis of which the appropriate rate is 3.65%. As a result, the UFR will be reduced with 15 basis points per year starting next year (meaning that it will be 4.05% in 2018). Hence, i twill take five years before the UFR reaches below 3.65%.!

(13)

13

actual swap curve is no longer used, and; (3) the rate of convergence from market rates at the LLP to the UFR. The main argument in favor of using an UFR type approach is that it avoids market impact and pro-cyclical behavior in the long end of the curve. On the other hand, using an UFR that is unrealistically high understates the present value of insurance obligations and thus overstate the EOF and SCR ratio. As noted by Tanner (2017), as a consequence of such concerns, given the current low interest rate environment, the EIOPA decided to reduce the UFR in annual increments of 15 basis points to a level of 3.65%. According to Foroughi and Crean (2017), in terms of sensitivity, the SCR ratio would decrease on average by 12% (as of 31 December 2016) in case of a 100 basis points decrease of the UFR.

In terms of general interest rate sensitivity of the SCR ratio, insurance companies find themselves in a catch-2210_{situation as a result of underlying differences in the exposure under the Solvency II and IFRS} frameworks. More specifically, insurance companies can focus on hedging their interest rate exposures to protect their: (1) SCR ratio; (2) profitability, or; (3) a combination of both. In practice insurance companies indeed seem to hold onto different approaches in managing their interest rate sensitivity. For example, according to Tanner (2016), Legal & General reported a close-to-zero sensitivity of its profitability in relation to the first half-year of 2016, while at the same time noting that if interest rates fell by 100 basis points its Solvency II ratio would decrease severely by 14 percentage points. On the other hand, some of Legal & General’s biggest competitors (i.e. Aviva, Prudential, and Standard Life) have declared to give priority to reducing the impact of interest rate movements on their SCR ratio. According to David Prowse, senior director at Fitch Ratings, in order to maintain investor confidence, it is important for insurance companies to ensure that their solvency metrics, among which the SCR ratio, are not too volatile. Hence, according to his understanding, the most common approach followed by insurance companies is to hedge their risk exposures in between IFRS and Solvency II metrics. The main reason why investors and stock analysts would focus on the volatility of the SCR ratio is because they are concerned that firms may fall below the regulatory capital threshold which can, for example, impact their ability to pay dividends.

The IFRS insurance accounting framework

In 1997, the predecessor of the International Accounting Standards Board (‘IASB’)11_{launched an} ambitious insurance contracts accounting standard project. Because of the complexity of the comprehensive reform it was decided in 2002 to split-up the project in two phases, namely IFRS 4 Phase I and IFRS 4 Phase II (from here on called IFRS 17). The IFRS 4 Phase I standard has been effective since 1 January 2005. However, it only imposes limited requirements and as a result mostly left the

10_{A catch-22 is a paradoxical situation from which a certain party cannot escape because of contradictory rules.}

(14)

14

practices applicable prior to the implementation of the standard unchanged. This effectively made it a temporary mechanism for accommodating Generally Accepted Accounting Principles (‘GAAP’) accounting for most insurance contracts. As a result, in terms of comparability, under the current applicable standard for insurance contracts there is (1) a lack of comparability between insurers12_{; (2)} non-uniform reporting of products within insurance groups13_{, and; (3) inconsistency with the revenue} reporting standards followed by other industries14_{. Furthermore, in terms of quality of financial} information, there is a lack of transparency of profitability. In addition, until the implementation of IFRS 17 (i.e. IFRS 4 Phase II) insurers are exempt from applying the criteria in International Accounting Standards (‘IAS’) 8 Accounting Policies, Changes in Accounting Estimates and Errors that is applicable for developing an accounting policy where no IFRS applies specifically to an item.15

The IFRS 17 standard published on the 18th of May 2017 by the IASB, on the other hand, covers the major comparability and transparency of profitability related issues and thus concerns a comprehensive reform for liability measurement. More specifically, the new framework replaces a huge variety of accounting treatments which improves the comparability of reporting between insurers as well as within insurance groups. Second, the revenue reported will reflect the services provided instead of the cash received which is in line with the reporting by other financial service industries. Third, a new structure of the statement of comprehensive income provides increased transparency on financial performance. For example, as displayed in Figure 4, under IFRS 17 there is a separation between insurance income & expenses (i.e. insurance service result) and investment income & expenses (i.e. net financial result) which enables investors to understand the origin of the profit or loss reported.

Figure 4: Statement of comprehensive income under IFRS 4 and IFRS 17

12 IFRS companies report insurance contracts using different practices. 13 Insurance contracts of subsidiaries are consolidated using different practices.

14 Revenues are currently reported on a cash basis and include deposits instead of being based on services provided.

15 Separate from the two phases described, the IASB also issued IAS 39 Financial Instruments: Recognition and Measurement which encouraged fair value measurement but enabled amortized cost to be used.!

(15)

15

Fourth, a Contractual Service Margin16_{(‘CSM’) is determined upon initial recognition of a new} insurance contract. Under the standard Building Block Approach (‘BBA’) the CSM is equal to the present value of future cash inflows minus present value of future cash outflows minus Risk Adjustment. The CSM is floored by zero17_{meaning that it cannot become negative. The purpose of recognizing a} positive initial CSM is to be able to recognize the profit related to the contract over the coverage period in a way that best reflects the service to be provided, instead of recognizing the full profit at day-one. According to Foroughi and Crean (2017), when comparing IFRS 17 to Solvency II, the principles underlying the discounting of the cash inflows and outflows, the recognition of an explicit Risk Adjustment (under Solvency II called the Risk Margin), and profit recognition are similar. Hence, it can be expected that, once implemented, IFRS 17 will improve the alignment of the accounting balance sheet with the Solvency II regulatory balance sheet. Although it should be noted that IFRS 17 will continue to be a hybrid between book and market value accounting. As of today, IFRS 17 together with the asset-side related standard, IFRS 9 (Recognition and measurement of financial instruments), should be implemented as of 2021.

2.2 Control variables

This section introduces the eight control variables based on IFRS figures that will be used in the multiple regression analyses in order to answer sub-questions 1.3 and 1.4. The eight independent variables are (1) total net margin, (2) insurance investment yield, (3) forecast error Book value per share (‘BPS’), (4) forecast error Earnings per share (‘EPS’), (5) net premiums earned, (6) insurance liabilities coverage ratio, (7) insurance liabilities to assets, and (8) insurance liabilities to equity. They can be divided into four categories, namely: operational efficiency, profitability, growth, and coverage / leverage ratios.

Operational efficiency

(1)! Total net margin (‘TNM’)

Change in total net margin is defined as the change in the net income as a percentage of the total net premiums earned and fee income (see formula 1). According to Kim et al. (1995), it is one of the key determinants of life insurer insolvencies. The expected sign of the coefficient is positive given that an increase in net margin signals an increased operating efficiency to investors.

∆"J+2%,"&(2"K%1')& = (J+2%,"&(2"K%1')&"2016Q4 − J+2%,"&(2"K%1')&"2016Q3) (1)

16 A component of the carrying amount of the asset or liability for a group of insurance contracts representing the unearned profit the entity will recognise as it provides services under the insurance contracts in the group.

17 In case the CSM would be negative at inception (the insurance contract is onerous) or would become negative during the coverage period (due to e.g. revised assumptions) losses should be recognised in the profit and loss account immediately.!!

(16)

16 Profitability

(2)! Insurance investment yield (‘IIY’)

Change in insurance investment yield is defined as the change in the return on invested assets (see formula 2). For my sample of listed European life and multiline insurers approximately 25% of the total income is classified as investment income. This is the result of the considerable investment portfolio held by insurers as coverage for their outstanding insurance contracts. Hence, the realized investment yield is of significant importance for their overall performance. According to Kim et al. (1995), it is one of the key determinants of life insurer insolvencies. The expected sign of the coefficient is positive given that an increased realized investment yield on invested assets signals more favorable investment conditions.

∆"0&VW1%&.(")&-(V2K(&2"/)(,X = (0&VW1%&.(")&-(V2K(&2"/)(,X"2016Q4 −

0&VW1%&.(")&-(V2K(&2"/)(,X"2016Q3) (2)

(3)! Forecast error book value of equity per share (‘BPS’)

The book value of equity per share forecast error is defined as the difference between the actual book value of common stockholder’s equity divided by the number of common shares outstanding and the book value per share estimate18_{relative to the book value per share estimate (see formula 3). According} to Collins et al. (1999), the book value of equity is a value-relevant proxy especially for expected future normal earnings for loss making firms. The expected sign of the coefficient is positive given that a positive forecast error signals a stronger equity position than expected.

Y+1(.% 2"(11+1")&"Z[* = >:?\>6"]^4"ABCDEFG8_?@`>?8a"]^4"ABCDEF

8_?@`>?8a"]^4"ABCDEF (3)

(4)! Forecast error earnings per share (‘EPS’)

The earnings per share forecast error is defined as the difference between the actual earnings per share and the estimated earnings per share relative to its current share price (see formula 4). According to Christie (1987), the current share price is the correct deflator of the earnings per share forecast error in the case of return studies. The expected sign of the coefficient is positive given that a higher earnings per share than expected is a positive surprise for investors which can potentially result in positive ARs.

Y+1(.%V2"(11+1")&"b[* =(>:?\>6"c^4"ABCDEFG8_?@`>?8a"c^4"ABCDEF)

_d>=8"e=@:8"ABCDEF (4)

(17)

17 Growth

(5)! Change in net premiums earned (‘NPE’)

Change in net premiums earned (see formula 5) is a proxy for revenue growth realized by the life and multiline insurance companies. Net premiums earned is defined as net premiums written adjusted by the change in net unearned premiums for a year. Net premiums earned is deemed more applicable than gross premiums earned because it excludes the reinsurance premiums. It should be noted that Kim et al. (1995) find mixed evidence for the significance of this variable as determinant of life insurance insolvencies. However, due to its significance as determinant for property-liability insurer insolvencies it was deemed important to be included in this study. The expected sign of the coefficient is positive given that revenue growth can be interpreted by investors as a proxy of market potential.

%"#ℎ%&'(")&"f(2"g1(K)WKV"(%1&(X =

h8?"e=8`@\`_"8>=98a"ABCDEFGh8?"e=8`@\`_"8>=98a"ABCiEF

h8?"e=8`@\`_"8>=98a"ABCiEF " (5)

Coverage / leverage ratios

(6)! Insurance liabilities coverage ratio (‘Coverage’)

Change in the insurance liabilities coverage ratio is defined as the change in the ratio of total investments divided by policy reserves (see formula 6). Policy reserves are defined as insurance reserves and liabilities for insurance and investment contracts (this includes insurance and investment reserves backing unit-linked contracts). This coverage ratio is included because it can be considered as an IFRS based (simple) alternative for the Solvency II ratio. The expected sign of the coefficient is positive given that a growth in coverage ratio should make investors more certain about the coverage of liabilities.

%"#ℎ%&'(")&"#+-(1%'("1%2)+ = j578=>k8"=>?@5"ABCDEFGj578=>k8"=>?@5"ABCDEH

j578=>k8"=>?@5"ABCDEH (6)

(7)! Insurance liabilities to assets

The insurance liabilities to assets variable is defined as the change in the ratio of policy reserves (see aforementioned definition under (6)) to total assets (see formula 7). The expected sign of the coefficient is negative given that an increase in the leverage ratio signals increased risk to investors about the coverage of insurance contract liabilities by available assets.

%"#ℎ%&'(")&"g+,)./"1(V(1-(V"2+"%VV(2V"1%2)+ = 6878=>k8""=>?@5"ABCDEFG6878=>k8"=>?@5"ABCDEH 6878=>k8"=>?@5"ABCDEH (7)

(8)! Insurance liabilities to equity

The insurance liabilities to shareholders’ equity is defined as the change in ratio of policy reserves (see aforementioned definition under (6)) to shareholders’ equity (see formula 8).

(18)

18

The expected sign of the coefficient is negative given that similar to the asset based leverage ratio (7) an increase in the leverage ratio signals increased risk to investors about the coverage of insurance contract liabilities.

%"#ℎ%&'(")&"g+,)./"1(V(1-(V"2+"(lW)2/"1%2)+ = 6878=>k8""=>?@5"ABCDEFG6878=>k8"=>?@5"ABCDEH 6878=>k8"=>?@5"ABCDEH (8)

To summarize, Table 1 provides an overview of the predicted signs for the selected independent variables.

Table 1: Summary of predicted signs of the explanatory variables

Explanatory variable Predicted Sign

Change in total net margin +

Change in insurance investment yield +

Forecast error in book value per share +

Forecast error in earnings per share +

Growth in net premiums earned +

Change in total investments / policy reserves +

Change in policy reserves / assets -

Change in policy reserves / equity -

Change in Solvency II ratio +

3. Research methods and techniques

First, the event study methodology used to calculate the ARs is introduced. Second, the one sample t-test and the nonparametric binomial t-tests performed in order to answer sub-questions 1.1 and 1.2 are introduced. Third, the multiple linear regression analysis used to test the relationship between the dependent variable, AR, and the explanatory variables such as the change in Solvency II ratio and the various control variables based on IFRS reporting are introduced. The regression analysis is performed to answer sub-questions 1.3 and 1.4.

3.1 Event Study

This research uses the results of an event study as input for the one sample t-tests and multiple linear regression analyses. As mentioned by MacKinlay (1997, p. 13) “the usefulness of such a study comes from the fact that, given rationality in the marketplace, the effects of an event will be reflected immediately in security prices.” In other words, an event study measures the impact of a specific event on the value of the firm.

For this research an event is classified as either a positive or a negative change in Solvency II ratio compared to the previous announced Solvency II ratio of the respective insurer. Given this definition both an estimation period and an event window are required. The estimation period is used to estimate

(19)

19

the market model parameters that will be used to calculate the realized ARs of a certain security during the event window on the day of the Solvency II ratio announcement. For this research the market model parameters are estimated using a 150 trading days long estimation period, from trading day -180 to -30. The event window itself only includes the day of the announcement of the Solvency II ratio by the respective life or multiline insurance company. This is sufficient because the relevant information is announced around 8:00 am in the morning. Hence, investors are expected to respond to the news on the same day.

In order to calculate the ARs on the announcement dates this research uses the Event Study Tool prepared by Oord (2017). This is done based on a market model that can be defined as a statistical model that relates the return of any given security to the return of the market portfolio.

First, the Event Study Tool calculates the daily returns for each sample company for both the estimation window (-180, -30) and the event window (-1, 0) as follows:

Rit = (Pit – Pit-1)/Pit-1 (9)

Where, Pit and Pit-1 are the respective daily prices for company ‘i’ at time ‘t’ and ‘t-1’.

Second, it calculates the daily returns of the market portfolio. This research uses the FTSE 35019_and the MSCI All Countries World20_{index as a proxies for the market index. According to MacKinlay} (1997), this is in line with the practice to use a broad based stock index for the market portfolio. This was done for each sample company for both the estimation window (-180, -30) and the event window (-1, 0) using formula (10).

Rmt = (It – It-1)/It-1 (10) Where, It and It-1 are the respective daily index values at time ‘t’ and ‘t-1’.

Third, it estimates the parameters αi and βi of the market model by regressing Rit on Rmt during the estimation window (-180, -30) using formula (11).

Rit = αi + βiRmt + ARit (11) Where,

Rit is the observed daily return for the share of a company ‘i’ at time ‘t’, Rmt is the observed daily return for the market index at time ‘t’,

αi is the estimate of the intercept for the company ‘i’,

βi is the estimate for the market beta for the shares of company ‘i’, and ARit is the abnormal return of company ‘i’ at time ‘t’ (i.e. the error term).

19 Mnemonic: FTSE350. Applicable to the observations in relation to insurers listed in the United Kingdom.

20 Mnemonic: MS1AWFE. Applicable to the observations in relation to insurers listed in the Netherlands, Belgium, France, Germany and Italy.!

(20)

20

Fourth, using the estimates of the parameters αi and βi of the market model, derived from the estimation window, the AR for each observation can be calculated using the return of the FTSE 350 or MSCI All Countries World index and that of the respective stock on the announcement day, using formula (12).

ARit = Rit – (αi + βiRmt) (12)

3.2 One Sample T-test

In order to answer sub-questions 1.1 and 1.2, related to the positive and negative dataset, respectively, one sample t-tests are performed in order to determine the value relevance of Solvency II ratio reporting. These one sample t-tests are used to examine whether the AR of the respective dataset is significantly different from zero. In addition, a nonparametric binomial test will be performed for each of the two datasets. These binomial tests are performed to test the null hypothesis of an even distribution (having an expected mean of 0.5). Under the null hypothesis the expected mean is 0.5 because half of the observed ARs are expected to be positive (therefore receiving the value 1) and the other half, negative (therefore receiving the value 0).

3.3 Multiple Linear Regression Analysis

As mentioned before, this research performs multiple linear regression analysis to test the relationship between the dependent variable, AR, and the independent variables.

First, a regression solely including the change in Solvency II ratio as an independent variable will be examined (see equation 13).

mn@= " oC∆*00@+ q@ (13)

Second, four regressions (see equations 14-17) are estimated each including the change in Solvency II ratio variable plus one of the four categories of independent variables introduced in section 2.2. These regressions are estimated to separately examine the effect of the independent variables included in each category on the ARs, while accounting for the main variable of interest being the change in Solvency II ratio. Operational efficiency mn_@ = " o_C∆Jfr_@+ o_A∆*00_@+ q_@ (14) Profitability mn@= " oC∆00s@+ oAZ[*@+ oHb[*@+ oF∆*00@+ q@ (15) Growth mn_@ = " o_C∆f[b_@+ " o_A∆*00_@+ q_@ (16) Coverage/Leverage ratios mn_@ = " o_C∆#+-(1%'(_@+ o_A∆^56@:;"t8_8=78_ u__8?_ @+ oH∆ ^56@:;"t8_8=78_ cv\@?; _@+ " oF∆*00@+ q@ (17)

(21)

21

Third, a regression of the full model is estimated including all independent variables (see equation 18).

Full model mn@ = " oC∆Jfr@+ oA∆00s@+ oHZ[*@+ oFb[*@+ oi∆f[b@+ oD∆#+-(1%'(@+ ow∆^56@:;"t8_8=78__{u__8?_} @+ ox∆ ^56@:;"t8_8=78_ cv\@?; _@+ " oy∆*00@+ q@ (18)

Fourth, a best fit regression is estimated based on the Akaike Information Criterion. More specifically, equations 19 and 20 represent the equation of the best fit model for the positive and negative changes in Solvency II ratio, respectively.

Best fit model - positive dataset

mn_@= " o_C∆00s_@+ o_Ab[*_@+ o_H∆f[b_@+ o_F∆*00_@+ q_@ (19)

Best fit model - negative dataset

mn_@ = " o_C∆00s_@+ o_A∆^56@:;"t8_8=78_

u__8?_ @+ oH∆

^56@:;"t8_8=78_

cv\@?; _@+ " oF∆*00@+ q@ (20) The correlation table of all the independent variables will be examined for an indication of multicollinearity. Furthermore, it will be examined whether the assumptions of the Gauss-Markov Theorem are satisfied.

(22)

22

4. Data

This study focuses on the Solvency II ratio changes announced by 22 life or multiline insurance companies in relation to the period between 2015 FY and 2016 FY. The final sample size of 22 insurance companies was selected from an initial sample of 102 European listed insurance companies based on information from ThomsonOne21. Table 2 provides an overview of the criteria and the number of firms that were excluded to arrive at the final sample size of 22 companies. As can be observed (34) companies were excluded due to their focus on the Property & Casualty business segments, (33) due to lack of data availability in the SNL Insurance database used extensively for data gathering in a later stage, (3) due to illiquidity, (3) merged during the sample period, (3) due to having a non-EUR or GBP reporting currency, (2) due to poor quality Solvency II reporting, (1) due to a main focus on bank activities, and (1) due to being a reinsurer.

Table 2: Sample selection of 22 European life or multiline insurance companies

Full sample of listed European Insurance companies 102

- Property & Casualty -34

- Data unavailable in SNL Insurance database -33

- Illiquid (Assets < 400 million or Market cap < 200 million) -3

- Firms that merged during sample period -3

- Non EUR or GBP as reporting currency (DK, NO, PL) -3

- Poor quality Solvency II ratio reporting -2

- Bank -1

- Reinsurance -1

Used sample size of listed European life or multiline

Insurance companies 22

Source: ThomsonOne & SNL Insurance

The 22 insurance companies included in the final sample are all listed in either of six European countries namely, the United Kingdom, the Netherlands, Belgium, France, Germany, or Italy. Considering the sample period between 2015 FY and 2016 FY a total of 66 Solvency II ratio changes were available. Out of these observations a total of 622_{were removed from the sample due to the unavailability of an} EPS estimate information making it impossible to calculate the Forecast Error in EPS. The remaining 60 observations consist of 34 increases in Solvency II ratio and 26 decreases in Solvency II ratio.

21 Based on an initial selection of all life and non-life insurance companies listed in any of 29 European countries, being: Austria, Belgium, Bulgaria, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Norway, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden, The Netherlands, and The United Kingdom.

(23)

23

Table 3 provides an overview of the distribution of the observations per country related to the positive and negative Solvency II ratio changes.

Table 3: Distribution of data-points across countries

Country Positive Solvency II ratio change

Negative Solvency II ratio change # % # % United Kingdom 11 32.4% 8 30.8% The Netherlands 5 14.7% 4 15.4% Belgium 1 2.9% 2 7.7% France 6 17.6% 3 11.5% Germany 7 20.6% 3 11.5% Italy 4 11.8% 6 23.1% Total: 34 100.0% 26 100.0%

Furthermore, Table 4 provides an overview of the distribution of the observations per reporting period. As can be observed, the observations are not equally distributed across the reporting periods because only a part of the insurance companies provide quarterly reporting.

Table 4: Distribution of data-points across reporting periods

Reporting periods Positive Solvency II ratio change

Negative Solvency II ratio change # % # % 2016 - FY 16 47.1% 5 19.2% 2016 - Q3 4 11.8% 6 23.1% 2016 - Q2 9 26.5% 11 42.3% 2016 - Q1 2 5.9% 4 15.4% 2015 - FY 3 8.8% 0 0.0% Total: 34 100.0% 26 100.0%

In relation to the data gathering process for the control variables the SNL Insurance Peer Analytics tool was used as the main source of information. The tool is essentially a database of financial information that it is directly based on the financial statements published by the respective insurers. It was used to gather the data required for the following control variables: total net margin; insurance investment yield; forecast error BPS (the actuals component); net premiums earned; insurance liabilities coverage ratio; insurance liabilities to assets, and; insurance liabilities to equity.

The remaining information related to the BPS estimate component of the forecast error BPS variable and both the EPS estimate and actual components for the forecast error EPS variable were sourced from leading providers of estimates and financial data. In case of the BPS estimate the mean estimate was obtained from Factset or S&P Cap IQ and in case both were available the average was used. In case of the EPS related information Table 5 provides an overview of the origin of the information.

(24)

24

More specifically, the mean EPS estimate was obtained from Factset, S&P Cap IQ or Thomson Reuters I/B/E/S. Furthermore, the actual EPS was obtained from Factset, S&P Cap IQ or directly from the Financial Statement reported by the respective insurer.

Table 5: Overview sources forecast error EPS variable

EPS estimate EPS actual

# % # %

Factset 38 63.33% 29 48.33%

S&P Cap IQ 8 13.33% 19 31.67%

Thomson Reuters I/B/E/S estimates 14 23.33% N/A N/A

Financial statement N/A N/A 12 20.00%

60 100.00% 60 100.00%

Descriptive statistics

The descriptive statistics of the positive changes in Solvency II ratio dataset are presented in Table 6. The independent variables are separated by the dotted lines into the four categories of control variables introduced in section 2.2. Considering the statistics of the positive change in Solvency II ratio independent variable the fact that its mean of 5.19% is larger than its median of 2.67% indicates a positively skewed distribution. Furthermore, as expected, the positive mean (1.83%) and median (1.89%) of the dependent variable, AR, are positively correlated with the change in Solvency II ratio. Table 6: Descriptive statistics positive dataset

Variable Mean Median

Standard deviation

Independent

Total net margin 1.30% -0.20% 13.46%

Insurance Investment Yield 0.52% 0.16% 3.86%

Book value per share -0.40% -0.40% 7.09%

Earnings per share 0.16% 0.15% 2.12%

Growth in net premiums earned 4.69% 1.11% 15.52%

Total Investments / policy reserves 0.53% 0.00% 2.09%

Policy reserves / total assets 0.15% -0.05% 3.41%

Policy reserves / equity 0.46% -0.21% 6.71%

Change in Solvency II ratio 5.19% 2.67% 5.71%

Dependent Abnormal return 1.83% 1.89% 4.22%

The descriptive statistics of the negative changes in Solvency II ratio dataset are presented in Table 7. Considering the statistics of the negative change in Solvency II ratio the size of its mean of -4.69% is similar in absolute size to that of the positive changes in Solvency II ratio (5.19%). However, its standard deviation of 2.47% (versus 5.71%) is considerably lower.

(25)

25

Furthermore, the negative mean (-0.64%) and median (-1.36%) of the dependent variable, AR, are positively correlated with the change in Solvency II ratio and indicate a positively skewed distribution. Table 7: Descriptive statistics negative dataset

Variable Mean Median

Standard deviation

Independent

Total net margin -0.15% 0.41% 5.95%

Insurance Investment Yield 2.41% 0.49% 6.21%

Book value per share 0.70% 1.54% 4.16%

Earnings per share -0.20% 0.00% 1.50%

Growth in net premiums earned 3.58% -1.08% 22.44%

Total Investments / policy reserves 0.20% 0.38% 2.03%

Policy reserves / total assets -0.96% -0.36% 2.29%

Policy reserves / equity 0.85% -0.49% 5.25%

Change in Solvency II ratio -4.69% -4.61% 2.47%

Dependent Abnormal return -0.64% -1.36% 5.05%

In addition, as displayed in Table 8, the descriptive statistics of the change in Solvency II ratio variable are split-up into two datasets (for both the positive and negative dataset) based on whether the Solvency II ratio change was smaller or larger than one standard deviation. As reported above, the applicable standard deviations are 5.71% and 2.47% for the positive and negative changes in Solvency II ratio, respectively. These splits are made because these subsets of data will be used in section 5.1 and 5.2 to test the intuition that larger changes in Solvency II ratio result in larger ARs. The statistics show that while the absolute mean and median of the small (<1 st.dev) changes in Solvency II ratio are comparable the absolute mean and median of the large (>1 st.dev) changes are not. More specifically, the mean and median of the negative subset are -5.81% and -6.35% while those of the positive subset are 12.42% and 12.69%, respectively.

Table 8: Split of the change in Solvency II ratio variable descriptive statistics

Small (<1 st.dev) Large (>1 st.dev) All observations

Negative Positive Negative Positive Negative Positive

Mean -1.68% 1.73% -5.81% 12.42% -4.69% 5.19%

Median -1.50% 1.44% -6.35% 12.69% -4.61% 2.67%

(26)

26

5. Results

5.1 Stock Price Response to Increases in Solvency II ratio

In order to answer sub-question 1.1, the alternative hypothesis H1: µAbnormal return > 0 was tested using a one sample t-test. The alternative hypothesis was tested against the null hypothesis H0: µAbnormal return = 0 given that no ARs are expected if the positive Solvency II ratio changes are not value relevant for investors. As reported in Table 9, the average AR based on 34 observations is 1.83%. The one-sided, one sample t-test performed on the AR has a t-statistic of 2.53 and a p-value of 0.008. Hence, the mean AR is significant at a 1% level. In addition, the dataset was split-up into two datasets based on whether the Solvency II ratio change was smaller or larger than one standard deviation (5.71%) of the Solvency II ratio changes. This was done to test the intuition that larger positive solvency II ratio changes result in larger ARs. The average ARs of the smaller and larger datasets of Solvency II ratio changes are 1.47% and 2.60%, respectively. When examining the level of significance, the AR of subset containing the small changes is significant at a 10% level and the AR of the large changes at a 1% level. Hence, these results support the aforementioned intuition.

Table 9: Results one-sample t-test positive Solvency II ratio changes

Positive Solvency II ratio changes

N AR t-statistic

Full sample: 34 1.83% 2.53***

Sub-samples:

Δ Solvency II ratio <1 st.dev 23 1.47% 1.437*

Δ Solvency II ratio >1 st.dev 11 2.60% 3.770***

*, **, and *** indicate significance at 10%, 5%, and 1%, respectively.

In addition, a nonparametric binomial test was performed as a robustness check. Out of the 34 observations, 61.8% of the ARs had a positive sign (classified as “success” and assigned the value ‘1’) while the remaining 38.2% had a negative sign (assigned the value ‘0’). Subsequently, the proportion of successes (0.618) is compared to the default proportion of 0.5 representing an even distribution of positive and negative ARs. The results, presented in Table 10, indicate that the proportion of ‘successes’ is significantly higher than the default proportion of 0.5 at a 10% level.

Table 10: Nonparametric Binomial test positive Solvency II ratio changes

Category N Observed Proportion Test Proportion

Group 1 1.00 21 0.618* 0.50

Group 2 0.00 13 0.382

Total 34 1.00

* indicates significance at 10%.

The results of both tests provide significant evidence to reject the null hypothesis in favor of the alternative hypothesis. Hence, it is concluded that the mean AR of this dataset is significantly larger than zero.

(27)

27 5.2 Stock Price Response to Decreases in Solvency II ratio

In order to answer sub-question 1.2, the alternative hypothesis H1: µAbnormal return < 0 was tested using a one sample t-test. The alternative hypothesis was again tested against the null hypothesis H0: µAbnormal return = 0. As reported in Table 11, the average AR based on 26 observations is -0.64%. The one-sided, one sample t-test performed on the AR has a t-statistic of -0.65 and a p-value of 0.261. Hence, the mean AR is not significant at a respected level (at least 10%). In addition, the dataset was split-up into two datasets based on whether the Solvency II ratio change was smaller or larger than one standard deviation (2.47%) of the Solvency II ratio changes. The average ARs of the smaller and larger datasets of negative Solvency II ratio changes are 1.27% and -1.35%, respectively. When examining the level of significance, the AR of the subset containing the small changes is insignificant and those of the large changes is significant at a 10% level. Hence, these results support the intuition that more negative changes in Solvency II ratio result into larger negative ARs.

Table 11: Results one-sample t-test negative Solvency II ratio changes

Negative Solvency II ratio changes

N AR t-statistic

Full sample: 26 -0.64% -0.65

Sub-samples:

Δ Solvency II ratio <1 st.dev 7 1.27% 0.477

Δ Solvency II ratio >1 st.dev 19 -1.35% -1.427*

*, **, and *** indicate significance at 10%, 5%, and 1%, respectively.

In addition, a nonparametric binomial test was performed as a robustness check. Out of the 26 observations, 61.5% of the ARs had a negative sign (classified as “success” and assigned the value ‘1’) while the remaining 38.2% had a positive sign (assigned the value ‘0’). When comparing the proportion of successes of 0.615 to the default proportion of 0.5, the results, presented in Table 12, indicate that the proportion of ‘successes’ is significantly higher at a 10% level.

Table 12: Nonparametric Binomial test negative Solvency II ratio changes

Category N Observed Proportion Test Proportion

Group 1 1.00 16 0.615 (*) 0.50

Group 2 0.00 10 0.385

Total 26 1.00

* indicates significance at 10%.

The results of the aforementioned tests are mixed. The overall one sample t-test provides insignificant evidence to reject the null hypothesis however, the subsample of large negative Solvency II ratio changes and the binomial test do provide limited evidence in favor of the alternative hypothesis being significant at 10%.

(28)

28 Asymmetric stock price response

When comparing the results of the one-sample t-test and nonparametric binomial test for the positive and negative datasets it is important to note that both datasets support the intuition that larger Solvency II ratio changes result in larger ARs. Hence, at least part of the asymmetric evidence for ARs can be explained by the fact that in absolute terms the positive dataset contains relatively larger Solvency II ratios changes as supported by Table 8 in section 4.

5.3 Multiple Linear Regression Analysis Increases in Solvency II ratio

First, this section will discuss the estimation results of the full and best fit model as presented in Table 13. Second, the estimation results, across all applicable regressions, of the key independent variables will be discussed. The estimation results of the various regressions are provided in Table 13.

When considering the estimation results of the full and best fit model the adjusted R2s are 0.139 and 0.238 and the AIC values are -3.419 and -3.653. Furthermore, the full model is significant at a 5% level and the best fit model at a 1% level. In order to determine whether the Ordinary Least Squares (‘OLS’) estimation method is the Best Linear Unbiased Estimator (‘BLUE’) certain tests in line with the assumptions of the Gauss-Markov Theorem were performed. First, the mean of the residuals of the full and best fit model were determined to be 0.0031 and 0.0018, respectively, and thus considered to be close to zero. Second, the scatterplots of the residuals of both models showed no clear indication that the residuals are heteroskedastic. Third, the correlograms of the residuals of both models indicated no evidence of autocorrelation in the error term. Fourth, when considering a threshold of |0.70|, based on Tabachnick & Fidell (1996), the cross correlation table of the independent variable does not provide an indication for multicollinearity. Hence, based on the above, together with the assumption that both models estimate a correct linear model, it can be concluded that the assumptions of the Gauss-Markov Theorem are satisfied and that the OLS estimation method is the BLUE.

When focusing on the estimation results in relation to the main variable of interest, the change in Solvency II ratio, it can be observed that its significance is widely supported across the estimated regressions. More specifically, it is significant at a 10% level in the regressions focused on operational efficiency (2), profitability (3), best fit model (7) and close to a 10% level of significance in the regression focused on growth (4) (p-value = 0.1032). Furthermore, it is significant at a 5% level in the coverage/leverage ratios regression (5) and close to a 5% level of significance in the regression including solely the change in Solvency II ratio (1) as an independent variable (p-value = 0.0525). In terms of economic significance, the coefficient is positive, as expected, and ranges from 0.1457 in the full regression to 0.2316 in the regression including the coverage/leverage ratios control variables. In the best fit model, the coefficient is 0.1477 meaning that, a one percent increase in Solvency II ratio results in an AR of 0.15%. Considering the average positive change in Solvency II ratio of 5.19%, as displayed

(29)

29

in Table 6, this would translate into an AR 0.78%. Hence, the empirical results indicate that for the positive dataset Solvency II ratio reporting continues to be value relevant to investors after controlling for IFRS based reporting. When considering the estimation results for the control variables it can be observed that the change in insurance investment yield and the growth in net premiums earned are significant at a 5% level in the regression covering their respective category and the full model. Furthermore, in the best fit model both variables are significant at a 1% level. In addition, in terms of economic significance the coefficients of the change in insurance investment yield variable are positive, as expected, and range from 0.4903 to 0.6283. In the best fit model, the coefficient is 0.5080 meaning that, an absolute increase in the insurance investment yield with one percent results in an AR of 0.51%. The significance and size of the insurance investment yield coefficient can be explained by the fact that, for the insurance companies included in this study, the realized investment income accounts for approximately 25% of the total income. In addition, in general the investment returns of insurance companies are under pressure due to the current low interest rate environment and therefore recovery in the realized returns in most likely well received by investors. The exceptional situation is even recognized by the EIOPA (2016) stress test that classified a continued low interest rate environment as one of the two tested stress scenarios. The coefficients of the growth in net premiums earned are positive, as expected, and range from 0.0978 to 0.1189. In the best fit model, the coefficient is 0.1189 meaning that, a one percent increase in net premiums earned results in an AR of 0.12%. The significance of the growth in net premiums earned coefficient can be explained by the strong average growth realized by the insurance companies during the sample period. More specifically, the average growth of 4.69%, as displayed in Table 6, is substantially larger than the Cumulative Average Growth Rate (‘CAGR’) of 2.87%, in life premiums, from 2011 to 2015, as displayed in Table 15 in appendix A.

Besides the aforementioned control variables, the forecast error of EPS is also included in the best fit model although its coefficient is insignificant at a 10% level. Its coefficient is positive, as expected, and equal to 0.4296.

When further examining the estimation results for the first five regressions it can be observed that the signs of the other control variables are in line with the expectation stated in section 2.2 except for the negative sign of the coverage variable (total investments/policy reserves). In addition, when considering the output for the full model the signs of the coefficients of the total net margin and policy reserves / equity variables flip from positive to negative and from negative to positive, respectively.

(30)

30 Table 13: Results multiple linear regression analysis concerning positive Solvency II ratio changes

This table looks at the determinants of Abnormal Returns (column 1-7). First, column 1 presents the regression results solely including the change in Solvency II ratio as an independent variable. Second, column 2-5 present the regression results including the change in Solvency II ratio and the variables belonging to either of the four categories of control variables. Third, column 6 presents the regression results of the full model including the change in Solvency II ratio and all control variables. Fourth, column 7 presents the regression results of a best fit model based on the Akaike Information Criterion (AIC). For definitions of all variables see section 2.2 on IFRS based control variables. The regressions are based on data points distributed across 5 financial reporting periods between 2015 and 2016 as previously presented in Table 4. The t-statistics are reported in parentheses. *, **, and *** indicate significance at 10%, 5%, and 1%, respectively.

Dependent variable: Abnormal Returns

(1) (2) (3) (4) (5) (6) (7) Change in Solvency II ratio Operational

efficiency Profitability Growth

Coverage/Leverage ratios Full model Best fit model

Total net margin 0.0779 -0.0158

(1.37) (-0.20)

Insurance investment yield 0.4903 0.6283 0.5080

(2.50)** (2.66)** (3.00)***

Book value per share 0.1174 0.1343

(1.10) (0.98)

Earnings per share 0.4397 0.5117 0.4296

(1.28) (1.39) (1.39)

Growth in net premiums earned 0.0978 0.1121 0.1189

(2.17)** (2.52)** (2.90)***

Total investments / policy reserves -0.2001 -0.2249

(-0.47) (-0.56)

Policy reserves / total assets -0.1732 -0.1335

(-0.54) (-0.35)

Policy reserves / equity -0.0438 0.0637

(-0.30) (0.43)

Change in Solvency II ratio 0.1964 0.1670 0.1829 0.1581 0.2316 0.1457 0.1477

(2.01)* (1.69)* (1.96)* (1.68) (2.08)** (1.29) (1.74)*

N 34 34 34 34 34 34 34

R2 _-0.064 _-0.005 _0.147 _0.073 _-0.041 _0.348 _0.307

Adj. R2 _-0.064 _-0.036 _0.062 _0.044 _-0.145 _0.139 _0.238

Akaike Information Criterion -3.401 -3.399 -3.445 -3.479 -3.246 -3.419 -3.653