Technical Findings on the Long-Term Guarantees Assessment Appendix 1: Methodology for the calibration of the adaptation (CCP) (as tested in the assessment)
14 June 2013 EIOPA/13/297
Blank page.
Contents 1.
Main technical conclusions ................................................................................ 4 1.1.
General framework ................................................................................... 4
1.2.
Harmonization .......................................................................................... 5
1.3.
Functioning of the adaptation ..................................................................... 5
1.4.
Practical implementation ............................................................................ 7
2.
Methodological findings .................................................................................... 8
3.
Methodology for deriving the representative portfolios. Data needs ..................... 14
4.
Composition of the representative portfolios ..................................................... 20
5.
Methodology for the derivation of the overall spreads ........................................ 21
6.
Methodology for the relevant portion of the spreads .......................................... 26
7.
Value of the input parameters referred to weights and spreads ........................... 30
8.
Values of the adaptations ............................................................................... 31
9.
Coordination between activation and calibration of the adaptation ....................... 34
10.
Composition of the representative portfolios .................................................. 36
10.1.
Central government bonds .................................................................... 36
10.2.
Central bank bonds .............................................................................. 36
10.3.
Other bonds ........................................................................................ 36
3
1. Main technical conclusions 1.1. General framework 1.
2.
3.
4 1
The framework used in LTGA exercise as basement regarding the adaptation would benefit from more clarity in key areas with a material impact either on the calibration or in the operationalization of the adaptation.
The most significant areas 1 where more clarity is suggested are the following ones: i.
Need to clarify (a) the definition of assets which should be considered in the representative portfolio; (b) assets to consider in the respective portfolios of bonds for the calibration of the currency-specific as well as of the nationalspecific adaptation, and (c) need to clarify which are their respective scopes of application (e.g. clarify the treatment of loans and structured bonds)
ii.
Need to solve the current circularity (the calibration of the adaptation is based on the spread of the relevant assets covering the technical provisions, whose value depends on the calibration of the adaptation, and therefore it is not possible to ascertain which assets cover the technical provisions without knowing firstly the adaptation)
iii.
Need to clarify the scope of application of the national adaptation, especially in the context of cross-border operations
iv.
Need to clarify the respective roles of EIOPA and national supervisory authorities regarding the activation/deactivation as well as the calibration of the national-specific adaptation. In particular it is unclear the procedure to apply by national supervisory authorities in order to send to EIOPA a request of activation, as it is unclear as well the interaction among EIOPA and the relevant national supervisory authority regarding the review of the activation/deactivation processes and the respective roles and processes for the calibration of the national-specific adaptation.
v.
Need to clarify the functioning where both currency and national adaptations may be activated (whether they can be applied at the same time, whether there is a prevalence in case of concurrence,…)
A major issue refers to the amount and scope of the currency-specific adaptation compared to the national-specific adaptation, because different understandings are possible. The framework used in LTGA exercise may lead to different representative portfolios and in the end to different amounts for the currency adaptation on the one hand and for the national adaptation in the same currency on the other hand. The following items describe more deeply each area
4.
From a purely technical perspective, EIOPA considers that it is feasible the implementation of both a currency and a national adaptation if the scope and calibration of both remains identical for all currencies other than the euro, and it is limited to the currencies of the EEA.
5.
On the contrary, EIOPA points out the practical complexity that would mean to put into practice a framework where the amount and scope of application of both adaptations differ for EU currencies other than the euro. This assessment refers to the calibration process and also to the activation process used to assess the existence of ‘temporary and exceptional stressed situations’. These difficulties will impact on undertakings, supervisors and EIOPA.
6.
Below practical difficulties are described in respect of adaptations for currencies out of EEA. Further than these technical difficulties, EIOPA suggests paying attention to the implications that might derive from declaring that markets out of Europe are in an exceptional stressed situation (e.g. reaction of market operators and macroeconomic consequences and responsibilities). This might be considered when deciding whether to restrict the currency-specific adaptation to European currencies only.
1.2. Harmonization 7.
EIOPA notes that the implementation processes for both the calibration and the activation of the adaptations are really demanding in terms of data sources and internal control. As a consequence a sufficient harmonization is a challenge that deserves appropriate analysis.
1.3. Functioning of the adaptation 8.
To the extent the adaptation mechanism (in either version) reflects the average profile of each insurance market; an individual undertaking’s asset exposures might be significantly different from the average market profile. The deviation from the risk profile of the individual insurers prevents that the adaptation incentives to invest in more risky assets. The amount and sign of the deviation will depend on the risk profile of the assets each undertaking really holds and the durations of its assets and liabilities
9.
From a macroeconomic point of view, to the extent that small undertakings contribute limitedly to the average profile of national markets (and hence to the level of the adaptation), the policy investments of the largest undertakings will have a major impact on the amount of the adaptation, and then on the own funds volatility of small undertakings. This may have as a consequence an alignment of policy investments.
10.
In order to avoid providing bad risk management incentives, EIOPA notes that spreads may materially reduce during the next twelve months after the initial application of the adaptation (this has been the real case of year 2012). This feature should be appropriately captured in the framework
5
11.
Closely linked to the features mentioned above, LTGA results show that the level of the adaptation may be volatile. For those undertakings with a profile sufficiently different from the market average, the volatility of the adaptation will be transmitted to their own funds, and hence to their solvency position.
12.
The national and currency adaptation are based on a representative portfolio concept. Therefore, in both adaptations numeric results are only understandable if it is considered the real composition of the assets of undertakings (type of bonds, credit quality, duration, currency, …). 1. On the one hand, when analysing the numerical results shown below for some Member States with high government spreads, there are some ‘apparently’ low values for the national-specific adaptation. This is due to the fly-to-quality investment policy applied by the insurance undertakings of such Members States. 2. On the other hand, when looking at the numerical results for some Member States with low government spreads, there are some values higher than expected. This is due to investment policies applied by the undertakings of such Member States looking for higher yields. 3. Finally, readers will observe that the adaptation for markets where undertakings mostly invest in bonds other than central government bonds (mainly corporate bonds), present a threefold feature in the period 20042012 analysed: a. Firstly, since the spreads of central government bonds of these markets usually have had a stable behaviour, the adaptation presents a lower level of variability compared to those other markets whose central government bonds have suffered volatile spreads, b. Secondly, in the case of markets mostly investing in corporate bonds is relevant to have in mind that these type of bonds use to trade with a spread compared to risk free rates, and hence the adaptation for all these markets have presented for all dates observed a certain level, although of limited amount. Therefore the adaptation for these markets is mostly explained by what might be labelled as the ‘natural’ spreads of corporate bonds.
6
c. But even among these markets the variability of the adaptation may be different. It is important to note that in some of the dates analysed, and for some credit quality steps, spreads of corporate bonds are materially dependent on the economic sector. In particular, banking and financial sectors have experienced periods of materially higher volatile spreads than other activities. Therefore even having two national markets with the same weight referred to the portfolio including corporate bonds, the respective adaptations may derive materially different values, pending on the concentration on one or other economic sector and the credit quality of the corporate bonds.
1.4. Practical implementation 13.
14.
EIOPA has considered that including in this appendix an exhaustive list of findings would be helpful to illuminate on the challenges for the implementation of the adaptations. Some points are easily solvable with the appropriate legal and technical toolkits, while a few of them are more challenging. Even in these cases, EIOPA is of the view that with an appropriate planning it is possible to cover all the areas involved.
In particular, deriving the four inputs (i.e. the two weights, ω _govt and ω _others , and the two spreads, S _govt and S _others ) which are necessary for the calculation of the adaptation has revealed quite challenging, especially regarding the national adaptation. Reasons for this finding refer to the following features a.
Nowadays, undertakings and supervisory authorities are not fully prepared to promptly provide on frequent basis the necessary disclosure of market values and spreads, including durations of bonds and loans. Should undertakings and national supervisory authorities set up in a short term appropriate implementation planning regarding the adaptation, EIOPA considers that it would be possible, without a material burden, to achieve the necessary preparedness in time for the entry into force of the full Solvency II framework. An appropriate follow up of the setting up and development of such planning is relevant. This appendix offers below a first approximation to the information infrastructure necessary to cover the gaps detected in the data available within the LTGA deadlines. Since most of the elements of this information are necessary for other purposes of Solvency II project (either belonging to Pillar I or to Pillar II areas), EIOPA considers that solving this lack of data will not cause a major additional impact for undertakings, supervisors or EIOPA.
b.
Regarding information from financial markets, the major limitations refer to the lack of yields curves for corporate bonds issued in EU currencies other than euro 2. Furthermore there are no yield curves for bonds of a specific national market (in fact, getting the information about geographical allocation of financial instruments is not such straight forward nowadays, as a consequence of worldwide activities). EIOPA cannot guarantee that this situation is solved before the application of Solvency II. Nevertheless there are some approximations based on the use of euro yield corporate curves with some adjustments. Although EIOPA calculations have provided apparently sensible outputs with such approximations, this field remains open to further analysis.
2
This partially applies to GBP area as well, to the extent the yields curves for corporates in such currency currently have an incomplete granularity
7
The same statement is applicable in respect of statistics of default and cost of downgrades used to calculate the portion of the spread that should not contribute to the adaptation. If the legal framework does not allow for approximations (subject to the appropriate controls and back testing) EIOPA and national supervisory authorities will straggle to get from the financial markets the necessary inputs to calibrate the adaptations. c.
15.
Furthermore, from a practical point of view and for the sake of harmonization, a material number of areas need either clarification or methodological development (see below).
Finally, the content of this report in respect of the calibration of the adaptations cannot be seen as a definitive outcome. Further analysis is relevant to better test the accuracy of the data gathered and the appropriateness of the suggested methodologies. In particular, to the extent that the adaptation cannot apply to the insurance business where the matching adjustment applies, and since the scope of the matching adjustment may vary to a great extent pending on Omnibus II final text, it is not possible at this stage to ascertain that the current figures will not materially change once data may be gathered excluding the assets covering the insurance business with matching adjustment.
2. Methodological findings 16.
This item tries to list the main difficulties found during LTGA, and the areas where clarification is necessary because there are divergent understandings with materially different impact.
17.
EIOPA considers that with an appropriate planning it is possible to cover all the areas described below
Definition of ‘representative portfolio’, either for the currency-specific or the national-specific adaptation)’. 18.
Once activated, the calculation and application of the adaptation requires a twofold perspective: a. Which assets should be considered in the calibration of the adaptation? b. Which liabilities may be discounted with the adaptation?
19.
8
The first item above (i.e. assets to be considered in the calibration of the adaptation) requires two steps: i.
Requirements for the assets to be included in the ‘representative portfolio’. For the purposes of this report EIOPA has considered the list set out in the
framework applied for LTGA exercise, as a closed list (bonds and loans, including mortgage loans, equities and properties) ii.
Requirements for the assets to be used to calculate the four parameters of the formula for the calibration of the adaptation (bonds and loans, including mortgage loans).
20.
Implementing the aforementioned schemes requires considering three dimensions: the currency of the asset, the currency of the liability the asset is covering, and the market where the insurance contract has been sold.
21.
If the ‘national adaptation’ is based on all the assets covering insurance obligations sold in a national financial market (regardless of the currency of both the assets or the insurance contract) its amount will systematically differ from the ‘currency adaptation’ (to the extent that this only refers to the assets expressed in the currency for which the adaptation is being calculated). This difference is already material for a number of national markets, according to the information available.
22.
From a practical point of view, to the extent that targets and scopes of both adaptations are different, in practice the implementation processes for activation and calibration of currency adaptation needs to be different than the implementation processes referred to the activation and calibration of the national adaptation. Therefore further than the coordination of activation and calibration within each adaptation, it will be necessary a third coordination among adaptations.
Treatment of loans 23.
On the treatment of loans, LTGA framework states that the representative portfolio of assets ‘… includes the following assets: bonds and loans, including mortgage loans; equity; property’. For the purpose of the calibration of ‘the adaptation’, the expression ‘bonds and loans, included mortgage loans’ seems to consider, ‘bonds’ as a different legal concept than ‘loans’ (this doubt would not exist if the text would have read ‘bonds, including loans and mortgage loans for this purpose’).
24.
Therefore, considered loans), or treatment,
25.
For the purpose of the LTGA and for the sake of an economic sound approach (principle of substance over form) it has been considered that those loans with features similar to fixed-income securities should be included in the calculations labeled as ‘bonds other than government bonds’ in the formula of the adaptation.
26.
Finally, an explicit clarification of the treatment of loans to policyholders might contribute to improve the legal certainty.
it might be uncertain whether the weights (and then the spread) in the formula of the adaptation refer exclusively to bonds (excluded those weights should include loans as well. Pending on their final the underlying figures for some markets may materially vary.
9
Main methodological issues regarding parameters referred to weights (ω _govt and ω _corp ). 27.
With the limitations mentioned below, the estimates of the two inputs referred to weights, (ω _govt and ω _others ), may be considered sufficiently reliable on general basis. Compared to the available data referred to previous dates, there is evidence that weights may materially change. It is not possible to assess whether these changes are due to exceptional drivers as those which motivate the adaptation or not. It will be necessary to test in the practice the impact on the four parameters used as inputs of changes in investment policies of undertakings. Furthermore, it will be necessary to consider that changes in the policies of undertakings to apply the matching adjustment, will influence the above referred weights as well.
28.
Although apparently the calibration only uses two general weights (for government and corporate bonds respectively), more detailed information on the composition of the government and corporate bonds categories will likely be necessary to derive appropriate spreads. In fact, the framework applied refers to spreads with the following expression: ‘…the relevant portion of an average spread on government bonds…’ (and the same for bonds other than government bonds) 3. Therefore, it is of utmost importance that undertakings develop and maintain appropriate data sets in order to achieve a harmonized implementation of all necessary components 4.
29.
Definition of ‘bond’ (e.g. to what extent structured products, whose spread does not derive exclusively from interest rates, may be included in such definition),
30.
As already mentioned, the current drafts envisage that for the national adaptation the composition of the representative portfolio has regard to all assets covering insurance contracts sold in a national financial market, (without explicit mention to either the currency of the asset or the currency of the liabilities). This raises a number of practical questions, assuming there are assets and liabilities expressed in different currencies: -
3
Which requirements should apply regarding the designation of assets covering insurance contracts sold in the national market (in order to avoid cherry picking and ascertain consistency with risk management)
An example may illuminate this point. Government bonds spread (S_govt) should reflect the spread of a portfolio representative of those government bonds actually held by insurance and reinsurance undertakings to cover the insurance contracts sold in each national market. While in some markets almost all government bonds held by insurance undertakings are issued by the domestic central government or central bank, in most of markets insurers invest also in bonds issued by international institutions and foreign central governments or central banks. Since those different issues may have quite different spreads, it is necessary to calculate some average spread. In doing this with an economic approach, the sub-weights of each sub-category of government bonds seem likely necessary.
4
10
This report only considers the data that undertakings need to manage in order to allow for an appropriate activation and calibration of the adaptation. At the light of those data needs, the report provides an opinion on whether it is materially burdensome or not to satisfy such needs. According to the IT state of art, it is assumed that any type of reporting based on an pure extraction of an already existing data base, will not require in itself a material cost.
-
whether the spread of each asset in a currency should refer only or not to the risk free curve for the relevant currency, having in mind that such spreads will be used to discount liabilities only expressed in the local currency
31.
For a given currency, the representative portfolio of assets for the currency adaptation has only regard to the average portfolio mix of insurers domiciled in the EU in that currency. This means that for a non-EU currency the representative portfolio for this currency is still only dependent on the investments of EU insurers in this currency. However, this may not be representative of the situation of the currency as a whole. Nevertheless this approach would ensure that the adaptation is based on the assets of the insurers who will apply it and is therefore effective.
32.
The technical flaws described advice to explore restricting the scope of the national adaptation to the assets covering liabilities expressed in the local currency with no currency mismatch. In case of a different approach, the previous steps reveal material flaws which are complex to solve.
33.
Identification of assets covering the own funds, since they should be excluded for the purposes of the calibration of ‘the adaptation’ according to the framework used in LTGA exercise,
34.
In the case of collective investments, application of the look through necessary to derive all the data needed for the purpose of the adaptation, has revealed also challenging for some markets where this type of investments is significant,
35.
Identification of bonds (and the relevant loans) covering technical provisions where ‘the adaptation’ may be applied is a challenging exercise in some markets,
36.
According to the framework applied in LTGA exercise, the calibration of ‘the national adaptation’ should be based on the spread of those assets that are representative of the investments made by insurance and reinsurance undertakings to cover the insurance and reinsurance obligations relating to insurance products sold in that national market. Where an insurance undertaking underwrites business both in its domestic market and in foreign markets, it is necessary to allocate assets distinguishing those related to contracts of the domestic market (and expressed in the local currency) and the rest of contracts. In most of markets this feature is currently immaterial, but there are already markets where insurance contracts written in foreign markets are really significant. In fact, in the future this situation might expand to other markets. In the case of the euro area this feature is most likely to appear and most likely to be challenging,
37.
Furthermore, since undertakings should identify the insurance contracts sold in each national market it will be necessary to set up appropriate criteria in order to identify where a contract is sold. Identifying this concept with any of the definitions of Article 13 of Directive 138/2009/EC, the identification of contracts sold in a national financial market may be challenging both legally speaking (at the light of the variety of possible cases) and from an operational point of view because undertakings and supervisors are currently not prepared to provide such data. However, these data could be made available in the future..
11
38.
In the case of insurance contracts provided under the freedom service regime or branches, the national supervisory authority of the host country cannot access to the data corresponding to such activities, since they remain in the authority of the home country. Such level of disclosure of the assets is not always available. However it could be required in the future
Main methodological issues regarding parameters referred to spreads (S _govt and S _corp ). 39.
Lack of data. While data on spreads of government bonds according to their credit quality are generally available (with limited exceptions), nevertheless spreads on corporate bonds are generally available only for the euro area and the sterling pound (in this case with a lower detail than the euro area).
40.
This means that for most of the other currencies either (a) there are no reliable spreads for different categories of corporate bonds; or (b) the spreads available are not publicly available (and then they do not meet the criteria set out in the framework applied in LTGA exercise); or (c) the spread is based on a financial reference which has an overall composition that may materially differ from the overall composition of the representative corporate bonds portfolios of the insurance undertakings of each market.
41.
12
There are at least two possible alternatives to move forward: i.
Firstly, using euro area corporate spreads as starting reference to derive the corporate spreads of other currencies,
ii.
Secondly using fixed-income indexes, where available
42.
Although both options have pros and cons, for the purpose of the calibration of the adaptation to be included in the LTGA report, the first option is selected. It is considered that this option (a) is easy to understand and replicate, (b) do not need material expert judgement, (c) may be easily applied to all currencies, and (d) relies on assumptions that might stand for most of foreign currencies in times of stressed national financial markets.
43.
EIOPA highlights that, although this option provides sensible results at this stage, it seems relevant to keep a door open to see whether there might be better practical solutions in a ‘transparent, objective and reliable manner’. Any alternative solution should be supported by a sufficient and appropriate back-testing and benefit of a practical implementation
44.
The specific case of the euro area. Although there are reliable spreads for ‘bonds other than government bonds’ (corporate bonds) in the euro area (appropriately
disclosed by credit quality step and economic sector), these spreads do not refer to each individual national financial market, but to the euro area as a whole. 45.
From other perspective, it is worth to mention that apparently a common set of mapping points is a desirable way to guarantee harmonization, although this solution should not be implemented in a rigid manner, but allowing for the treatment of specific cases, such as those markets where undertakings are mostly invested in a certain type of specific financial markets (e.g. covered bonds).
46.
Regarding availability of data, and due to the utmost consequences of ‘the adaptation’, it should be required evidence that the financial references used in the mapping exercise as model points for some types of bonds, sufficiently reflect the spreads of ‘the representative portfolio’ of the national market concerned for those types of bonds 5. This is a demanding exercise for those markets where the general set of mapping points is not sufficient. Furthermore this point is interlinked with the necessity of a sufficient granular detail in the information, although provided a sound balance with its burden is reached, as proposed above.
47.
Duration. Since spreads are materially dependent on the term considered 6 (i.e. the duration of the assets), it is relevant to confirm that the determination of the overall spread needs to allow for the duration of the assets considered in the calculation of the adaptation, specific to each national financial market and date. This allowance makes the calculation closer to the features of the representative portfolio, although it increases its volatility (adding an additional factor for variation) and the efforts required to produce the necessary inputs.
48.
This report has assumed that in order to provide a sufficiently sensitive calibration, it is relevant to allow for the duration of bonds portfolios as an input to the process of calibration of ‘the adaptation’. In this manner such process would reflect the different levels that spreads may present pending on the term of the instruments. Furthermore, to the extent that ‘the adaptation’ is intended to apply to all cash flows of the insurance undertakings of the national financial market concerned, its sensitiveness to the average duration may have material impact on the actual value of long term cash flows.
49.
Therefore insurance undertakings will need to have sufficiently granular information on the disclosure of government bonds by credit quality steps and their
5
This sentence only refers to the national-specific adaptation, since in the case of the currency-specific adaptation the focus is not on a particular country /market but on the currency
6
An example may illustrate this point. Let’s use an undertaking that uses higher credit quality steps to allocate long term investments (e.g. 10 years for AA bonds) and low credit quality steps for shorter exposures (e.g. 2 years for BBB bonds). `Using an average duration of 6 years for both AA and BBB bonds would likely produce a materially different spread than calculating the spread firstly for AA bonds and BBB bonds separately and then combining the two spreads with a financially sound approximation.
13
respective durations in order to calculate a sufficiently meaningful average (and similarly for other bonds). 50.
Unrated bonds. The treatment of unrated bonds is also an area of material importance in some markets. For the time being there is no methodological framework on how to allocate unrated bonds into the relevant credit quality steps, and therefore it is not possible to ascertain that all insurance undertakings apply consistent approaches. The method applied to carry out this exercise has its obvious influence in the spread.
51.
Structured products. Finally there is evidence that a material portion of bonds currently held by EU insurers embeds structures, in such a manner that the calculation of their spread and duration is not straight forward. Furthermore, a part of the spread of these bonds does not come from interest rates, but from premium of the embedded elements.
52.
In order to have a reference, for some markets the importance of these structured bonds is similar to the importance of one of the most popular instruments, i.e. bullet/plain vainilla bonds. For the time being, there is no precise information on the scope of these structures and their potential impact in the calibration of the adaptation. 53. Finally it is important to make explicit that credit risk adjustment for all currencies has been applied according to LTGA specifications. One of the lessons of LTGA exercise is the necessity of progressing in the methodologies of that adjustment, in order to improve its homogeneity among different currencies.
3. Methodology for deriving the representative portfolios. Data needs 54.
14
EIOPA has conducted a survey among national supervisory authorities in order to identify the current state of art related to the four parameters that should feed in the calibration of ‘the adaptation’. The template used for this purpose was the following one:
F IGURE 1
55.
National supervisory authorities were requested to fill in the template for the assets covering the insurance business expressed in their local currencies. For a few markets, where business expressed in foreign currencies (in most of cases the euro) are material, separate templates for assets covering technical provisions expressed in such foreign currencies were sent as well.
56.
Data referred to market values at year end 2011, in line with LTGA data. Furthermore a.
Only assets covering technical provisions were included in the survey. When assets covering technical provisions were not directly identifiable, the total amount of asset (as well as for the asset classes) was scaled down to the amount of the technical provisions.
b.
Assets covering unit-linked business were excluded from the survey. As mentioned in the executive summary, it has not been possible even an approximation to the assets that might be under the scope in the future the matching adjustment, since the range of different versions of this mechanism tested in LTGA has been very wide.
c.
Government bonds mean exposures to central governments and central banks, in line with the framework applied in LTGA exercise
d.
Covered bonds, as defined in Article 163 SR7 of the Informal draft Level 2 (October 2011), were allocated to the ‘Bonds other than government bonds’ category.
e.
For indirect exposures (e.g. in funds that have invested in the above mentioned categories - bonds or equities) a look through approach was applied, where possible.
57.
LTGA has considered as representative portfolio all assets included in the template with the only exception of the category ‘Other assets’.
58.
Data provided by national supervisory authorities covered in a vast majority of cases the whole national market.
Collection of information 59.
LTGA exercise has shown that currently it is not possible to provide in an efficient and timely manner the reliable information necessary for the calibration of the adaptations. Nevertheless EIOPA considers it will be possible an accurate calculation if undertakings store the information listed in the dataset below. Such requirement would not be materially burdensome for undertakings, to the extent that most of the information is already necessary for other purposes.
60.
The advantage of undertakings having this information appropriately stored is that it would allow to provide insurers with an automatic toolkit in order to extract the information relevant for the calculation of the adaptation (including appropriate summary reports)
15
61.
At the light of the difficulties experimented by both national supervisory authorities and undertakings to fill in the aforementioned survey, it is relevant to take special care about the necessities of information for an appropriate calculation of the four parameters used in the calibration (ω _govt , ω _corp , S _govt and S _corp ).
62.
At this moment the following three dimensions have been identified in order to make possible that national supervisory authorities and EIOPA gather the information necessary for EIOPA calibration of the adaptation: a. For government bonds, national supervisory authorities and EIOPA cannot derive a reliable estimation of the parameters unless they collect the following information: i.
As first dimension, disclosure of the Solvency II market value as in the nonstressed balance sheet, according to buckets with homogeneous spreads. In the case of government bonds, the credit quality of the issue may be a good proxy to define those buckets.
ii.
Additionally, as second dimension, it is necessary to calculate the duration of the government bond portfolio in each category, since the spread materially depends on such duration.
b. For ‘bonds other than government bonds’, national supervisory authorities and EIOPA cannot derive a reliable estimation of the parameters unless they collect the following information i.
ii.
16
Disclosure of the Solvency II market value as in the non-stressed balance sheet, according to buckets with homogeneous spread. The identification of these buckets is more complex since corporate bond spreads depend to a large extent of credit quality and economic sector. The disclosure by economic sectors is relevant since currently, bonds with the same credit quality have materially different spreads pending on the economic sector they mainly belong to. The disclosure by countries for the purpose of the national adaptation is really demanding and it is uncertain whether its cost eventually balances the higher accuracy in the calculation. As in the case of government bonds, it is necessary to calculate the duration of portfolios in each category.
63.
As additional third dimension, the two sets of information set out above should be differentiated by each national financial market where the insurance contract has been sold.
64.
EIOPA considers it is possible an accurate calculation if undertakings store the information listed in the following table. Grey cells identify those concepts that either undertakings already need to manage for purposes other than the adaptation, or the concepts may be automatically calculated (i.e. the concepts do not mean any additional burden for management of undertakings).
BONDS
LOANS
Identification (ISIN, CUSIP…)
Internal identification of the loan
Type of identification (ISIN, CUSIP, …) Type of bond 7
Type of loan
Currency of the issuer
Currency of the debtor
Credit quality of the asset
Credit quality of the debtor
Currency of the asset
Currency of the loan
Does the bond cover insurance obligations where the adaptation may apply (even if not activated yet) (yes/no)
Does the bond cover insurance obligations where the adaptation may apply (even if not activated yet) (yes/no)
Does the bond cover insurance obligations where the adaptation is applied (once it is activated) (yes/no)
Does the bond cover insurance obligations where the adaptation is applied (once it is activated) (yes/no)
Market where the insurance contract covered with the asset is sold
Market where the insurance contract covered with the asset is sold
Currency of the insurance contract
Currency of the insurance contract
Market value
Market value
Yield
Yield
Duration
Duration
65.
Please, note that the table does not mean necessarily for undertakings a reporting requirement to supervisors, but just a management requirement. This paper does not aim to discuss reporting requirements related to the adaptation. Note as well that only the last three items of information need regular update, while the rest of concepts are rather stable in time.
66.
The following table tries to summarize the data referred above with a tentative disclosure of ‘bonds other than government bonds’ into homogeneous categories of spread.
7
Firstly central governments/ Central Banks, International Institutions; and secondly Governments and Local Authorities, Financial, Industrial, Technologies, Utilities and others.
Regional
17
Market value in the Solvency II nonstressed balance sheet
Credit quality 0
1
2
3
4
5
6
Total
I. Adaptation of the domestic financial market. Assets covering insurance contracts sold in the domestic financial market expressed in the local currency I. Assets neither covering (a) own funds, (b) index/unit-linked contracts, (c) contracts with matching adjustment, (d) liabilities other than technical provisions, (e) assets covering own funds Government bonds Banking bonds Other bonds
financial
Industrial bonds Others bonds Loans, mortgages
incl.
Subtotal bonds Equities covering this type of business Properties covering this type of business Subtotal representative portfolio for the purposes of the adaptation Other assets covering representative portfolio
this
business
not
included
in
the
Total assets covering this type of business, sold in the domestic financial market
II. Insurance contracts susceptible of using the adaptation sold in the domestic market, expressed in foreign currencies (where material)
18
Assets covering this type of business, sold in domestic market, in
foreign currency X Assets covering this type of business, sold in domestic market, in foreign currency Y … Total insurance business (susceptible of use the adaptation) sold in domestic market in foreign currency
III. Insurance contracts susceptible of using the adaptation sold in markets other than the domestic market (all currencies). (only where material) (to be sent to the relevant supervisory authorities of the host financial markets) Assets covering this type of business, sold in market X and currency Y … Total assets covering this type of business, sold in markets other than domestic
IV. Information to reconcile with total assets of Solvency II non-stressed balance sheet Assets covering index and unit linked contracts Assets covering contracts where the matching adjustment applies Assets covering liabilities other than technical provisions Assets covering own funds Total asset Solvency II non-stressed balance sheet F IGURE 2
67.
On data necessary regarding the duration of assets, they may be derived from the database mentioned above applying automatic procedures, then with no additional cost for insurance undertakings.
68.
It seems highly desirable that all undertakings apply the maximum level of harmonization in respect of the granularity and the methods used to produce the
19
input needed for the calibration of ‘the adaptation’. At the light of the impact of ‘the adaptation’, this maximum harmonization seems necessary both from a macroeconomic perspective (e.g. for the purpose of financial stability) and from a supervisory perspective.
4. Composition of the representative portfolios 69.
None of the sources used for the calibration of the adaptations provides the complete range of data necessary for an accurate calculation. Therefore a number of workarounds have been necessary.
70.
EIOPA feels that jointly considered the data used and the workarounds implemented, the conclusions of this report are robust and would very likely remain under a fully accurate calculation. Please note that these workarounds should not be considered to be specified for the future.
Adaptation per currency 71.
Under the methodological framework described above Figure 3 shows the approximated composition considering all assets covering technical provisions for insurance business other than unit and index-linked contacts (thus, with the caveat of not excluding business with matching adjustment, due to its uncertain scope)
F IGURE 3: C OMPOSITION OF THE ASSETS COVERING TECHNICAL PROVISIONS WHERE THE ADAPTATION MAY APPLY
Source: specific survey national supervisory authorities (february 2013). Data 31-12-2011
20
72.
Grey part of bars identifies those assets covering technical provisions where the adaptation may apply, that are excluded from the ‘representative portfolio’ having in mind the definition of this portfolio 8.
73.
Regarding the assets not considered in the calibration of the adaptation, Figure 3 shows that countries may be grouped into three obvious categories: a. Markets where assets excluded mean around or less than 10 per cent of total assets. It is the case of the largest markets (France, Italy, Germany and UK) perhaps due to their higher retention of risks (recoverables from reinsurance and SPVs are excluded from the ‘representative portfolio’). Some medium sized markets belong to this category as well (e.g. Austria, Belgium, Czech Republic, Portugal and Sweden). b. On the opposite side, there are two markets (Bulgaria and Malta) where assets excluded mean the major part of assets covering technical provisions, c. For the rest of countries, assets excluded from the ‘representative portfolio’ mean about 25 per cent of total assets.
National Adaptation 74.
Although there have not been specific data for a fully accurate assessment of the composition of the representative portfolio for the purposes of the national adaptation, having in mind available statistics on insurance contracts sold in foreign markets and assets expressed in other currencies, it is considered at this stage that in general terms the composition and weights applied for the adaptation per currency may be used for the calibration of the national adaptations without producing material distortions.
5. Methodology for the derivation of the overall spreads 75.
Regarding the calculation of overall spreads (S _govt and S _corp ), the following chart provides an overall description of the major steps needed for such purpose.
76.
The two main challenges refer to the so-called ‘mapping’ exercise and the calculation of ‘the portion of the spread’ that actually constitutes ‘the adaptation’. The current item refers to ‘mapping’ while the following one expands on the calculation of the ‘portion’.
8
Please, note the point made above, regarding the treatment of loans, including mortgages, mortgages and units (investments funds and UCITS).
21
F IGURE 4
22
77.
The mapping exercise is challenging both in terms of granularity of information needed, complex methodology and availability of data.
78.
On granularity and as mentioned above when dealing with collection of information, the mapping exercise should theoretically lead to as many financial references as materially different categories/behaviours of spread may be observed. For other bonds further than the disclosure by credit quality steps, the additional dimension referred to economic sector seems theoretically relevant, although it raises practical problems.
79.
Nevertheless the theoretically desirable granularity finds material limitations due to a twofold feature: lack of data (there are no reliable data) and availability of data (data exist, but their access is restricted or they proceed from a financial institution with some type of indirect interests on the data). Both have been developed in the item devoted to methodological findings.
80.
From a legal view an essential point refers to the sources of information that may be used to derive the spreads. The common practice is based on ‘financial curves’ produced by concrete specialized providers. Others alternatives do seem neither practicable, nor cost-efficient, nor appropriate in view of the objectives of the calculation.
81.
Assuming it is possible to use ‘financial references’ provided by generally accepted sources of financial information, (under the relevant requirements), no major issues have been noted in accessing to spreads of government bonds according to their credit quality.
82.
Nevertheless as mentioned in the executive summary, spreads on corporate bonds are generally available only for the euro area and the sterling pound (in this case with a lower detail than the euro area)
83.
The selected option to solve this lack of data is based on a straightforward use of the financial information available with a high degree of reliability. The following
expression summarizes the proposed approach for the calculation of the spread of a corporate bond for a given currency other than the euro: S _corporates_currency_X = f ( S _corporates_euro , Y _risk_free_rate_currency_X , Y _rsik_free_rate_euro factor_correction 9 ) 84.
,
A possible approximation might have in mind that rating providers use to link the credit quality of corporate bonds to the credit quality of the respective national government bonds. The calculation might look like something similar to where the credit quality of the bond is equal or better than the national government credit quality S _corp_currency_X = S _corp_same_credit_quality_government_X [ * factor_correction ] otherwise
85.
S _corp_currency_X = S _corp_same_creditquality_euro
Another alternative might consider that markets spread in a currency keep some similarity to the respective comparison of the respective risk free rate curves in such currency and the euro S _corp_currency_X = S _corp_same_creditquality_euro + k * ( Y _risk_free_rate_currency_X Y _rsik_free_rate_euro )
86.
These approaches implicitly rely on some assumptions that may not be applicable through the whole economic cycle. Nevertheless, considering a currency other than the euro in an eventual situation where the national financial market of such currency faces a stressed situation, the assumptions underlying the method above suggested may be sufficiently realistic. Even not reaching the ideal confidence level, the fact of having a method based on reliable and observable data from financial markets seems an element to balance some limited inaccuracy of the assumptions.
87.
The formula used for LTGA is the following one. This formula should be considered as a practical expedient to deliver usable results with the deadlines available. It does not mean any proposal for the future, and hence this topic will need further analysis: S _corp_currency_X = S _corp_same_creditquality_euro + 0.5 * ( Y _risk_free_rate_currency_X Y _rsik_free_rate_euro )
9
This correction factor might be included to allow for some features not captured in the other inputs. Obviously the values of correction factor should be mostly based on expert judgement for the time being
23
Calculation process for the overall spreads
24
88.
During February 2013 EIOPA carried out a specific survey to support the quantification of the adaptation. This survey delivered reliable values to determine the weights to be applied to the governments and corporate bond spreads in the final formula for the quantification of the adaptation as well as the relative contribution from the different markets in the euro zone for the euro adaptation. Nevertheless for the purpose of the determination of the overall spreads on bonds, additional information is needed.
89.
For this purpose, specific information contained in LTGA inputs has been used in order to calculate sufficiently reliable overall spreads. Although LTGA inputs are quite granular, there are still assumptions to take to cover specific areas of the calculation of the overall spreads.
90.
Data from LTGA spread sheet (sheet ‘BS+) are used to build the two following standardized portfolios: I.
A portfolio of central government bonds and international institutions, composed of 32 types of bonds according to the issuer 10 (one type per European country plus one EEA international institutions and the remaining for non-EEA international institutions)
II.
A portfolio with 52 types of bonds. It is composed firstly of 32 bonds similar to the first portfolio, but whose issuers are non-central authorities. The 20 remaining bonds reflects corporate bonds belonging to financial, industrial, utilities and other corporate bonds category (with five credit quality steps for each economic sector, from 0 to 4 or higher)
91.
For each participant these two portfolios are built, including information on market values for each bond and durations.
92.
In the calculation of the adaptation per currency, only bonds in local currency are considered. For other bonds, their market value is set to zero.
93.
In the calculation of the national adaptation, all bonds are taken into account. This implies the assumption that all of them cover insurance obligations of contracts sold in the national financial market. In general, this is not a strong assumption and the calibration of the adaptation obtained is sufficiently reliable and accurate.
94.
In the following step these two portfolios are aggregated at country level. Market values of bonds are simply summed up, while for durations a weighted average is
10
LTGA data do not contain information on the currency of the bonds. In absence of any other reliable assumption, it is assumed that the currency of a government bond is the local currency of the issuer. Bonds from international institutions are expressed in the local currency of the undertaking (this assumption does not have a material impact, since these type of bonds do not have any contribution to the adaptation, to the extent that they are considered to be risk free). The only impact of this assumption refers to the weights, although having in mind the low proportion that these bonds mean compared to the total of bonds, such impact is negligible.
derived (having in mind the formula of duration, the weights are the market values). 95.
Once mapped each national financial markets in the two aforementioned portfolios, four magnitudes are calculated
96.
Firstly, the process derives for each type of bond the relevant risk free rate according to its currency and duration 11. For consistency with other areas of LTGA exercise, swap curves (appropriately adjusted for credit and basis risk) are considered as risk free rates. Where there has been evidence of the lack of representativeness of swap curves, the relevant government curve has been qualified as risk free. (in LTGA exercise this criterion has been applied for the currencies of Bulgaria and Poland). In these cases, obviously government bonds do not contribute to the adaptation.
97.
Secondly the relevant market yield is calculated in similar manner. In this case further than the duration, the yield curve is selected according to the either issuer (for government issuers) or the economic sector and credit quality (for corporate bonds). Where a certain type of bond lacks of yield curve, the most similar type of bond is used (its yield curve is applied).
98.
Finally the part of the spread that cannot contribute to the adaptation is quantified (the following sub-item describes this calculation).
99.
The spread net of the ‘portion’ not allowed for the adaptation (with a floor nil) is added to the risk free rate, forming the rates for the adaptation.
100. This process is carried out for each type of bond of each of the two portfolios initially described. To enable the determination of an overall adaptation aggregating the individual results that are differentiated by each bond and duration, the following calculation is performed. 101. Two calculations are carried out for each of the two portfolios (central government bonds separately from ‘other’ bonds), assuming both of them are zero-coupon portfolios -
Vector of zero coupon values to maturity, obtained accruing market values declared in LTGA for each type of bond, using as capitalization rate the risk free rate of each type of bond. Current value = market value declared in LTGA inputs.
-
Same vector but using as capitalization rate the following one: total yield bond – part of the spread that cannot contribute to the adaptation. Current value = market value declared in LTGA inputs.
102. Let’s imagine a representative portfolio with only two types of central government bonds. The two following calculations are applied (and similarly for ‘other bonds’ portfolio) 11
For this purpose a linear interpolation is applied. Using other methods does not have a material impact on the final calibration of the adaptation.
25
Type 1
Type 2
Total
1,000
800
1,000 + 800
Duration
3 years
7 years
3 / 7 years respectively
Risk free rate
3.20%
3.50%
Yield
4.20%
3.80%
40 bp
10 bp
Vector RFR cash flows
1.000 * ( 1+0.032 )^3 = 1,099
800 * ( 1+0.035)^7 = 1,017
Vector CCP cash flows
1.000 * ( 1+0.038 )^3 = 1,118
Market value
Portion considered
not
800 * ( 1+0.037)^7 = 1,032
2,116
2,150
T ABLE 1
Using the right column refers to ‘Total’, vector RFR delivers an internal effective rate 3.396 12, while Vector CCP delivers and internal effective rate 3.735 103. The difference among both internal effective rates is taken as the value for the adaptation corresponding to the portfolio. 104. This calculation is applied to the portfolio of “central government bonds” and separately to the portfolio of “other bonds”, and feed in the formula of the calibration. 105. EIOPA contribution has assessed the adaptation for all end years in the range 2004-2012. Since LTGA technical specifications provide quantitative inputs only for years 2004, 2009 and 2011, where practicable interpolated estimates have been applied for the other dates. 106. Since LTGA data on disclosure of bonds and market values only refer to year end 2011, for the sake of comparability it has been assumed that the composition of the portfolios of bonds remained unchanged during the period considered.
6. Methodology for the relevant portion of the spreads 26 12
{ 1099 * 1.03396-3 } + { 1017 * 1.03396-7 } = 1800
107. This item describes the methodology applied to calculate the part of market spreads that according to the framework applied in LTGA exercise cannot contribute to the adaptation. The rest of the spread is considered to calculate the adaptations following the methodology previously described. Conceptual underlying the calculation for corporate bonds 108. The framework applied in LTGA exercise envisages that the adaptation shall be determined based on the following formula 𝐴 = 𝑤𝑔𝑔𝑔 ∙ 𝑚𝑚𝑚�𝑆𝑔𝑔𝑔 ; 0� + 𝑤𝑐𝑐𝑐𝑐 ∙ 𝑚𝑚𝑚�𝑆𝑐𝑐𝑐𝑐 ; 0�
109. Parameters S are intended to capture the relevant portion of the spread of a bond (understood as the spread between the interest rate that could be earned from the bond and the rates of the basic risk-free interest rate term structure), where “the portion shall not be attributable to a realistic assessment of expected losses or unexpected credit risk on the assets. The portion shall not be attributable to any other risk.“ 110. The calculation of ‘the portion’ for corporate bonds is based on the decomposition of spreads into the following parts: 13 1. Expected losses: This component reflects an allowance for expected, average default losses. It is sometimes also referred to as default risk premium. It varies over time as the bonds’ cash flow viability changes. 2. Unexpected losses: This component reflects the reward required by the holder of a credit-risky bond for carrying the implied market risk of that bond. This component is also sometimes referred to as allowance for unexpected defaults or referred to as credit risk or spread risk premium. It varies over time as market risk aversion changes. 3. Liquidity premium: The liquidity premium reflects the reward required by the holder of the bond to compensate for the illiquidity costs implied by the illiquid bond. Thus, it also reflects the illiquidity of the market as such. Conceptual underlying the calculation for corporate bonds 111. For the purpose of the determination of the adaptation according to the legal texts as stated above, the adaptation is defined to be the share of the spread that cannot be attributed to a realistic assessment of expected and unexpected losses. The adaptation does therefore capture the liquidity premium component of the spread as well as that part of the unexpected losses due to changing market risk aversion. 112. For this purpose three elements have been practiced:
27 13
For the purpose of this exercise other components such as expense or taxation effects are neglected.
a) The reflection of probability of default expressed in annual basic points that is interpreted as investors’ requirement for taking the risk of the expected probability of default of a bond. The expectation of a default is based on historical default probabilities taken from default statistics combined with an assumption on the loss given default, which is assumed to be 30% for corporates. For this purpose the optimum solution should be based on statistics of defaults on the one hand as short as possible to provide some reflection of the cycle, and on the other hand sufficiently long to gather a sufficient number of observations necessary to guarantee reliable outputs (i.e. a period of 10-15 years). This element has been derived separately for all rating categories and durations, however, no further distinction is made e.g. by economic sector or region. The necessity for further differentiation will need to be analysed in the future (e.g. regarding the differentiation among economic sectors, on the one hand such disclosure may be justified by sufficiently different statistics of default and downgrades, but on the other hand such differentiation may bias insurers policy investments and then produce undesirable macroeconomic consequences). Although some statistics with these properties have been reached, in absence of sufficient information about the process for their construction, due to time constrains, a single statistic referred to the last 30 years has been used. b) The reflection of the cost of downgrades, expressed as well in annual basic points, is based on the probability of being downgraded as well as the cost in case of downgrade. The probability of being downgraded is taken from transition matrices derived and published by rating agencies. The cost in case of downgrade is based on the observation of market spreads for different durations and rating categories on financial markets. This calculation considers the cost of downgrades for the whole duration of the bond. Similar considerations regarding the data inputs used (statistics of downgrades) apply to this calculation as well. c) A third element reflective of the volatility of the probability of default (the spread an investor requires as a safety margin in addition to the expected probability of default) is linked to the standard deviation of the historical probability of default rates. Based on long term statistics, an overall estimation of 77 per cent of the first element is applied. Conceptual underlying the calculation of sovereign bonds 113. Theoretically the conceptual framework described for corporates could also apply to sovereign bonds as well, provided appropriate statistics are available.
28
114. Nevertheless due to the limited number of exposures and their specific features in the European market, there are no reliable statistics for defaults which are needed for the purpose of the estimation of the first and third element as outlined above and this lack of sufficient number of homogeneous observations seems likely to remain in the future.
115. As a practical expedient, the same figures used for government bonds in the matching adjustment have been applied to the calculation of the part of the spread that cannot contribute to the adaptation. 116. An alternative would have been to base the estimation of the spread component relating to the probability of default on the probabilities of default for corporates of the same rating category and duration combined with a separate estimation of the loss given default for sovereigns. This approximation is based on the assumption that the rating assigned by the Credit Rating Agencies is comparable for both, corporates and sovereigns (e.g. a AAA corporate bond is considered having the same probability of default than a AAA government bond). 117. In the case of the estimation of the cost of downgrading, a separate calculation could have been performed for sovereigns, as actual and reliable statistics for sovereigns are available. 118. Both for corporate and government bonds, the calculation of the portion is open to refinements, in particular regarding (a) the volatility of the probability of default ; (b) default statistics and statistics of downgrading; (c) granularity with respect to regions or economic sectors (see above the consideration on the pros and cons of these differentiations). Treatment of bonds with very low rates 119. Regarding bonds with very low credit quality (below credit quality step 4), it is deemed relevant to retain them in the process of calculation of weights and overall spreads feed in the formula of the calibration. Nevertheless, there is sufficient and repeated evidence that spreads on those bonds mostly reflect the high expectance of default, and that such high expectance is not fully captured in the methodology described above for the calculation of the part of the spread that cannot contribute to the adaptation. 120. Therefore this area needs further development in order to avoid bad risk management incentives. 121. For the purpose of LTGA exercise, no adjustment has been applied to these bonds in absence of a sufficiently agreed methodology. This omission has a material impact only in respect of the adaptation of two Member States (Greece and Cyprus), and therefore their figures below should be read with this material caveat.
29
7. Value of the input parameters referred to weights and spreads Weights for the adaptation per currency 122. Having in mind all aforementioned, the following table provides a first outlook of the values of parameters referred to weights at the end of 2011. Each parameter has been calculated at national level and taking into account as ‘reference portfolio’ the sum of exclusively bonds, loans, equities and properties in accordance with the relevant framework. Govts bonds
Other bonds
Govts bonds
Other bonds
Total bonds
Austria
16.8%
52.3%
69.0%
Italy
65.8%
26.2%
92.0%
Belgium
55.7%
35.7%
91.4%
Bulgaria
52.2%
10.1%
62.3%
Latvia
76.9%
7.7%
84.6%
Lithuania
87.5%
0.3%
87.9%
Croatia
77.8%
6.8%
84.5%
Luxemburg
60.0%
23.0%
83.0%
Czech R.
64.5%
28.2%
92.7%
Malta
67.1%
20.9%
88.1%
61.3%
32.3%
93.6%
9.8%
62.6%
72.4%
Total bonds
Denmark
19.8%
56.0%
75.8%
Netherlands
Estonia
44.8%
48.0%
92.9%
Norway
Finland
11.7%
55.4%
67.1%
Poland
94.6%
3.4%
98.0%
France
33.7%
46.6%
80.3%
Portugal
36.5%
52.5%
89.0%
Germany
12.1%
76.1%
88.2%
Slovakia
51.1%
43.4%
94.5%
Greece
62.8%
21.0%
83.8%
Slovenia
44.0%
48.1%
92.1%
Hungary
86.3%
10.1%
96.4%
Spain
38.6%
49.9%
88.5%
Iceland
77.2%
9.3%
86.4%
Sweden
17.9%
48.0%
65.9%
Ireland
77.8%
15.4%
93.2%
UK
30.5%
51.0%
81.6%
T ABLE 2: V ALUE OF PARAMETERS Ω _ GOVT , Ω _ OT HERS S URVEY REFERRED TO 31-12-2011
123. Weights for the euro are 32.3 and 52.4 per cent for parameters ω _govt , ω _others respectively.
30 F IGURE 5: W EIGHTS USED FOR THE CALIBRATION OF THE NATIONAL ADAPTATION ( W _ GOV AND W _ CORP ). 3112-2011
124. Since ‘the adaptation’ and the ‘matching adjustment’ are incompatible, and the final scope of the ‘matching adjustment’ is still uncertain, Figure 5 should be considered with the caveat mentioned in the executive summary. Weights for the national adaptation 125. As mentioned above, in absence of specific data for a fully accurate assessment of the composition of the representative portfolio, statistics on insurance contracts sold in foreign markets and assets expressed in other currencies, suggest that in general terms the composition and weights applied for the adaptation per currency may be used for the calibration of the national adaptations without producing material distortions. Duration of the portfolios of bonds 126. The following table derives from LTGA data and reflects the market durations applied to the calculation of the national adaptation
T ABLE 3
8. Values of the adaptations Currency adaptation
31
T ABLE 4
National adaptation
T ABLE 5
Understanding the calibration of the adaptations 127. The level of the adaptation has three major drivers: a. Firstly, the real composition of the portfolios of bonds that the insurers of each national market hold. In this respect
32
-
It is necessary to have in mind that not all bonds are included in the calculation, but only those allowed in the formula of the adaptation. Below this report shows the composition of assets for each national market according to LTGA data,
-
Credit quality is one of the key factors influencing the level of the adaptation
b. Secondly the duration of each type of bond, to the extent that spreads are materially different pending on the duration of the assets
c. Finally the respective weights of the two high-level portfolios used in the calculation. 128. Therefore, an analysis mostly based on the name of the Member State will allow neither to understand the figures above nor to capture how actually the adaptation works. 129. Some guidance that readers may use in their analysis might be: 1. On the one hand, when analyzing the numerical results shown below for some Member States with high government spreads, there are some ‘apparently’ low values for the national-specific adaptation. This is due to the fly-to-quality investment policy applied by the insurance undertakings of such Members States. 2. On the other hand, when looking at the numerical results for some Member States with low government spreads, there are some values higher than expected. This is due to investment policies applied by the undertakings of such Member States looking for higher yields. 3. Finally, readers will observe that the adaptation for markets where undertakings mostly invest in bonds other than central government bonds (mainly corporate bonds), present a threefold feature in the period 20042012 analysed: a. Firstly, since the spreads of central government bonds of these markets usually have had a stable behaviour, the adaptation presents a lower level of variability compared to those other markets whose central government bonds have suffered volatile spreads, b. Secondly, in the case of markets mostly investing in corporate bonds is relevant to have in mind that these type of bonds use to trade with a spread compared to risk free rates, and hence the adaptation for all these markets have presented for all dates observed a certain level, although of limited amount. Therefore the adaptation for these markets is mostly explained by what might be labelled as the ‘natural’ spreads of corporate bonds. c. But even among these markets the variability of the adaptation may be different. It is important to note that in some of the dates analysed, and for some credit quality steps, spreads of corporate bonds are materially dependent on the economic sector. In particular, banking and financial sectors have experienced periods of materially higher volatile spreads than other activities. Therefore even having two national markets with the same weight referred to the portfolio including corporate bonds, the respective adaptations may derive materially different values, pending on the concentration on one or other economic sector and the credit quality of the corporate bonds.
33
9. Coordination adaptation
between
activation
and
calibration
of
the
130. Generally speaking, there are two main approaches to define activation indicators of ‘the adaptation’ a. The first one might be ‘named’ as ‘macroeconomic approach’ and would use a set of indicators referred to financial markets in general (e.g. spreads on government bonds, situation of equity markets,…). Eventually other indicators might be added such as inflation rates, economic growth or budget stability. Having in mind drafts of legal texts used for LTGA exercise, this option has been disregarded. b. The second approach would identify activation indicators according to the features of the ‘representative portfolios’ as defined in the relevant framework. There is a separate item containing a detailed description of activation indicators according to this approach. The following considerations are limited to the coordination of activation processes with the calibration of the adaptation (i.e. prevention of the paradoxical situation where either ‘the adaptation per currency’ or ‘the national adaptation’ are activated but its calibration delivers a negligible amount of the adaptation -or inversely in the case of deactivation). Activation indicators based on the representative portfolio 131. Independently of the concrete definition of this approach and from an operational point of view, mapping the representative portfolio of each currency/ national financial market is a challenging exercise due to the wide range of data needed and the lack of reliable data to derive the spreads of corporate bonds for a material number of currencies / national financial markets. Nevertheless this challenge should be solved, in any case, for the calibration as well. 132. Furthermore, the experience shows that the composition of the representative portfolio may materially change in a rather short period. Hence activation indicators would need to evolve similarly. This may make challenging the definition of informative thresholds. 133. From a practical point of view, the more closely activation indicators reflect the representative portfolios, the more challenging the implementation of the activation becomes, to the extent that changes in the representative portfolio (e.g. duration) will likely lead to changes in the activation indicators.
34
134. Additionally, referring activation indicators to the ‘representative portfolio’ does not guarantee per se a sufficient coordination with the calibration of ‘the adaptation’ for those markets where assets other than bonds mean a material part of the representative portfolio.
135. Figure 6 below shows the weights that would be used in the calibration for both government bonds and for other bonds. The gap between the aggregated weights in each national financial market and 1, shows the part of the representative portfolio that would be used in the activation assessment of ‘the adaptation’, but cannot be considered in the calibration. (gap up to 1 = part of the representative portfolio not considered in the calibration) EIOPA survey referred to 31-12-2011
F IGURE 6: W EIGHTS USED FOR THE CALIBRATION OF THE NATIONAL ADAPTATION ( W_ GOV AND W_ CORP ). 31-12-2011 (gap up to 1 = part of the representative portfolio not considered in the calibration)
136. On the other side, there are several markets where their representative portfolios are almost exclusively composed by bonds. In these cases the activation indicator might end is so similar to the calibration mechanism, that de facto activation would become just a simple threshold of the spread.
35
10. Composition of the representative portfolios
36
10.1.
Central government bonds
10.2.
Central bank bonds
10.3.
Other bonds
37 T ABLE 6
38 T ABLE 7
T ABLE 8
39
