The Geneva Papers, 2015, 40, (209–231) © 2015 The International Association for the Study of Insurance Economics 1018-5895/2015 www.genevaassociation.org
Measuring Risk Premiums Using Financial Reports and Actuarial Disclosures Jochen Zimmermanna, Stefan Veithb and Johannes Schymczykc a
Faculty of Business Studies and Economics, University of Bremen, Hochschulring 4, Bremen 28359, Germany. Federal Financial Supervisory Authority of Germany (Bundesanstalt für Finanzdienstleistungsaufsicht—BaFin), Marie-Curie-Straße 24-28, Frankfurt am Main 60439, Germany. E-mail: stefan.veith@bafin.de c PricewaterhouseCoopers, Friedrichstraße 14, Stuttgart 70174, Germany. b
Insurance companies increasingly augment their financial reports by releasing actuarial measures— the so-called embedded value—to supply information about the value of their life insurance activities. Both accounting and actuarial measures differ with respect to the timeliness of profit realisation and its reliability, and their performance in yielding information may differ. This paper asks if and how embedded values help in assessing risk premiums. We estimate multifactor market models in the spirit of Fama and French, and find that actuarial disclosures are superior to financial accounting in estimating these risk premiums. They further add information to financial reports as an estimator for growth opportunities. The Geneva Papers (2015) 40, 209–231. doi:10.1057/gpp.2014.17 Keywords: embedded value; financial accounting; multifactor market model; life insurance Article submitted 21 August 2013; accepted 14 April 2014; published online 27 August 2014
Introduction Currently, two methods for informing shareholders about the life insurance business compete: financial accounting and actuarial accounting within the embedded value framework. Although financial reporting is mandatory, insurance companies throughout the European Union increasingly issue voluntary reports based on actuarial measures. The major difference between the two is that actuarial methods use a valuation at market prices and include profits from existing insurance business that is unrealised in traditional financial accounting.1 Supporters of actuarial accounting argue that it captures more of an insurer’s economic value than the corresponding measure of financial accounting.2 Also, embedded values consider the sale of an insurance contract as the productive moment in insurance, not the bearing of risk over time, which would call for a pro-rata realisation of profits. Thus, one can argue that applying actuarial valuation harms information processing: embedded values capitalise parts of future income; they cannot be measured reliably and open up possibilities 1 2
O’Keeffe et al. (2005). Horton (2007).
The Geneva Papers on Risk and Insurance—Issues and Practice
210
for earnings manipulation.3 They may also not reflect the long-term perspective of life insurance properly.4 The lines of the argument become increasingly blurred as financial accounting under International Financial Reporting Standards (IFRS) has evolved from a mainly transaction-based framework to include more and more fair values over the last years—this inclusion of more fair values potentially makes the information reported by IFRS more similar to that reported by embedded values. Although tensions between the financial and actuarial accounting approaches have dominated the debate in the insurance industry for many years, empirical research in this area is rare.5 We contribute to the empirical strand of the debate and analyse whether financial or actuarial accounting information performs better in estimating risk premiums of life insurers. We thus conceptualise risk as the driver of premiums in excess of returns earned on riskless financial papers without making any conjectures about the underlying sources of this risk (operations, information risk, asset liability mismatch, etc.). Our analysis then gauges these premiums based on the Fama and French6 approach: For example, in this study the book-to-market ratios capturing the value/growth opportunities are either constructed based on financial or actuarial accounting information, or put in addition to them. We hypothesise and find the embedded value to yield superior information in estimating the risk premiums as tested in both a three- and a four-factor model. This result is robust towards several controls. Our paper makes three contributions. First, it extends the literature on the information characteristics of actuarial measurement models in insurance accounting. We can take advantage of a currently unique setting in which financial and actuarial measures are published at the same time. Investigating into the use of embedded values is also topical with the current exposure draft to modify the IFRS 4 Insurance Contracts currently in force towards a more value-based perspective. Second, related research shows that actuarial measures are value relevant and increase stock liquidity.7 To our knowledge, we are the first to address the issue whether actuarial information can be used to estimate risk premiums—this can be useful in assessing the costs of capital of life insurance firms. Third, our study adds to the literature by using multifactor market models for accounting research purposes. Existing studies examine the role of additional risk factors formed by characteristics of annual accounts or the effect certain accounting practices have on market pricing,8 but little research exists employing alternative accounting information directly to determine additional risk factors. The remainder of the paper is organised as follows: The next section describes main characteristics of the competing accounting measurement concepts for equity and profit. The subsequent section reviews the literature on size and growth as risk factors to develop a hypothesis. The latter section derives the main models and describes the sample used. The penultimate section contains the main results as well as two additional analyses. The final section concludes the paper. 3 4 5 6 7 8
Serafeim (2008). Meyer (2005). Most prominent contributions are Klumpes (2002); Horton (2007) and Serafeim (2011). Fama and French (1993). Horton (2007) and Serafeim (2011), respectively. Francis et al. (2005) and Ernstberger and Vogler (2008).
Jochen Zimmermann et al Measuring Risk Premiums Using Financial Reports and Actuarial Disclosures
211
The rise of actuarial disclosures Even though all payments between insurer and policyholder are determined at the moment of signing a contract, their exact quantity and timing depend on future realisations. In this setting, policyholders and the company (i.e. its shareholders) have different reporting demands.9 From the policyholders’ perspective, it is important to realise whether the insurer will be able to meet its contractual obligations in the long run or at least during the specific contract period. The shareholders ask whether insurance contracts give rise to profits both in the short and long run and require dividend payments—hence they demand timely information about the corporation as a whole. Both perspectives also conflict in defining the point in time when income accrues: upon signing, when payments become legally enforceable, or upon realisation, when the specific amounts are known with adequate certainty.10 Financial and actuarial accounting approaches answer this question differently, even though both standardisation bodies—the International Accounting Standards Board (IASB) for financial accounting and the CFO Forum for embedded value accounting—set out rules that should be useful for investors.11 Taking a closer look reveals that the IASB tends to map future in- and outflows of the firm’s resources through the perspective of the accrual principle,12 which then is to convey relevant information on (dis)investment decisions. The CFO Forum’s standard, for its part, merely sets out to inform on the present value of future distributable net inflows,13 without requiring a realisation principle. To this end, the specific provisions of financial and actuarial accounting mainly fall apart with respect to the unwinding of profits from insurance contracts. Financial accounting for insurance companies has also been used as a regulatory instrument to foster stability in the insurance sector and, as such, follows a policyholder perspective to guarantee the solvency and going-concern of an insurance company. Consequently, financial accounting typically realises income from insurance contracts upon inception; hence this contract type is per se viewed as an uncompleted transaction with no identifiable value. In subsequent periods, profit accrues as the difference between the premium inflows and anticipated outflows with respect to the final payment that are gradually deferred in an insurance provision. Usually, the final payment is smaller than the expected capital inflows, giving rise to a margin that reflects compensation for risk bearing and entrepreneurial profits which is revealed in a time-proportional manner. In Europe, financial insurance accounting was significantly harmonised via the European Commission’s Insurance Accounting Directive in 1991 (91/674/EEC). It introduced that the acquisition cost is not fully recognised in the period of signing the insurance contract but deferred over its duration, which resulted in less conservative profits. However, it still did not provide fully symmetric revenue recognition, as profits inflate towards the end of a contract.2 More importantly, the directive reinforced the perception of insurance contracts as uncompleted transactions. Still as of today, the current IFRS 4 Insurance Contracts represents a preliminary standard whose measurement rules relate back to national accounting practices
9 10 11 12 13
Horton et al. (2007) and De Mey (2009). Klumpes (1999). IASB FW.OB2-4 as well as MCEV Principle 1 (CFO Forum, 2009b) bear a similar reading. IASB FW.OB13, 17. MCEV.BC6.
The Geneva Papers on Risk and Insurance—Issues and Practice
212
and allow U.S. treatments in some cases. However, the IFRS framework introduced more and more fair value measurements especially when dealing with financial instruments (as set out in IAS 39 and IFRS 9). For that reason, group earnings are not based on a purely transaction-based realisation principle, but are also to reflect market-consistent value changes immediately. To this end, income as reported under IFRS consequently lacks information regarding the value of signed but not yet realised insurance contracts, but tends to include information on value changes with respect to the insurer’s investments. In the meantime, actuarial valuation has been proposed to fill this gap in corporate reporting. It uses the embedded value (EV) concept,14 which gained importance in the U.K. in the 1980s when it was increasingly used for portfolio valuations in business mergers.15 Lacking investor-oriented financial accounting norms for life insurance companies, the industry started publishing actuarial measures as an instrument of external communication in the early 2000s.16 As a consequence, the embedded value framework was gradually refined to suit its new application. In 2004, representatives of the insurance industry published the European Embedded Value (EEV) principles to harmonise the formerly company-specific EV practices for shareholder communication. These principles were refined with more elaborate modelling techniques and renamed market consistent embedded value (MCEV), as issued in 2008. The major changes in the MCEV concern an increased level of items to be disclosed, as well as implementing as many marketbased (or “market consistent”) estimates towards, for example, interest rates and future cash flows as possible.17 In contrast, the EEV framework more strongly relied on firmspecific assumptions which made measurement and estimation dependent on managerial discretion to a larger extent. To this end, the harmonisation of the embedded value methodology was intended to increase comparability across firms.18 Since its early standardisation in 2004, an ever-growing number of life insurance firms, especially in Europe, have disclosed a separate (and sometimes audited) report containing these actuarial measures. Currently, two-thirds of listed European life insurers voluntarily disclose embedded value information. In contrast to financial accounting, the value-based actuarial metric builds on the perception that the profit of an insurance contract accrues as soon as the contract is initially signed.19 The value of the policy is the positive difference between probability weighted future cash in- and outflows discounted with the company’s cost of capital.1 Actuarial accounting is designed to anticipate the expected net profit of any written insurance business, which financial accounting does not do. In this respect, both rule sets differ: in the initial periods of an insurance contract, actuarial accounts accrue the net profit of the contract and, in subsequent years, its annual unwinding.20 Financial accounting rules, on the contrary, do not report on these effects in the first period, but typically give rise to a proportionate realisation of net income in subsequent years. The main difference between the two sets of
14 15 16 17 18 19 20
Originally set out by Anderson (1959). Horton and Macve (1997). Towers Perrin (2005). CFO Forum (2009a, b). Serafeim (2011). Klumpes (2005). Horton et al. (2007).
Jochen Zimmermann et al Measuring Risk Premiums Using Financial Reports and Actuarial Disclosures
213
measurement rules therefore is made up by earlier profit recognition in actuarial accounts. In the broader picture, both types of accounting concepts tend to overlap when fair values of financial instruments give rise to a non-transaction-based interpretation of the financial accounting realisation principle. Considering all aspects, it can be assumed that investors prefer the information inherent in the actuarial value of unrealised business, which constitutes a major part of an insurance company’s value not (yet) revealed in financial accounts.21 This reports in a more timely manner about a value change in their investment, which makes embedded values more useful in decision-making. The above can be illustrated by the following example. Absent perfect markets, the market value of equity (ME) of a life insurance firm is made up of three components. The first component is the book value of equity (BE) that embraces items already realised in financial accounting; in the language of actuarial valuation, the book value of equity corresponds to the value of realised insurance business (VRB). The second element is the actuarial estimate of written but yet unrealised insurance business (VUB). The third part is formed by the investors’ assumptions towards the value of future business (VFB) that is realised in neither reporting framework. These variables are connected via the simple relation stated in Eq. (1): ME ¼ BE + VUB + VFB:
(1)
From an actuarial perspective, the market value of equity is the embedded value plus the value of future business as shown in Eq. (2). Eqs. (1) and (2) are connected, as the embedded value corresponds to the sum of the book value of equity and the value of unrealised business; see Eq. (3): ME ¼ EV + VFB;
(2)
EV ¼ BE + VUB:
(3)
Figure 1 depicts the relationship in the standard case that VUB is positive. This implies that the embedded value usually incorporates a larger share of the market value of equity than financial accounting does. In this normal case, VFB is also positive. However, embedded values do not fully capture the firm’s economic value, since they include neither growth beyond the current fiscal year nor intrinsic goodwill. Since the current IFRS 4 leaves many accounting choices for preparers, the IASB has completely redrafted the standard. For now, IFRS 4 relies on national GAAP in some cases, and IAS 8 permits the use of U.S. GAAP if no international standard or interpretation regulating a specific problem exists. The second exposure draft of this project, ED/2013/7 issued in June 2013, signals a move towards value-based reporting—as did the first exposure draft of 2010. Despite criticism by national standard setters in the first round,22 the IASB did so far not decide to deviate substantially from the initial exposure draft. Similar to the MCEV, the subsequent measurement of insurance contracts builds on expected discounted net cash flows. In a major aspect, the current exposure draft of IFRS 4 differs from the embedded value framework: day-one profits from selling new contracts will not be permitted. If the currently discussed rules come into force, the realisation of financial 21 22
Dickinson and Liedtke (2004). DRSC (2010); ANC (2011) and IASB (2011).
The Geneva Papers on Risk and Insurance—Issues and Practice
214
VFB VUB ME EV BE
Figure 1. Stylised relations between financial and actuarial items. Variable definitions: BE is the book value of equity, VUB is the value of unrealised business, EV is the embedded value, VFB is the value of future business and ME is the market value of equity.
accounting profits will still to a large extent be deferred by means of a contractual service margin.23 The IASB decided that subsequent changes in estimates impact this margin to some degree (the so-called “unlocking”), which was one of the six major topics discussed in the comment phase that concluded in January 2014.24 However, most of the concerned bodies criticised the proposed unlocking as being not far-reaching enough.25 The different processing of unrealised profits in financial and actuarial accounting is also supported by the views of the CFO Forum, which envisages both types of information system as complementary.26 Since this prominent difference is hence likely to remain even after the new IFRS 4 has come into effect, investigating into the economic characteristics of embedded values will remain a topical issue.
Hypothesis development Investigating into the risk premiums of life insurers can focus on the causation or the estimation of risk: The first approach asks which factors might have an influence on the risk premium, like the possibilities for earnings manipulation in financial reporting, the overall accounting quality level adopted by management, or the extent of risk reporting and management.27 Results in this domain have implications for firms in finding strategies to reduce the cost of capital. The second approach investigates into which of the available reporting tools yields more suitable information in estimating risk premiums. This treats the firm as a black box, but has implications for both preparers and users regarding the usefulness of either tool. This paper follows the latter approach, and in doing so, we use empirical market models to measure the effect of two different information sources for risk premiums. 23 24 25 26 27
ED/2013/7 IFRS 4.18. IASB (2014). CFO Forum (2013); DRSC (2013); FRC (2013). CFO Forum (2006). Höring and Gründl (2011).
Jochen Zimmermann et al Measuring Risk Premiums Using Financial Reports and Actuarial Disclosures
215
The pertinent literature distinguishes between systematic and further risk factors.28 Systematic risk cannot be diversified and is captured in the market risk premium, that is, the the asset’s (or portfolio’s) sensitivity towards market returns. There is ample evidence that a single-factor market model such as the CAPM is not very likely to reflect the riskiness of an investment fully, because the further risk factors can be diversified and are not individually priced. ICAPM- or APT-style models were proposed to capture the risk premiums other than inherent in the market portfolio.29 While some authors identified economic variables,30 characteristics like firm size and value/growth dominate as risk factors.31 We also follow the latter route, defining risk as the size and value/growth risk premiums in excess of the market risk premium as part of the Fama and French6 models. This enables us to analyse how financial and actuarial accounting information can help measure these risk premiums. Prior literature consistently documented size effects: Firms with a small absolute market value of equity earn larger stock market returns than firms with a big market capitalisation.32 One explanation is that small firms face adverse borrowing conditions, but this effect could not be consistently traced to a pricing of default risk.33 In addition to size, the book-to-market ratio (i.e. the BE/ME ratio) is a further well-established risk factor. Usually, a growth (value) stock exhibits a low (high) book-to-market ratio. If BE/ME is low, then the financial accounting equity of a firm is small compared with its market value, and this is interpreted as investors anticipating a growth potential which has not yet materialised in accounting.34 This neoclassical interpretation is backed by a behavioural approach which finds investor overreactions towards firm-specific good or bad news leading to over- or underpricing.35 It still remains unclear what size and BE/ME exactly measure. Recent articles demonstrate that they may be proxies for macroeconomic variables such as GDP, demand and inflation, and that they reflect a firm’s investment opportunity set.36 In sum, size and value/growth are both proxies for future firm profitability,6 that is the perspective on risk inherent in this approach focuses on the expected risk premium and not on other non-diversified sources of total firm risk. The gap in literature we address is which information channel better reports about future growth. This is especially interesting in the life insurance industry where market participants have access to two accounting measures. We motivate our analysis by prior research that so far found three different properties with respect to the competing disclosures. First, incentives were investigated why insurance companies disclose actuarial accounts voluntarily. Evidence was found that firms increasingly incline towards disclosing actuarial information with higher amounts of written but not yet realised insurance business.37 Second, evidence was provided that embedded value adopters enjoy higher stock liquidity, a fact pointing towards less information asymmetries
28 29 30 31 32 33 34 35 36 37
Black et al. (1972). Merton (1973) and Ross (1976). For example, Chen et al. (1986); Jorion (1990) and El-Sharif et al. (2005). Fama and French (1992, 1993, 1995, 1996, 1997). Already Banz (1981); Reinganum (1981); more recently O’Brien et al. (2010). Perez-Quiros and Timmermann (2000); Dichev (1998); cf. Vassalou and Xing (2004) Smith and Watts (1992). Lakonishok et al. (1994) and La Porta et al. (1997). Vassalou (2003); Simpson and Ramchander (2008) and Petkova (2006). Klumpes (2002).
The Geneva Papers on Risk and Insurance—Issues and Practice
216
due to actuarial disclosures.18 Finally, prior studies have shown that embedded value information explains (changes in) market values of insurance companies better than financial accounting does.38 This general finding regarding information quality also extends to cases of mergers and acquisitions.39 Given these empirical findings it is astonishing to also find sporadic insurance firms stopping their voluntary embedded value reporting. This behaviour can be explained by the firm-specific reporting costs in the light of economic benefits of the disclosure.40 A further empirical investigation into cost/benefit considerations of actuarial information is, however, beyond the scope of this paper. Consequently, we formulate the following hypothesis bearing on the different information properties of financial and actuarial accounting: Hypothesis:
The estimation of the firm size and value/growth risk premiums is superior when based on embedded value instead of financial accounting information.
“Superiority” is understood in this context as (additional) statistical significance of modified Fama and French6 models. We consider three approaches in forming size and value/growth portfolios: financial accounting only, actuarial accounting only, and actuarial information in addition to financial accounting. Their formulation will be discussed in the subsequent section.
Methodology, data and sample Model specifications To analyse how financial and actuarial information can be used to estimate risk premiums, we compare augmented and single-factor market models. Our benchmark is the simple capital asset pricing model41 (CAPM, here labelled as model (A)), using White42 standard errors to control for heteroskedasticity. It includes the monthly return on the common stock of firm i (Ri), the monthly return on the risk-free asset (Rf), as well as the monthly valueweighted return on the market portfolio (Rm) and takes the following form: (A) Ri - Rf ¼ α + β Rm - Rf + ε: We expect that further factors—in our case: size and growth—should be significantly different from zero. A straightforward way to assess this is by adding the financial accountingbased Fama and French6 factors to the initial model. This approach is widely employed in previous literature, and was applied to insurance when comparing risk factors between various industries.43 It is also found in research on the insurance sector, especially with respect to property-liability insurers.44 To isolate the effect of size, we categorise an insurance firm as either big or small depending on whether it ranges above or below the 38 39 40 41 42 43 44
Horton (2007) and Wu and Hsu (2011). Paetzmann and Lippl (2013). Hail (2011). The CAPM was already applied to the life insurance industry by Hoyt and Trieschmann (1991). White (1980). For example, Fama and French (1997) and Elyasiani et al. (2011). Cummins and Phillips (2005).
Jochen Zimmermann et al Measuring Risk Premiums Using Financial Reports and Actuarial Disclosures
217
median of the sample market capitalisation at the end of every year t. Growth opportunities are captured by the life insurers’ book-to-market values. This ratio is obtained by dividing the book value of equity45 of fiscal year t by the market value as of the beginning of July in t+1. Ranking the sample according to book-to-market, the highest 30 per cent of our firms are selected as the high (value) portfolio, the intermediate 40 per cent as medium portfolio and the lowest 30 per cent as low (growth) portfolio. All sample firms are then attributed to a size and a growth portfolio. Six portfolios emerge at the intersection of the two risk factors, namely small/high, small/medium, small/low, big/high, big/medium and big/low. The valueweighted return of small over big stocks (small minus big portfolio) is the average of all three portfolios comprising small stocks minus the average of all three portfolios comprising big stocks, and captures the size risk premium. Accordingly, the value-weighted return of value over growth stocks (high minus low portfolio) is the difference between the average of the two high portfolios and the average of the two low portfolios, and captures the value/growth risk premium. This procedure yields a size factor that is free from effects driven by growth, and vice versa.6 In sum, the risk factor for size (SMB) is calculated as the monthly valueweighted return of a size factor-mimicking portfolio (conditioned by HML), and the risk factor for value/growth (HML) is calculated as the monthly value-weighted return of a value/ growth factor-mimicking portfolio based on financial accounting book-to-market ratios (conditioned by SMB); both factors are calculated out of the sample. Model (B) is estimated for our life insurance portfolio on a monthly basis using a regression of the following form46: Ri - Rf ¼ α + β Rm - Rf + γ 1 SMB + δ1 HML + ε: (B) Since the financial accounting book value of equity is crucial in estimating both risk premiums, it is worth discussing the underlying economic intuition. If BE/ME is low, we observe a steep difference between both values, indicating growth potential. From a life insurer’s perspective, this means that the sum of VUB and VFB is large (at least in relation to ME). For an outside investor, it is hard to predict to what extent this observed gap is made up of unrealised but already written contracts or of anticipated future business. A direct test for the quality of embedded values is to construct Fama and French6 portfolios on the corresponding actuarial indicator. Repeating the steps for forming SMB and HML outlined above, we replace BE/ME by EV/ME. In doing so, we leave all parameters other than the accounting information source unchanged (i.e. the sample, portfolio formation dates or relative portfolio cut-off points). This leads to six new portfolios at the intersection of size and growth. More specifically, the size factor based on embedded value information (SMBEV) is calculated as the monthly value-weighted return of a size factor-mimicking portfolio (conditioned by HMLEV), and the value/growth factor based on embedded value information (HMLEV) is calculated as the monthly value-weighted return of a value/growth factor-mimicking portfolio based on embedded value-to-market ratios (conditioned by SMBEV). The regression equation differs from the previously described model (B) only by replacing the standard financial accounting-based by actuarial accounting-based variables. Consequently, the 45 46
See description in Fama and French (1993). See, for example, Gompers et al. (2010) and Hearn et al. (2010) for recent studies employing the same model with a similarly small sample.
The Geneva Papers on Risk and Insurance—Issues and Practice
218
equation for model (C) is estimated as follows: Ri - Rf ¼ α + β Rm - Rf + γ 2 SMBEV + δ2 HMLEV + ε:
(C)
Deriving size and growth factors based on embedded values now incorporates achieved profits that are not yet realised in financial accounting, that is, the VUB component. An outside investor is then able to observe directly VFB, which may be informative when estimating the growth potential. Since we examine life insurers that release both types of equity information, it can be argued that investors use all available data in making assumptions about the future.47 Expressed differently: if BE and EV are disclosed to the market, not only the exact value of future business is known, but also, the value of unrealised business can be approximated in a more reliable way. To assess if the difference between the embedded value and the financial accounting book value of equity can be used to estimate growth potentials, we construct a high-minus-low factor based on a VUB/ME ratio. This mimics the return of firms with relatively large amounts of written (and in financial accounting terms, unrealised) accrued actuarial profits over firms with relatively small amounts of that value. Technically, the selection procedure for the embedded value increment (EVI) factor is in line with the one for the HML factor, and it is calculated as the monthly value-weighted return of an embedded value increment factor-mimicking portfolio.48 The equation for the four factor regression (model D) is estimated as follows: (D) Ri - Rf ¼ α + β Rm - Rf + γ 1 SMB + δ1 HML + μEVI + ε: In sum, the three augmented market models assess the available book value information in all possible ways: model (B) asks for the (VUB+VFB)/ME ratio, model (C) for the VFB/ME ratio and model (D) for the VUB/ME ratio. To find evidence for our hypothesis—that embedded value information is superior to financial accounting—we expect three requirements to be fulfilled. First, the size and value/growth risk factors based on other actuarial information exhibit a higher statistical significance than those based on financial accounting. Second, the additional information in actuarial disclosures is manifested in statistically significant risk factors (over and above the risk premium estimated using financial accounting). Finally, risk factors estimated from actuarial accounting induce a significantly higher explanatory power over the single-factor market model. A comparison like this can easily be accomplished by using a Vuong49 test that is valid as long as the dependent variable and sample are the same.50 Based on the prior 47 48
49 50
This also corresponds to the intended usage of embedded values, see CFO Forum (2009a). To form the EVI factor, all equities within the sample are ranked using accounting equity and embedded value data from fiscal year t and the market value as of the beginning of July in t+1. The highest 30 per cent of the firms are selected as the value portfolio, and the lowest 30 per cent as growth portfolio. Following the approach of Francis et al. (2005), the Fama and French (1993) second and third risk factors are based on financial accounting valuation. That is, the additional EVI portfolios are not used to create additional SMB and HML subportfolios, as this would result in 18 portfolios (Ernstberger and Vogler, 2008). Owing to our limited sample size, we are unable to form equations at the intersection of the three named portfolios. Vuong (1989). For two reasons we do not compare the model R2s: First, the coefficient of determination does not play the dominant role in market model evaluation as it does in, for example, value relevance research. Second, we do not inquire into the overall explanatory power of the models, but into the one of explanatory variables.
Jochen Zimmermann et al Measuring Risk Premiums Using Financial Reports and Actuarial Disclosures
219
empirical literature on embedded values, we predict actuarial information to be the superior explanatory variable in all named three requirements. Sample selection and descriptive statistics Our data sample consists of all listed life insurance companies in the EU that voluntarily disclose actuarial information according to the EEV principles in the time period from 2005 to 2010. Our observation period starts in 2005 for two reasons: First, we mitigate issues of lacking comparability in actuarial reports before harmonisation through the EEV framework. Second, by 2005 all life insurers have adopted IFRS, providing comparability of the financial accounting data also. We include those firms in our sample that are categorised as life insurers in Datastream. This database yields 21 life insurance companies listed on European stock exchanges, 14 of which disclose additional financial statements on an EEV basis for the period under consideration.51 More explicitly, we considered the following entities in the remainder: Aegon, Aviva, AXA, Chesnara, CNP Assurances, ING, Irish Life and Permanent, Legal & General, Mediolanum, Old Mutual, Prudential, Standard Life, St James’s Place and Storebrand. The final sample contains 726 firm-month observations. Besides the accounting numbers and stock prices obtained from Datastream, we manually collected all actuarial capital and income numbers from embedded value reports. Further, we used the 30 days Euro Interbank Offered Rate (EURIBOR) as the risk-free rate, and the Datastream Europe portfolio as market index. We acknowledge three major limitations stemming from this selection process: First, the latter part of the sample period falls into the (still ongoing) financial crisis which also adversely affected the stock markets. Since we apply multifactor market models in our analysis, their results may be biased due to these external shocks. Second, our sample does not exclusively consist of unique life insurers; most firms run other business lines, like operating unit trusts. Including these firms is common practice in prior research on embedded value market effects,2 and seems to be unproblematic considering the fact that the average share in revenues stemming from life insurance is at 74 per cent for our sample firms.52 Third—and related—the resulting number of firms included is low. However, this makes sense from an economic perspective since this represents a total population sample of listed life insurance firms in Europe, and it seems to be unproblematic from an econometric perspective given other studies in this domain with similar sample sizes.53 Panel A in Table 1 shows mean values for the raw data of book value of equity, embedded value and market value of equity. The incremental incorporation of the value components of realised, unrealised and future business is reflected in the mean values of the three 51
52
53
We analysed the properties of firms disclosing and not disclosing embedded values. Untabulated results show that both subsamples significantly differ in terms of size (measured as total assets), but not in terms of BE/ME and profitability (measured as return on assets, operating return and financial return). Larger firms thus tend to provide the embedded value information, which we explain by the cost of voluntary disclosures. As disclosing embedded values is invariant to the other major tested firm characteristics, the difference in size is not likely to be material for our analysis, in particular since we do not compare both subsamples. To calculate the equal-weighted sample mean regarding the share of the life business, we used the 2010 financial reports of the sample firms and extracted the total revenues along with those revenues pertaining to life business lines as disclosed in the segment reporting section. Gompers et al. (2010); Hearn et al. (2010).
The Geneva Papers on Risk and Insurance—Issues and Practice
220
Table 1 Descriptive statistics Average Panel A: Raw data (in € million) BE 9,180 EV 11,892 ME 12,005 Average
SD
Min
Max
10,289 10,886 14,621
161 159 138
40,226 38,490 69,620
Low BE/ME
Medium BE/ME
High BE/ME
Panel B: Financial and actuarial accounting book-to-market, formed on size (ME) and BE/ME portfolios Financial accounting equity-to-market (BE/ME) Average 0.982 Small 0.322 0.677 2.313 Big 0.516 0.760 1.108 Actuarial accounting equity-to-market (EV/ME) Average 1.280 Small 0.814 Big 0.979
0.941 1.002
2.509 1.279
Actuarial increment-to-market (EV-BE)/ME Average 0.324 Small Big
0.395 0.098
0.279 0.337
0.494 0.470
Where BE is the book value of equity; EV is the embedded value; ME is the market value of equity.
aggregates, where EV is on the average 29.7 per cent higher than BE, and ME 5.9 per cent higher than EV. Even though the former difference seems to be high—which may to some degree result from the specifics of the sample period—this is a first evidence that our hypothesis holds: embedded values on average capture more current value creation than financial accounting does. Notably, the predicted higher volatility of actuarial over financial capital is not reflected in the standard deviation of the two measures; they are similar. Panel B illustrates the distribution of book-to-market ratios based on financial accounting, actuarial accounting and expressed as embedded value increment. As subsamples, the six classical Fama and French6 portfolios at the intersection of size (i.e. ME) and BE/ME are used. For the sake of comparability, these portfolios remain unchanged here for all three ratios. The average financial accounting book-to-market ratio is close to 1. We attribute this to the fact that during the financial crisis, the life insurers’ market values were more strongly affected by a downturn than the book values were. The finer low-medium-high breakdown shows that the portfolio means vary between 0.322 and 2.313. Analysing size effects, big firms have (compared with small firms) a higher BE/ME in the low and medium portfolios. We do not observe consistently higher ratios for small firms as Fama and French6 did. The actuarial accounting book-to-market ratio shows the same distribution as BE/ME, but always with higher values. Most notably, the average EV/ME equals 1.280. This implies that investors predicted for the sample that the value of future business is negative. This can
Jochen Zimmermann et al Measuring Risk Premiums Using Financial Reports and Actuarial Disclosures
221
indeed be the case when interest rates remain low and guaranteed payments are (relatively) high, which was the case during the financial crisis. Finally, the actuarial incrementto-market ratio (EV-BE)/ME is always positive. Our finding is in line with the assumption that actuarial accounting incurs profits that are not yet realised in financial accounting: there is a positive value of unrealised business, and this effect is largest for low BE/ME firms.
Empirical findings Main model Panel A of Table 2 shows the regression estimates for our main models. In column 1, the single factor CAPM is set out as the benchmark model (A). The returns of the life insurance firms in the sample display an exposure towards the market portfolio of 1.07 and an explanatory power of 60.3 per cent. These results are in line with prior literature of estimating asset pricing models for the life insurance sector.54 Model (B) is the Fama and French6 three-factor regression, as displayed in column 2. The results are similar with respect to the beta coefficient and explanatory power. Out of the Fama and French6 factors, HML shows a significantly negative risk premium, whereas SMB is insignificant. The negative coefficient for HML can be interpreted to mean that investors believe growth firms outperform value firms during the observation period. This relation generally should be positive, but also a changing preference of the market in different time periods can be assumed, reflecting changing economic environments in which investors have altering risk preferences.55 The alternative three-factor model (C) using embedded value information is displayed in the third column. As opposed to the previous regression, both additional risk premiums— SMBEV and HMLEV—are associated with highly significant coefficients in this model. Again, the value/growth premium remains negative, affirming the market’s prediction during the observation period. The size premium is positive indicating that the market estimates small firms outperform big firms. The column on the right-hand side of Table 2, Panel A, reports model (D), the specification incorporating information about the value of written but not yet realised insurance business as a fourth risk premium, EVI. Adding the actuarial information to the otherwise financial accounting-based risk factors changes their estimates materially: SMB becomes significant and positive, similar to employing embedded values for portfolio selection. HML remains strongly negative, but the factor loading increases steeply compared to model (B). The EVI factor also displays a significantly negative loading; this is in line with a relatively high (low) EVI indicating little (high) future growth potential. Below we will present an additional analysis whether a low EVI factor truly reveals growth opportunities. All three models show an intercept statistically not different from zero, indicating that the model is in equilibrium.6
54 55
Chen et al. (2001). Fama and French (1996).
The Geneva Papers on Risk and Insurance—Issues and Practice
222
Table 2 Risk estimates from financial and actuarial accounting Model (A) Panel A: Regression results Intercept 0.003 (0.578) β (Rm-Rf) 1.071*** (25.957) γ1SMB δ1HML
Model (B)
Model (C)
Model (D)
0.004 (0.864) 1.093*** (26.989) 0.018 (0.539) −0.041** (−2.150)
0.001 (0.121) 1.042*** (26.040)
0.001 (0.249) 1.072*** (27.444) 0.081** (2.249) −0.077*** (−3.497)
γ2SMBEV
0.052*** (3.555) −0.051*** (−3.414)
δ2HMLEV μEVI N Adj. R2 F
726 0.603 673.743***
Panel B: Vuong (1989) tests Model (B) compared with model (A) Model (C) compared with model (A) Model (D) compared with model (A)
726 0.604 272.09***
−0.142** (−2.550) 726 0.609 207.544***
726 0.613 247.474***
−1.2854 −2.3303** −1.9199*
This table reports coefficients and values for two-tailed t-tests estimated for a linear regression with White (1980) standard errors. *, ** and *** denote levels of significance of 0.10, 0.05 and 0.01, respectively. Model ð AÞ : Ri - Rf ¼ α + β Rm - Rf ; Model ðBÞ : Ri - Rf ¼ α + β Rm - Rf + γ 1 SMB + δ1 HML + ε; Model ðC Þ : Ri - Rf ¼ α + β Rm - Rf + γ 2 SMBEV + δ2 HMLEV + ε; Model ðDÞ : Ri - Rf ¼ α + β Rm - Rf + γ 1 SMB + δ1 HML + μEVI + ε; where Ri−Rf is the monthly excess return of company i over the risk-free asset (30 days EURIBOR); Rm−Rf is the monthly excess return of the market portfolio (DataStream Euro) over the risk-free asset; SMB is the monthly value-weighted return of a size factor-mimicking portfolio (conditioned by HML); HML is the monthly value-weighted return of a value/growth factor-mimicking portfolio based on financial accounting book-tomarket ratios (conditioned by SMB); SMBEV is the monthly value-weighted return of a size factor-mimicking portfolio (conditioned by HMLEV); HMLEV is the monthly value-weighted return of a value/growth factormimicking portfolio based on embedded value-to-market ratios (conditioned by SMBEV); EVI is the monthly return of an incremental information in actuarial accounting (defined as (EV-BE)/ME) factor-mimicking portfolio.
Panel B of Table 2 illustrates the statistical tests regarding the differences in explanatory power of the respective independent variables. As to the financial accounting-based estimation of risk premiums (model B), we do not find a significant difference over the CAPM using a Vuong49 test. Using in turn the embedded value-based modification (model C) yields an increment over the market portfolio that is significant at the 5 per cent level. A replacement of the accounting book value by its actuarial counterpart increases the
Jochen Zimmermann et al Measuring Risk Premiums Using Financial Reports and Actuarial Disclosures
223
explanatory power of the risk premiums. Similarly, a significant increase in explanatory power (at the 10 per cent level) is found when estimating these premiums from financial and actuarial data (model D). Based on our hypothesis, we derived three statistical requirements for identifying quality differentials for alternative accounting methods. First, we assumed that risk premiums based on the actuarial accounting rule set should exhibit a higher statistical significance than those based on financial accounting. We find that this is the case for HML (significant at the 5 per cent level for financial accounting vs the 1 per cent level for embedded values), and even stronger for the SML factor. Size only gains significance when the underlying portfolio is constructed at the intersection of embedded values (i.e. tested as SMBEV). Our second requirement was that additional risk premiums estimated based on embedded value information are statistically significant. We find this to be the case: If EVI is added, the further information constitutes a significant coefficient and also increases the significance of the financial accounting-based SMB factor. And third, we find risk factors based on actuarial information to yield a statistically significant higher explanatory power than financial accounting-based risk factors. With respect to our sample, we find that actuarial measures are better able to explain the value/growth and size risk premiums for this kind of stocks. Relying on actuarial information is directly informative about the future earnings potential of the life insurance collective. This is empirical evidence for the notion that embedded values better report on business yet unrealised in financial accounting, giving rise to a better estimation of future growth potentials. This simple regression approach estimates risk premiums at firm level. However, most firms in our sample run at least one additional business line, for example, the management of unit trusts. For this reason, the aggregate risk estimates may vary with the exposure of every business line. One way to overcome this issue is using full-information industry betas, assuming that a firm-level beta is the weighted average of the divisional betas.56 Cummins and Phillips44 present a recent application of this approach to the insurance sector. Extending their contribution, we run a single-factor, a Fama and French6 three-factor and an embedded value three-factor, full-information beta model. Untabulated results corroborate our main findings: Using embedded value instead of financial accounting information increases the significance of these risk premium estimations.
Additional analyses Controlling for growth To examine the robustness of our results further, we investigate the question as to whether portfolios based on embedded values truly separate the firms into growth and value subsamples. Stated differently: Do embedded values yield robust information about the growth risk premium? In both three-factor models, the negative loading on the HML and HMLEV factors indicates the investors’ prediction that growth firms will outperform value stocks. The fourfactor model (model D) captures growth opportunities in both, negative HML and EVI factors. One proxy for future profits might be the embedded value flow measure value of new 56
Kaplan and Peterson (1998).
The Geneva Papers on Risk and Insurance—Issues and Practice
224
business (VNB) as disclosed in the actuarial report. This component of actuarial income maps additional information regarding signed policies in the fiscal year under scrutiny, which should then allow more precise predictions towards potential future growth opportunities. To analyse this relation, we construct an additional factor mimicking the return of firms with a high amount of income from new business over firms with a low contribution from new business to income. We use the value of new business return on (the same year’s) financial accounting equity to ensure comparability among the sample firms. According to this VNB measure, all firms are ranked and sorted into high, medium and low portfolios, applying the same sorting methodology as for the previous value/growth factors. In this respect, firms in this high (low) portfolio exhibit relatively high (low) new business-related actuarial earnings. The factor for growth (VNB) is calculated as the monthly value-weighted return of a value of new business factor-mimicking portfolio; again, we do not relate this additional factor with the size and growth Fama and French6 factors. If actuarial accounting rates of return are a proxy for future growth, we expect VNB to be significantly positive. The augmented regressions (E) and (F) take the following form: Ri - Rf ¼ α + β Rm - Rf + γ 1 SMB + δ1 HML + μEVI + νVNB + ε; (E) Ri - Rf ¼ α + β Rm - Rf + γ 2 SMBEV + δ2 HMLEV + νVNB + ε: (F) Table 3 reports the results of the two models including the value of new business as an additional factor for the growth risk premium. In both models, the VNB factor is highly significant and weakly positive, as predicted. All other slopes show unchanged signs and significance, indicating an overall robustness of our main findings. These findings support the results of the main model, since the positive relation of the value of new business factor is consistent with our findings indicating the assumption that growth life insurances outperformed value stocks during the estimation period. The result supports our assumption that actuarial measures can be used to select value/growth portfolios, and in doing so are more precise than accounting data. Controlling for institutional effects, moments and seasonality In this section, the main model is augmented to control for three further issues: Specific institutional factors, the risk premium on moments and seasonality effects. To begin with the institutional factors, when comparing the embedded value with the financial accounting framework, a shift in the time series becomes evident: In 2008, the MCEV superseded the EEV concept. This may have given rise to augmented information quality of embedded value disclosures.18 Additionally, a cross-sectional effect is apparent: the current IFRS 4 allows the use of national and U.S. GAAP to the discretion of management. To control for both effects, we include a 2008 dummy as well as firm fixed effects into our regressions. Apart from the premiums on value and growth, a third factor was documented explaining the cross-sectional variance of returns: the price momentum. It was shown that shares which have performed well in the recent past maintain their outperformance for several months.57 This is usually attributed to an overreaction of investors, that is, a behavioural effect.58 57 58
Jegadeesh and Titman (1993). De Bondt and Thaler (1985).
Jochen Zimmermann et al Measuring Risk Premiums Using Financial Reports and Actuarial Disclosures
225
Table 3 Additional analysis controlling for actuarial income
Intercept β (Rm−Rf) γ1SMB δ1HML
Model (E)
Model (F)
0.002 (0.441) 1.077*** (27.587) 0.117*** (2.689) −0.113*** (−3.627)
0.001 (0.151) 1.032*** (25.549)
γ2SMBEV
0.074*** (3.778) −0.058*** (−3.654)
δ2HMLEV μEVI νVNB N Adj. R2 F
−0.186*** (−2.910) 0.007*** (3.098) 726 0.611 167.42***
0.007*** (3.142) 726 0.615 188.744***
This table reports coefficients and values for two-tailed t-tests estimated for a linear regression with White (1980) standard errors. *, ** and *** denote levels of significance of 0.10, 0.05 and 0.01, respectively. Model ðEÞ : Ri - Rf ¼ α + β Rm - Rf + γ 1 SMB + δ1 HML + μEVI + νVNB + ε; Model ðF Þ : Ri - Rf ¼ α + β Rm - Rf + γ 2 SMBEV + δ2 HMLEV + νVNB + ε; where Ri−Rf is the monthly excess return of company i over the risk-free asset (30 days EURIBOR); Rm−Rf is the monthly excess return of the market portfolio (DataStream Euro) over the risk-free asset; SMB is the monthly valueweighted return of a size factor-mimicking portfolio (conditioned by HML); HML is the monthly value-weighted return of a value/growth factor-mimicking portfolio based on financial accounting book-to-market ratios (conditioned by SMB); SMBEV is the monthly value-weighted return of a size factor-mimicking portfolio (conditioned by HMLEV); HMLEV is the monthly value-weighted return of a value/growth factor-mimicking portfolio based on embedded value-to-market ratios (conditioned by SMBEV); EVI is the monthly return of an incremental information in actuarial accounting (defined as (EV-BE)/ME) factor-mimicking portfolio; VNB is the monthly value-weighted return of a value of new business factor-mimicking portfolio.
Carhart (1997) is the first to derive a portfolio mimicking this buying strategy. To add the factor WML (winners minus losers) to the previously defined regressions controlling for actuarial accounting information, we construct two subsamples:59 the winner portfolio containing the average returns of those sample firms within the highest 30 per cent throughout an 11-month period ending one month prior to t, and the loser portfolio with the lowest 30 per cent.60 Every month, the portfolios of the momentum risk factor WML are calculated as the monthly equal-weighted return of a price momentum factor-mimicking portfolio, while predicting WML to be consistently positively priced. Additionally, the 59
60
See for recent applications of the momentum factor, e.g. L'Her et al. (2004), Chordia and Shivakumar (2006) and Fama and French (2010). Carhart (1997).
The Geneva Papers on Risk and Insurance—Issues and Practice
226
augmented models (G) and (H) include a dummy for year 2008 (D2008) as well as firm fixed effects (FIRMi), and are written in the following ways: X α2;i FIRMi + β Rm - Rf + γ 1 SMB Rf ¼ α0 + α1 D2008 + + δ1 HML + ωWML + μEVI + νVNB + ε; Ri - Rf ¼ α0 + α1 D2008 +
X
ðGÞ
α2;i FIRMi + β Rm - Rf + γ 2 SMBEV
+ δ2 HMLEV + ωWML + νVNB + ε:
ðHÞ
As size, value/growth and momentum were documented to explain stock returns crosssectionally, seasonality was found to be an issue from a time-series perspective. Several contributions showed that the U.S. stock market exhibits statistically significantly higher returns in Januaries as compared with the rest of the calendar year61; the same result was found for April in the U.K.62 Most consistent evidence was found for the so-called tax-loss selling hypothesis arguing that investors sell loss firms at fiscal year-end to lower their income tax burden, and that prices rebounce one month after that.63 Since seasonality seems to be a major driver for the size and value/growth risk premiums,64 we test whether this effect also applies to our sample with embedded value-based information. To do so, we run three versions of models (G) and (H) each: one with data from the complete calendar year, one excluding January and one excluding January and April since our sample also comprises U.K.-based life insurers.65 If the previously documented seasonality is also an issue in our analysis, we expect varying significance for our coefficients. Table 4 summarises the results for the named six specifications. Model (G) incorporates financial accounting-based size and growth factors, an embedded value increment and an actuarial income-based control for growth—augmented by a factor for momentum; this latter additional coefficient generally carries the expected positive sign. Thus having added WML, the 2008 MCEV dummy as well as firm fixed effects does not change the estimates of the regression parameters materially. Compared with the results of the prior section, neither the full-year nor the non-January models show qualitatively different coefficients. Excluding both January and April makes the EVI effect insignificant. This is in line with prior literature that turn-of-the-(tax)-year effects inter alia drive the growth risk factor loading. Model (H) uses one actuarial accounting-based size, and two growth factors as well as a control for the momentum effect. In all three specifications, the addition of WML does not alter the other coefficients’ impact. Seasonality does not play a role with respect to the embedded valuebased risk factors, but does for the momentum effect when excluding January and April. In sum, we document that estimating risk premiums from actuarial accounts is robust to the control for institutional effects and momentum trading strategies. We also find weak seasonality issues for both the momentum and embedded value increment factors. However, this result is less prominent than prior literature suggests. 61 62 63 64 65
For example, Rozeff and Kinney (1976); Tinic and West (1984) and Chordia and Shivakumar (2006). Gultekin and Gultekin (1983) and Chen and Singal (2004). See for a review—also for other hypotheses on seasonality in returns—Chen and Singal (2004). Daniel and Titman (1997). Ours is similar to the approach of Chan and Faff (2003).
Jochen Zimmermann et al Measuring Risk Premiums Using Financial Reports and Actuarial Disclosures
227
Table 4 Additional analysis controlling for institutional effects, moments and seasonality Model (G)
Intercept α1D2008 β (Rm-Rf) γ1SMB δ1HML γ2SMBEV
Model (H)
Full
Non-JAN
Non-JAN, Non-APR
Full
Non-JAN
Non-JAN, Non-APR
−0.009** (−2.123) 0.021** (2.015) 1.027*** (26.399) 0.171*** (3.825) −0.124*** (−4.081)
−0.011** (−2.409) 0.028** (2.489) 1.045*** (25.914) 0.164*** (3.616) −0.125*** (−4.061)
−0.012** (−2.412) 0.029** (2.458) 1.046*** (24.470) 0.136*** (2.969) −0.105*** (−3.455)
−0.009** (−2.038) 0.025** (2.259) 1.007*** (26.268)
−0.010** (−2.375) 0.031*** (2.750) 1.020*** (25.684)
−0.012*** (−2.679) 0.036*** (3.013) 1.033*** (25.497)
0.072*** (3.656) −0.036** (−2.403) 0.204** (2.417)
0.069*** (3.383) −0.038** (−2.466) 0.200** (2.346)
0.074*** (3.650) −0.045*** (−3.057) 0.077 (0.830)
0.006*** (2.889) 726 0.622 169.465***
0.006** (2.538) 670 0.619 160.132***
0.006** (2.526) 614 0.616 149.586***
δ2HML
EV
ωWML μEVI νVNB N Adj. R2 F
0.305*** (3.710) −0.102** (−2.117) 0.008*** (3.324) 726 0.625 159.698***
0.288*** (3.476) −0.116** (−2.383) 0.008*** (3.188) 670 0.623 151.016***
0.212** (2.254) −0.080 (−1.602) 0.006** (2.506) 614 0.614 138.136***
This table reports coefficients and values for two-tailed t-tests estimated for a linear regression with White (1980) standard errors. *, ** and *** denote levels of significance of 0.10, 0.05 and 0.01, respectively. X Model ðGÞ : Ri - Rf ¼ α0 + α1 D2008 + α2;i FIRMi + β Rm - Rf + γ 1 SMB + δ1 HML + ωWML + μEVI + νVNB + ε; X α2;i FIRMi + β Rm - Rf + γ 2 SMBEV
Model ðH Þ : Ri - Rf ¼ α0 + α1 D2008 +
+ δ2 HMLEV + ωWML + νVNB + ε; where Ri−Rf is the monthly excess return of company i over the risk-free asset (30 days EURIBOR); D2008 is a dummy for year 2008; FIRMi is the fixed effect for firm i; Rm−Rf is the monthly excess return of the market portfolio (DataStream Euro) over the risk-free asset; SMB is the monthly value-weighted return of a size factor-mimicking portfolio (conditioned by HML); HML is the monthly value-weighted return of a value/growth factor-mimicking portfolio based on financial accounting book-to-market ratios (conditioned by SMB); WML is the monthly equalweighted return of a price momentum factor-mimicking portfolio; SMBEV is the monthly value-weighted return of a size factor-mimicking portfolio (conditioned by HMLEV); HMLEV is the monthly value-weighted return of a value/ growth factor-mimicking portfolio based on embedded value-to-market ratios (conditioned by SMBEV); EVI is the monthly return of an incremental information in actuarial accounting (defined as (EV-BE)/ME) factor-mimicking portfolio; VNB is the monthly value-weighted return of a value of new business factor-mimicking portfolio. The results on firm fixed effects remain unreported for the sake of parsimony.
Conclusion This paper reviewed the role of different profit concepts for life insurers from a capital market valuation perspective. The life insurance industry uses two reporting systems: financial accounting and the actuarial embedded value. Whereas the former defers profit recognition and prorates ensuing profits, the latter relies on expected values and is said to report in a more timely manner about value creation.
The Geneva Papers on Risk and Insurance—Issues and Practice
228
In particular, we hypothesised that embedded value information is superior in estimating the risk premiums vis-à-vis financial accounting disclosures. Our results suggest this hypothesis holds. This is due to the fact that actuarial accounting includes the value of written but not yet realised insurance business. Further, this additional information reveals parts of future earnings that cannot be derived from financial accounts. Our results also hold when using the information about value creation from the actuarial income statement as alternative growth approximation and when controlling for institutional effects, momentum strategies and seasonality. Our empirical results contribute to the debate on value-based measurement models for insurance contracts. We find evidence that actuarial models are a better source of information when expected returns and their resulting risk premium are to be estimated. Our study also extends the literature of the incorporation of multifactor models in empirical accounting research. We find evidence that the inclusion of additional factors based on accounting figures can lead to an improvement of the performance of market models. Still, our investigation also has limitations. Our models implicitly assume that the risk premiums are systematic by nature. While this hypothesis is in line with previous work,66 it still remains a matter of debate in current accounting research. Secondly, it is well documented that market models exhibit instabilities with respect to the parameter estimates.67 We constrained our study to the European setting and to start only in 2005 to provide for a maximum intertemporal and between-firm comparability. It is also evident that the second half of our sample period falls into the financial crisis which caused major turbulences on the European stock markets. Finally, the sample of firms analysed is limited due to the small number of listed life insurers and by the extent of voluntary embedded value disclosures. Taking together all issues the findings of this study may be limited to the sample under scrutiny. Once a longer time-series is available, future research can address coefficient stability issues as well as the impact of the crisis on market models. This will show to which extent embedded value disclosures offer a generally more robust measure for risk premiums than financial accounting.
Acknowledgements The authors would like to thank workshop participants at the EAA Annual Congress 2011, the referees and workshop participants of the VHB Annual Congress 2011, as well as Kerstin Lopatta for their valuable comments and suggestions on earlier versions of the paper. The views expressed in this paper are solely those of the authors and not necessarily those of the Federal Financial Supervisory Authority of Germany or of PricewaterhouseCoopers.
References ANC (2011) Comment Letter concerning ED IASB Insurance Contracts (Autorité des Normes Comptables). Anderson, J.C. (1959) ‘Gross premium calculations and profit measurement for non-participating insurance’, Transactions of Society of Actuaries 11(30): 357–420. Banz, R.W. (1981) ‘The relationship between return and market value of common stocks’, Journal of Financial Economics 9(1): 3–18.
66 67
Botosan (2006); Easley and O’Hara (2004) and Leuz and Verrecchia (2004). Brooks et al. (1998).
Jochen Zimmermann et al Measuring Risk Premiums Using Financial Reports and Actuarial Disclosures
229 Black, F., Jensen, M. and Scholes, M. (1972) ‘The capital asset pricing model: Some empirical tests’, in M. Jensen (ed.) Studies in the Theory of Capital Markets, New York: Praeger Publishers, pp. 79–121. Botosan, C.A. (2006) ‘Disclosure and the cost of capital: what do we know?’ Accounting and Business Research 36(Special Issue): 31–40. Brooks, R.D., Faff, R.W. and Ariff, M. (1998) ‘An investigation into the extent of beta instability in the Singapore stock market’, Pacific-Basin Finance Journal 6(1–2): 87–101. Carhart, M.M. (1997) ‘On persistence of mutual fund performance’, The Journal of Finance 52(1): 57–82. CFO Forum (2006) Elaborated Principles for an IFRS Phase II Insurance Accounting Model, Elaborated Principles and Basis for Conclusions, http://www.cfoforum.nl/letters/elaborated_principles.pdf. CFO Forum (2009a) Market Consistent Embedded Value: Basis for Conclusions, http://www.cfoforum.nl/ downloads/MCEV_Basis_for_Conclusions_October_2009.pdf. CFO Forum (2009b) Market Consistent Embedded Value: Principles, http://www.cfoforum.nl/downloads/MCEV_ Principles_and_Guidance_October_2009.pdf. CFO Forum (2013) Comment letter concerning Exposure Draft ED/2013/7 Insurance Contracts. Chan, H.W. and Faff, R.W. (2003) ‘An investigation into the role of liquidity in asset pricing: Australian evidence’, Pacific-Basin Finance Journal 11(5): 555–572. Chen, C.R., Steiner, T.L. and White, A.M. (2001) ‘Risk taking behaviour and managerial ownership in the United States life insurance industry’, Applied Financial Economics 11(2): 165–171. Chen, H. and Singal, V. (2004) ‘All things considered, taxes drive the January effect’, The Journal of Financial Research 27(3): 351–372. Chen, N.-F., Roll, R. and Ross, S.A. (1986) ‘Economic forces and the stock market’, The Journal of Business 59(3): 383–403. Chordia, T. and Shivakumar, L. (2006) ‘Earnings and price momentum’, Journal of Financial Economics 80(3): 627–656. Cummins, J.D. and Phillips, R.D. (2005) ‘Estimating the cost of equity capital for property-liability insurers’, The Journal of Risk and Insurance 72(3): 441–478. Daniel, K. and Titman, S. (1997) ‘Evidence on the characteristics of cross sectional variation in stock returns’, The Journal of Finance 52(1): 1–33. De Bondt, W.F.M. and Thaler, R. (1985) ‘Does the stock market overreact?’ Journal of Finance 40(3): 793–805. De Mey, J. (2009) ‘Reporting on the financial performance of life insurers’, The Geneva Papers on Risk and Insurance—Issues and Practice 34(2): 228–241. Dichev, I.D. (1998) ‘Is the risk of bankruptcy a systematic risk?’ The Journal of Finance 53(3): 1131–1147. Dickinson, G. and Liedtke, P.M. (2004) ‘Impact of a fair value financial reporting system on insurance companies: A survey’, The Geneva Papers on Risk and Insurance—Issues and Practice 29(3): 540–581. DRSC (2010) Comment Letter concerning Exposure Draft ED/2010/8 Insurance Contracts (Accounting Standards Committee of Germany). DRSC (2013) Comment letter concerning Exposure Draft ED/2013/7 Insurance Contracts (Accounting Standards Committee of Germany). Easley, D. and O'Hara, M. (2004) ‘Information and the cost of capital’, The Journal of Finance 59(4): 1553–1583. El-Sharif, I., Brown, D., Burton, B., Nixon, B. and Russell, A. (2005) ‘Evidence on the nature and extent of the relationship between oil prices and equity values in the UK’, Energy Economics 27(6): 819–830. Elyasiani, E., Mansur, I. and Odusami, B. (2011) ‘Oil price shocks and industry stock returns’, Energy Economics 33(5): 966–974. Ernstberger, J. and Vogler, O. (2008) ‘Analyzing the German accounting triad—“Accounting premium” for IAS/ IFRS and U.S. GAAP vis-à-vis German GAAP?’ The International Journal of Accounting 43(4): 339–386. Fama, E.F. and French, K.R. (1992) ‘The cross-section of expected stock returns’, The Journal of Finance 47(2): 427–465. Fama, E.F. and French, K.R. (1993) ‘Common risk factors in the returns on stocks and bonds’, The Journal of Financial Economics 33(1): 3–56. Fama, E.F. and French, K.R. (1995) ‘Size and book-to-market factors in earnings and returns’, The Journal of Finance 50(1): 131–155. Fama, E.F. and French, K.R. (1996) ‘Multifactor explanations of asset pricing anomalies’, The Journal of Finance 51(1): 55–84. Fama, E.F. and French, K.R. (1997) ‘Industry costs of equity’, Journal of Financial Economics 43(2): 153–193.
The Geneva Papers on Risk and Insurance—Issues and Practice
230 Fama, E.F. and French, K.R. (2010) ‘Luck versus skill in the cross-section of mutual fund returns’, The Journal of Finance 65(5): 1915–1947. Francis, J., LaFond, R., Olsson, P. and Schipper, K. (2005) ‘The market pricing of accruals quality’, Journal of Accounting and Economics 39(2): 295–327. FRC (2013) Comment letter concerning Exposure Draft ED/2013/7 Insurance Contracts (Financial Reporting Council of the UK). Gompers, P.A., Ishii, J. and Metrick, A. (2010) ‘Extreme governance: An analysis of dual-class firms in the United States’, The Review of Financial Studies 23(3): 1051–1088. Gultekin, M.N. and Gultekin, N.B. (1983) ‘Stock market seasonality: International evidence’, Journal of Financial Economics 12(4): 469–481. Hail, L. (2011) ‘Discussion of consequences and institutional determinants of unregulated corporate financial statements: Evidence from embedded value reporting’, Journal of Accounting Research 49(2): 573–594. Hearn, B., Piesse, J. and Strange, R. (2010) ‘Market liquidity and stock size premia in emerging financial markets: The implications for foreign investment’, International Business Review 19(5): 489–501. Höring, D. and Gründl, H. (2011) ‘Investigating risk disclosure practices in the European insurance industry’, The Geneva Papers on Risk & Insurance—Issues & Practice 36(3): 380–413. Horton, J. (2007) ‘The value relevance of “realistic reporting”: Evidence from UK life insurers’, Accounting and Business Research 37(3): 175–197. Horton, J. and Macve, R. (1997) UK Life insurance: Accounting for Business Performance, London: FT Financial Publishing. Horton, J., Macve, R. and Serafeim, G. (2007) An Experiment in ‘Fair Value’Accounting? The State of the Art in Research and Thought Leadership on Accounting for Life Assurance in the UK and Continental Europe, London: Centre for Business Performance, ICAEW. Hoyt, R.E. and Trieschmann, J.S. (1991) ‘Risk/return relationships for life-health, property-liability, and diversified insurers’, The Journal of Risk and Insurance 58(2): 322–330. IASB (2011) Effect of board redeliberations on ED Insurance Contracts, International Accounting Standards Board (IASB) Staff Paper. IASB (2014) Insurance contracts: Outreach and comment letter analysis, International Accounting Standards Board (IASB) Staff Paper. Jegadeesh, N. and Titman, S. (1993) ‘Returns to buying winners and selling losers: Implications for stock market efficiency’, The Journal of Finance 48(1): 65–91. Jorion, P. (1990) ‘The exchange-rate exposure of U.S. multinationals’, The Journal of Business 63(3): 331–345. Kaplan, P.D. and Peterson, J.D. (1998) ‘Full-information industry betas’, Financial Management 27(2): 85–93. Klumpes, P.J.M. (1999) ‘Measuring the profitability of UK proprietary life insurers’, The British Accounting Review 31(2): 185–204. Klumpes, P.J.M. (2002) ‘Incentives facing lie insurance firms to report actuarial earnings: Evidence from Australia and the UK’, Journal of Accounting, Auditing & Finance 17(3): 237–258. Klumpes, P.J.M. (2005) ‘Managerial use of discounted cash-flow or accounting performance measures: Evidence from the U.K. life insurance industry’, The Geneva Papers on Risk and Insurance—Issues and Practice 30(1): 171–186. L'Her, J.-F., Masmoudi, T. and Suret, J.-M. (2004) ‘Evidence to support the four-factor pricing model from the Canadian stock market’, Journal of International Financial Markets, Institutions and Money 14(4): 313–328. La Porta, R., Lakonishok, J., Shleifer, A. and Vishny, R. (1997) ‘Good news for value stocks: Further evidence on market efficiency’, The Journal of Finance 52(2): 859–874. Lakonishok, J., Shleifer, A. and Vishny, R.W. (1994) ‘Contrarian investment, extrapolation, and risk’, The Journal of Finance 49(5): 1541–1578. Leuz, C. and Verrecchia, R.E. (2004) Firms' capital allocation choices, information quality, and the cost of capital, working paper, Univsity of Pennsylvania. Merton, R.C. (1973) ‘An intertemporal capital asset pricing model’, Econometrica 41(5): 867–887. Meyer, L. (2005) ‘Insurance and international financial reporting standards’, The Geneva Papers on Risk and Insurance—Issues and Practice 30(1): 114–120. O'Keeffe, P., Desai, A., Foroughi, K., Hibbet, G., Maxwell, A., Sharp, A., Taverner, N., Ward, M. and Willis, F. (2005) ‘Current developements in embedded value reporting’, British Actuarial Journal 11(3): 407–496. O’Brien, M.A., Brailsford, T. and Gaunt, C. (2010) ‘Interaction of size, book-to-market and momentum effects in Australia’, Accounting & Finance 50(1): 197–219.
Jochen Zimmermann et al Measuring Risk Premiums Using Financial Reports and Actuarial Disclosures
231 Paetzmann, K. and Lippl, C. (2013) ‘Accounting for European insurance M&A transactions: Fair value of insurance contracts and duplex IFRS/U.S. GAAP purchase accounting’, The Geneva Papers on Risk and Insurance—Issues and Practice 38(2): 332–353. Perez-Quiros, G. and Timmermann, A. (2000) ‘Firm size and cyclical variations in stock returns’, The Journal of Finance 55(3): 1229–1262. Petkova, R. (2006) ‘Do the Fama-French factors proxy for innovations in predictive variables?’ The Journal of Finance 61(2): 581–612. Reinganum, M.R. (1981) ‘Misspecification of capital asset pricing: Empirical anomalies based on earnings’ yields and market values’, Journal of Financial Economics 9(1): 19–46. Ross, S.A. (1976) ‘The arbitrage theory of capital asset pricing’, Journal of Economic Theory 13(3): 341–360. Rozeff, M.S. and Kinney, W.R. (1976) ‘Capital market seasonality: The case of stock returns’, Journal of Financial Economics 3(4): 379–402. Serafeim, G. (2008) Relevance vs. reliability: Evidence from embedded value accounting, working paper, Harvard University. Serafeim, G. (2011) ‘Consequences and institutional determinants of unregulated corporate financial statements: Evidence from embedded value reporting’, Journal of Accounting Research 49(2): 529–571. Simpson, M.W. and Ramchander, S. (2008) ‘An inquiry into the economic fundamentals of the Fama and French equity factors’, Journal of Empirical Finance 15(5): 801–815. Smith, C.W. and Watts, R.L. (1992) ‘The investment opportunity set and corporate financing, dividend, and compensation policies’, Journal of Financial Economics 32(3): 263–292. Tinic, S.M. and West, R.R. (1984) ‘Risk and return: January vs. the rest of the year’, Journal of Financial Economics 13(4): 561–574. Towers Perrin (2005) European embedded value—the story so far, working paper, Towers Perrin Tilinghast, London. Vassalou, M. (2003) ‘News related to future GDP growth as a risk factor in equity returns’, Journal of Financial Economics 68(1): 47–73. Vassalou, M. and Xing, Y. (2004) ‘Default risk in equity returns’, The Journal of Finance 59(2): 831–868. Vuong, Q. (1989) ‘Likelihood ratio tests for model selection and non-nested hypotheses’, Econometrica 57(2): 307–333. White, H. (1980) ‘A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity’, Econometrica 48(4): 817–838. Wu, R.C.-F. and Hsu, A.W.-H. (2011) ‘Value relevance of embedded value and IFRS 4 insurance contracts’, The Geneva Papers on Risk and Insurance—Issues and Practice 36(2): 283–303.
About the Authors Jochen Zimmermann is Professor of Accounting at the University of Bremen where he holds the Chair of Accounting and Control. His research covers insurance accounting, accounting regulation and the interplay of accounting-based regulations such as Solvency II with financial reporting. Stefan Veith received his doctoral degree from the University of Bremen. Currently, he works as senior policy officer at the Federal Financial Supervisory Authority of Germany. His research interests cover the economic effects of financial accounting enforcement as well as of financial and non-financial disclosures. Johannes Schymczyk holds a PhD in Economics from the University of Bremen. His PhD thesis from 2012 focuses on the role of economic measures in international financial reporting for life insurance companies. He currently works at PricewaterhouseCoopers.