Technical Findings on the Long-Term Guarantees Assessment APPENDIX 2 Potential design for the activation and deactivation of the adaptation (CCP) 14 June 2013 EIOPA/13/298
Technical Findings on the Long-Term Guarantees Assessment APPENDIX 2 Potential design for the activation and deactivation of the proposed adaptation (countercyclical premium)
1
Disclaimer: the current appendix to EIOPA LTGA report aims to offer a complete picture on how the adaptation of the risk free rate measure could work in its current design contained in the framework of the LTGA. In particular, the appendix presents the EIOPA’s current thinking on how to complete LTGA exercise which assumes that the adaptation of the risk free rate is activated in different scenarios. The appendix also picks up on how the current deactivation mechanism could be implemented in practice. Based on the given framework for the LTGA, the current appendix assumes that as for the activation, the indicators would only inform the relevant decision making body and form the basis for the discretionary activation decision. Appendix 1 of this report, which deals with the methodology used for the calibration of the adaptation, already covers the main conclusions on this measure (including activation and deactivation), regarding the general framework, harmonization, functioning of the adaptation and its practical implementation, which are not repeated in this appendix. Additionally Appendix 1 introduces considerations on the coordination of the activation and the calibration of the adaptation. The current appendix suggests a set of indicators which are seen as relevant for informing the process of the activation and deactivation of the adaptation. The assessment is mainly based on a time series while taking into account limited historic data availability in various countries of the EU. It is acknowledged that such a historic assessment of the activation and deactivation of the adaptation is limited as historic data is not available in a sufficiently granular manner. With the implementation of Solvency II more data will be available, particularly regarding the investment portfolios held by European insurance undertakings. The activation and deactivation thresholds presented in this appendix are for illustrative purposes; none of these should be seen as an explicit proposal for the actual calibration of the activation or deactivation of the adaptation at this stage – it is even recommended not to mechanistically rely on any specific threshold. The current appendix does not discuss the governance process for the activation/deactivation since it is purely focused on the definition of the sharp, steep and unforeseen fall in financial markets as included in the relevant framework.
2
Contents Main technical findings ........................................................................................... 5 1.
General principles for the activation of the adaptation.......................................... 6 1.1.
Context ................................................................................................... 6
1.2.
Definition of sharp, steep and unforeseen fall in financial markets ................... 7
1.3. Criteria for assessing the suitability of indicators feeding into the activation and deactivation decision .......................................................................................... 7 2.
Set of indicators and data availability ................................................................ 8
3.
Maximum Drawdown Model .............................................................................. 9 3.1.
Calculation of maximum drawdowns in the equity market .............................. 9
3.2.
Drawdown distributions in the equity market .............................................. 12
3.3.
Indicator analysis in the equity market ...................................................... 16
3.4.
Crisis periods in the equity market ............................................................ 19
3.5.
Indicators for other asset classes .............................................................. 23
3.5.1 Government bonds................................................................................ 24 3.5.2 Corporate bonds ................................................................................... 26 4.
5.
Combination of individual indicators ................................................................ 27 4.1.
Portfolio approach ................................................................................... 27
4.2.
“X-out-of-Y” approach ............................................................................. 32
Additional indicators ...................................................................................... 34 5.1.
Supportive indicators ............................................................................... 34
5.1.1. Volatility of foreign exchange rates ........................................................ 34 5.1.2. Volatility of short-term interbank rates ................................................... 36 5.2.
Complementary indicators ........................................................................ 38
5.3.
Real-economy indicators .......................................................................... 38
6. Interplay between the activation/deactivation of the adaptation and its size calibration .......................................................................................................... 39
7.
6.1.
General considerations ............................................................................ 39
6.2.
Deactivation of the CCP ........................................................................... 39
Conclusion ................................................................................................... 42
ANNEX 1: COMPOSITE SYSTEMIC STRESS INDICATORS .......................................... 43 ANNEX 2: DIFFERENCES BETWEEN PRICE AND RETURN INDICES ............................. 46 ANNEX 3: SENSITIVITY OF RESULTS TO CHANGES IN THE ASSUMED AVERAGE DURATION OF BOND HOLDINGS ........................................................................... 48 ANNEX 4: CORRELATION ANALYSIS ...................................................................... 50 ANNEX 5: ACTIVATION ANALYSIS FOR DIFFERENT ACTIVATING/DEACTIVATING QUANTILES OF SIMULATED MAXIMUM DRAWDOWN DISTRIBUTION (MSCI COUNTRY
3
INDICES, PRICE INDEX), WHEN THE ADAPTATION IS ASSUMED TO BE ACTIVE FOR AT LEAST A YEAR, I.E. 261 TR ................................................................................... 51 ANNEX 6: ACTIVATION ANALYSIS FOR DIFFERENT ACTIVATING/DEACTIVATING QUANTILES OF SIMULATED MAXIMUM DRAWDOWN DISTRIBUTION (PORTFOLIOS), WHEN THE ADAPTATION IS ASSUMED TO BE ACTIVATED FOR AT LEAST A YEAR, I.E. 261 TRADING DAYS............................................................................................. 55 ANNEX 7. MAXIMUM DRAWDOWNS OF PORTFOLIOS AND INDIVIDUAL ASSET CLASSES CALCULATED OVER A PERIOD OF 1 YEAR FOR VARIOUS COUNTRIES ........................ 59
4
Main technical findings The current appendix to EIOPA LTGA report outlines the design of a set of financial market indicators which could be used to support a decision by EIOPA for activating and deactivating an adaptation to the risk-free rate. For illustrational purposes, the appendix also presents an evaluation of the functioning of these indicators using historic data from the financial markets. Such a historic analysis can only provide a first understanding of the underlying mechanics and how thresholds could be set that, where exceeded, would indicate the need to activate an adaptation. Future implementation under Solvency II can be based on a broader dataset than the calculations applied for this exercise. In particular, more granular data will be available regarding the investments of insurance undertakings (“representative portfolio”) which is not easily available for past periods. The analysis suggests that the market indicators are based on the following four types of time series: primary indices, supportive indices, complementary indices and real economy indices. The main emphasis is placed on the following primary indices: government bond spreads, corporate bond spreads, equity indices. While the development of these time series would be directly linked to the market value of assets held by insurance companies, some other time series could be used as supportive information to cross-check the analysis. Such supportive indices can e.g. be the observed volatility of foreign-exchange rates (against EUR and/or USD), or the observed volatility of short-term interest rates. •
Composite indicators of systemic risk like the CISS indicator of the ECB or other indicators can complement the presentation of results.
Finally, it is relevant to compare the findings resulting from the analysis of financial market developments against time series from the real economy which could signal a crisis situation. Such indicators can e.g. be inflation rates (consumer prices and maybe producer and/or import prices), industrial production, or confidence indicators (e.g. purchase managers or consumers). •
Over time it would be necessary to re-assess the explanatory power of the set of indices used for the analysis: First, insurance undertakings’ portfolios might change as, for instance, additional asset classes become more important. Second, future financial crises might reveal new patterns of shock transmission, so that the addition of new primary and supportive indices would complete the analysis.
•
In case certain indices for a specific country or a specific currency are not available, a suitable proxy should be used – throughout this historic analysis, either Euro area or EU indices have been used as proxies. It is understood that an important basis of any decision to activate the adaptation would be an assessment on whether a proxy which has been used for the analysis is indeed suitable in the specific situation.
According to the framework used in LTGA exercise , an activation of the adaptation should be restricted to falls in financial market prices that are i) unforeseen, ii) sharp and
5
iii) steep. Maximum drawdowns are considered to be a suitable measure for determining such a situation. This measure calculates the maximum cumulated loss during a (rolling) observation period: the maximum loss from a local market maximum to a local market minimum. It therefore considers all information available during the observation period. •
Maximum drawdowns should be calculated for a portfolio. While in the historic analysis the weights of different asset classes in this portfolio were assumed to be constant, future data availability under Solvency II reporting will allow for regular (annually or quarterly) updates of portfolio weights.
•
Where a set of market indicators is used to support a decision on the use of an adaptation, this requires the setting of appropriate thresholds that would indicate an activation or a de-activation of an adaptation.
It is recommended to base the activation on a quantile threshold which is sufficiently high to meet the “unforeseen” criterion. In order to prevent the pricing-in of activation decisions and thus considerably reducing the power of the measure, it is not recommended to disclose all the details of the mechanism. For the same reason, it is not advisable to set a fixed quantile; instead, some discretionary room for maneuver is warranted. Deactivation should also be based on a quantile threshold which should be lower than the activation quantile, i.e. the situation in financial markets should have stabilized, at least to a certain extent. This stabilization should be double-checked against the supportive and real-economy indices. As for the activation, the indicator would only inform the decision making body and form the basis for the discretionary activation decision.
1.
General principles for the activation of the adaptation
1.1.
Context
The framework used for LTGA exercise defines three preconditions for the application of an adaptation of the risk-free interest rate, also known as Counter-Cyclical Premium (CCP) 1. These are: a)
the spread between credit risk-free and liquid assets in the representative portfolio is caused by illiquidity or excess credit risk of the issuer;
b)
it is likely that the excessive spread mentioned under (a) will lead companies to engage in fire sales of parts of their asset holdings;
c)
there is a fall in financial markets which is unforeseen, sharp and steep.
This appendix is intended to contribute to the development of a framework for setting up a clear, transparent and swift decision process by EIOPA on whether and in which situations the CCP will be applied. It is focused on the description of a set of market
6 1
Both terms “adaptation” and “counter-cyclical premium (CCP)” will be used interchangeably throughout this document.
indicators which EIOPA could take into account for determining a stressed situation of financial markets for a given currency.
1.2.
Definition of sharp, steep and unforeseen fall in financial markets
According to the framework applied, an activation of the CCP should be restricted to falls in financial market prices that are i) unforeseen, ii) sharp and iii) steep. In order to provide a parameterisation of these three sub-criteria the following is suggested: •
Regarding (i): A market price fall is qualified as being “unforeseen” if the reference measure takes on values that are “non-normal”, in the sense that such values are expected to be observed only rarely. In statistical terms, this means that “unforeseenness” is measured by a deviation from model expectations, defined on an assumed or empirical distribution of the reference measure and taking into account an appropriate level of confidence. Alternatively, and probably more correctly, one could devise an equilibrium model for the markets of interest and then use this model to pin-point situations where actual market prices are significantly different from model predictions. However, while such an approach is possible, it is judged that a first approximation to the draft Level 2 text is achieved better by a purely empirical approach rather than the modelling avenue. First, every model is subject to model-risk; second, it is time-consuming to conceptualise and to calibrate; and third, it may be challenging to make a complicated modelling framework completely transparent and replicable. Due to these practical considerations, i.e. the deadline under which the preparatory CCP activation work is done, as well as the mentioned structural reasons, i.e. replicability, transparency and time-consistency, it has been suggested to explore the empirical definition-avenue.
•
Regarding (ii) and (iii): It is not clear whether it is necessary to draw a distinction between sharp and steep market movements. One interpretation could be that sharpness refers to the initiation of the price decreases, i.e. that the fall is initiated by a sudden large one-day negative return, which is then followed by a series of additional days with negative returns. Together, this chain of negative returns would then together form a sharp and steep market movement. However, for practical reasons it is suggested not to draw this distinction between sharpness and steepness, and thus to parameterise a steep and sharp market movement simply as a chain of consecutive negative movements with a cumulative negative return larger than a certain pre-determined threshold value. Such consecutive return chains are also called “draw-downs”. Once the draw-downs of a given timeseries have been determined, it is naturally also possible to generate the distribution of the draw-downs and thus to determine the appropriate threshold value on the basis of a confidence level. This is done here for the above mentioned variables.
1.3. •
Criteria for assessing the suitability of indicators feeding into the activation and deactivation decision 7 This appendix does not aim for a concrete recommendation on the final calibration of the indicators which would inform activation or deactivation decision, instead it presents a set of criteria which could be used to decide on the actual calibration. According to this set of criteria, the (set of) indicators to be used:
•
should correctly depict those situations when insurers potentially get into trouble, due to a fall in financial markets which is o
sharp,
o
steep,
o
unforeseen.
•
should not depict situations when insurers, on aggregate, do fine.
•
should not lead to too frequent activation or deactivation decisions.
•
should be transparent.
2.
Set of indicators and data availability
It is proposed to have four types of indicators: primary indicators, supportive indicators, complementary indicators and real economy indicators. Most emphasis will be placed on the following primary indicators: •
Government bond spreads,
•
Corporate bond spreads,
•
Equity indices.
In this version of the document, spreads are derived by subtracting the overnight interest swap rate (OIS) – as a proxy for risk-free rates – from the redemption yield of the bond indices. There are however two analytical drawbacks to such an approach: •
The duration of the bonds and the OIS rates do not match – the government bond indices have a synthetic maturity of three years while the OIS rates are available for a maturity of one month.
•
The use of OIS rates as a proxy for the risk-free interest rate which is inconsistent with the draft Level 2 text.
Therefore the future implementation to derive the historic development of the risk-free interest rate term structures might be based on the currently available specifications (i.e. those used for the long-term guarantee impact assessment) – for the maturities of government bond and corporate bond indices, the relevant risk-free rates from these term structures should be used as a basis to calculate the spreads. While the development of these time series would be directly linked to the market value of assets held by insurance companies, some other time series could be used as supportive indicators to cross-check the analysis. Such supportive indicators can e.g. be
8
•
observed volatility of FX rates (against EUR and/or USD),
•
observed volatility of short-term interest rates.
Composite indicators of systemic risk like the CISS indicator of the ECB or other indexes can complement the presentation of results. However, given their underlying indicators
they would probably not add too much new information. See Annex 3 for further details on the composition of these indices. Finally, it is recommended to double-check the findings resulting from the analysis of financial market developments against indicators from the real economy which could signal a crisis situation. Such indicators could e.g. be •
inflation (consumer prices and maybe producer and/or import prices),
•
industrial production,
•
confidence indicators (e.g. purchase managers or consumers).
These indicators are widely available on a monthly basis and would cover various external shocks to the real economy which could translate into financial market stress. At this stage, it is thought such indicators should be used more to support the deactivation mechanism as they are usually lagging financial market developments. Given the rather different nature of financial crises, it is not expected that every crisis would be observable in all above-mentioned real economy indicators: While in some instances a crisis might be associated with an inflationary environment, other crises might occur in a surrounding characterized by sluggish demand and/or low consumer confidence.
3. 3.1.
Maximum Drawdown Model Calculation of maximum drawdowns in the equity market
The maximum drawdown approach identifies the largest return in a period of cumulated returns. Maximum drawdowns have been calculated using the formula below:
𝑀𝑎𝑥𝐷𝑟𝑎𝑤𝑑𝑜𝑤𝑛[0,𝑇]
⎛ = −𝑚𝑖𝑛𝑡∈[0,𝑇] ⎜ ⎝
𝑥 1 + 𝑙𝑛 � 𝑡 � 𝑥0
𝑥 1 + 𝑚𝑎𝑥𝑠∈[0,𝑡] �𝑙𝑛 � 𝑠 �� 𝑥0
⎞ − 1⎟ ⎠
With x t and x s being the current states in the considered period ([0,t] , [0, T] resp., T is the last value in the period of 1 year) Note: the maximum drawdown is shown as a positive number! The drawdown is calculated over a defined period T of 1 year (i.e. 261 trading days), using a rolling window on the data set. Maximum drawdown is a widely used risk measure in financial portfolio analysis, where it is assumed to better reflect the risk preferences of investors. It rather complements other risk measures (like value at risk) than replacing them, as each of the risk measures reveals different aspects of portfolio risk, and there will be no “best” risk measure.
9
Maximum drawdown calculates the maximum cumulated loss during an observation period: the maximum loss from a local market maximum to a local market minimum. It therefore considers all information available during the observation period (not just the difference between starting value and end value like value-at-risk). Additionally, it is a cumulative loss, not a one-period loss. An additional advantage of this approach is that it does not rely on any a priori assumptions about the return distribution or timedependence of returns (i.e. as a cumulative loss measure it can deal with auto-correlated returns, in contrast to the VaR-approach). 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0,2 0,1 0
MSCIEMU_MaxDrawdowns 261 days
MSCIEMU_MaxDrawdowns 66 days
F IGURE 1: M AXIMUM D RAWDOWNS CALCULATED OVER A PERIOD OF MONTHS (66 TRADING DAYS ) FOR MSCI EMU
1 YEAR (261 TRADING DAYS) AND 3
Calculated for MSCI EMU over a rolling window of 261 (trading) days (1 year) and 66 trading days (about 3 months), the maximum drawdown over 3 months is much more volatile than over a 1-year period (as would be expected), but as the adaptation mechanism provides for a minimum activation period of 1 year, once it is provoked, there is no use for the information a more volatile maximum drawdown could provide. The maximum drawdown approach considers all desirable aspects of methodology informing the activation decision:
10
•
the market changes should be sharp – the maximum drawdown value will be more prominent in this case;
•
the market changes should be steep – the maximum drawdown approach considers consecutive losses as a cumulative loss measure;
•
the market changes should be unforeseen – this will be captured by using a quantile approach for the activation mechanism based on the maximum drawdown distribution.
Using historic return values of a time series, a bootstrapping method is used to calculate possible performance scenarios of the time series. The bootstrapping method may be applied to all time series from which returns can be derived in a reasonable manner. Currently, pieces of the length of 3 months are cut randomly out of the used time series and built together to form a new scenario time series of the length of the original time series. In this way, a set of possible performance scenarios can be calculated. The drawdowns determined on these scenarios provide a drawdown distribution that can be used to evaluate the quality of the above calculated drawdowns for each time series. Maximum Drawdowns and Trigger at 70% quantile for MSCI EMU log returns m axdd Log MSCI EMU Trigger LogReturns 70% quantile 1,2
1,0
0,8
0,6
0,4
0,2
0,0
F IGURE 2: MAXIMUM D RAWDOWNS CALCULATED OVER A PERIOD OF 1 YEAR FOR MSCI EMU, AND THRESHOLDS BASED ON 70% QUANTILE OF MAXIMUM DRAWDOWN DISTRIBUTION FOR 500 SCENARIOS
In the above figure, the maximum drawdown is calculated over a rolling window of 1 year, the drawdown distribution is determined using 500 scenarios from the bootstrapping algorithm, and a quantile of 70% of this scenario-distribution is used as (indicative) threshold for the activation mechanism. The bootstrapping algorithm may need more than 500 runs to obtain stable quantiles, but a comparison of 500 runs and 1000 runs for the above shown MSCI EMU price index shows rather close results. Therefore, for equity data the bootstrapping algorithm is run for 500 simulations, to reduce computation time.
11
MSCI EMU 500 simulations 1000 simulations 50% 0.1654 0.1647 55% 0.1811 0.1800 60% 0.1981 0.1968 65% 0.2164 0.2149 70% 0.2367 0.2345 75% 0.2617 0.2591 80% 0.2893 0.2873 85% 0.3226 0.3198 90% 0.3667 0.3642 95% 0.4370 0.4370 97.5% 0.5037 0.5067 99% 0.5875 0.5932 T ABLE 1: COMPARISONS OF SOME QUANTILES FOR MAXIMUM DRAWDOWN DISTRIBUTION CALCULATED FOR SIMULATIONS ( LEFT COLUMN ) AND 1000 SIMULATIONS ( RIGHT COLUMN ) OF MSCI EMU PRICE INDEX
500
•
The bootstrapping algorithm may also be run on several time series together to take into account the correlation structure of the data.
•
While for this analysis price indices have been use, one might argue that return indices would provide a more relevant picture as they include dividends and interest. Firms may use dividends in particular to give a better impression of conditions during stressed times. This may shift the activation/deactivation periods out of sync with actual market conditions. This effect is examined closer by EIOPA and the analysis shows that there is no significant difference in activation periods between the two approaches. From an analytical point of view, nevertheless, the use of price indices seems advantageous since usually the time series of price indices are available for a longer period, so historic assessments
are more valid. 3.2.
Drawdown distributions in the equity market
The following graphs compare the distributions of maximum drawdowns over a period of 1 year, calculated on historic values of MSCI EMU price index daily log returns only, and calculated for 500 scenarios of possible developments of MSCI EMU daily log returns. The distribution for the scenarios indicates a much wider but smoother distribution with a heavier tail: values up to 1.15 can occur for the simulated values, where the maximum drawdown values for the historic distribution are not larger than 0.8. Therefore the simulated distribution covers a wider range of possible drawdown behavior than the historic distribution with the advantage of a much smoother progression. This helps to avoid large changes in the quantile values.
12
F IGURE 3: M AXIMUM DRAWDOWN DISTRIBUTION CALCULATED FOR 500 SIMULATIONS OF DAILY MSCI LOG RETURNS ( UPPER GRAPH ) AND FOR HISTORIC DAILY MSCI EMU LOG RETURNS ( LOWER GRAPH )
EMU
MSCI EMU Quantile simulation distribution Quantile original distribution 50% 0.165 0.134 55% 0.181 0.151 60% 0.198 0.173 65% 0.216 0.212 70% 0.237 0.276 75% 0.262 0.305 80% 0.289 0.336 85% 0.323 0.366 90% 0.367 0.430 95% 0.437 0.565 97.5% 0.504 0.665 99% 0.587 0.737 T ABLE 2: COMPARISONS OF SOME QUANTILES FOR MAXIMUM DRAWDOWN DISTRIBUTION CALCULATED FOR SIMULATIONS ( LEFT COLUMN ) AND THE ORIGINAL HISTORIC VALUES ( RIGHT COLUMN ) OF MSCI EMU
500
Table 3 provides an overview of the drawdown distributions for various MSCI country indices. For each country, the left column depicts the results of the bootstrapping simulation while the right column shows the historically observed distribution. Note that the lengths of the time series differ across countries so that drawdown distributions are not directly comparable – for further details refer to Annex 2.
13
It can be seen that for most countries the equity index exhibits a wider distribution of maximum drawdowns than for the MSCI EMU. This is in line with empirical evidence that indices from smaller markets tend to be less diversified and therefore more volatile. Table 3 shows that only for some larger markets like e.g. France, Germany, Netherlands, Spain, United Kingdom, the drawdown distribution is very similar to the one of the MSCI EMU.
14
Quantiles 50% 55% 60% 65% 70% 75% 80% 85% 90% 95% 97,5% 99% Quantiles 50% 55% 60% 65% 70% 75% 80% 85% 90% 95% 97,5% 99% Quantiles 50% 55% 60% 65% 70% 75% 80% 85% 90% 95% 97,5% 99%
Austria Belgium Bulgaria Estonia Finland France Czech Republic Denmark simulated historic simulated historic simulated historic simulated historic simulated historic simulated historic simulated historic simulated historic 0,186 0,180 0,152 0,130 0,321 0,311 0,225 0,218 0,147 0,126 0,233 0,211 0,246 0,228 0,166 0,146 0,202 0,201 0,168 0,151 0,351 0,363 0,239 0,234 0,160 0,156 0,248 0,236 0,267 0,256 0,180 0,161 0,220 0,218 0,185 0,167 0,385 0,429 0,255 0,246 0,173 0,182 0,265 0,256 0,290 0,289 0,196 0,175 0,239 0,234 0,203 0,205 0,424 0,483 0,274 0,258 0,187 0,205 0,287 0,284 0,315 0,332 0,215 0,196 0,259 0,253 0,222 0,229 0,472 0,518 0,295 0,271 0,203 0,226 0,313 0,298 0,343 0,374 0,237 0,259 0,283 0,269 0,244 0,261 0,531 0,635 0,320 0,302 0,221 0,246 0,342 0,355 0,374 0,410 0,263 0,315 0,313 0,295 0,272 0,284 0,611 0,745 0,350 0,339 0,242 0,265 0,380 0,427 0,412 0,443 0,293 0,346 0,354 0,346 0,310 0,326 0,734 0,957 0,392 0,405 0,267 0,311 0,430 0,552 0,460 0,496 0,330 0,377 0,415 0,424 0,368 0,401 0,926 1,528 0,452 0,450 0,306 0,345 0,503 0,785 0,522 0,605 0,376 0,450 0,532 0,579 0,466 0,572 1,175 1,692 0,550 0,643 0,383 0,410 0,616 0,929 0,627 0,802 0,445 0,531 0,692 1,093 0,620 0,833 1,365 1,750 0,645 0,697 0,464 0,611 0,715 1,003 0,724 0,895 0,513 0,591 0,901 1,218 0,764 1,115 1,580 1,791 0,752 0,768 0,547 0,689 0,839 1,092 0,849 0,983 0,606 0,657 Ireland Italy Poland Germany Greece Hungary Lithuania Netherlands simulated historic simulated historic simulated historic simulated historic simulated historic simulated historic simulated historic simulated historic 0,168 0,141 0,298 0,283 0,243 0,241 0,211 0,182 0,207 0,189 0,215 0,196 0,142 0,121 0,259 0,243 0,184 0,150 0,322 0,307 0,261 0,263 0,229 0,202 0,223 0,222 0,231 0,200 0,155 0,126 0,280 0,287 0,202 0,169 0,349 0,365 0,281 0,286 0,248 0,226 0,242 0,244 0,253 0,210 0,172 0,135 0,304 0,312 0,222 0,202 0,380 0,415 0,303 0,307 0,270 0,253 0,263 0,275 0,281 0,228 0,192 0,146 0,331 0,333 0,248 0,245 0,414 0,460 0,327 0,334 0,295 0,290 0,286 0,295 0,310 0,262 0,215 0,197 0,362 0,357 0,280 0,298 0,452 0,489 0,357 0,415 0,322 0,338 0,312 0,344 0,348 0,278 0,239 0,247 0,399 0,413 0,317 0,347 0,498 0,524 0,399 0,480 0,354 0,369 0,343 0,385 0,401 0,293 0,268 0,292 0,443 0,485 0,359 0,385 0,555 0,575 0,455 0,521 0,398 0,397 0,378 0,423 0,459 0,299 0,314 0,373 0,494 0,539 0,408 0,487 0,631 0,702 0,541 0,567 0,467 0,463 0,425 0,466 0,550 0,535 0,372 0,426 0,562 0,595 0,670 0,697 0,449 0,520 0,678 0,752 0,490 0,606 0,749 1,026 0,679 0,915 0,592 0,769 0,498 0,527 0,564 0,687 0,862 1,106 0,792 1,009 0,702 1,195 0,564 0,722 0,762 0,705 0,514 0,679 0,794 0,823 0,669 0,727 1,000 1,186 0,924 1,044 0,828 1,331 0,646 0,807 0,870 0,716 0,602 0,759 0,957 0,975 Portugal Romania Slovenia Spain Sweden United Kingdom simulated historic simulated historic simulated historic simulated historic simulated historic simulated historic 0,206 0,191 0,308 0,281 0,182 0,173 0,188 0,170 0,183 0,158 0,153 0,133 0,221 0,215 0,338 0,318 0,196 0,188 0,202 0,195 0,198 0,173 0,165 0,150 0,239 0,234 0,375 0,332 0,210 0,199 0,219 0,220 0,216 0,197 0,179 0,169 0,257 0,267 0,417 0,340 0,232 0,208 0,237 0,257 0,237 0,218 0,193 0,183 0,277 0,299 0,469 0,363 0,258 0,232 0,257 0,297 0,260 0,284 0,209 0,196 0,300 0,331 0,529 0,401 0,292 0,262 0,280 0,324 0,286 0,328 0,227 0,212 0,328 0,364 0,594 0,784 0,329 0,289 0,309 0,354 0,317 0,391 0,252 0,235 0,361 0,412 0,677 0,979 0,371 0,387 0,346 0,385 0,356 0,432 0,283 0,279 0,405 0,461 0,779 1,185 0,433 0,601 0,390 0,408 0,406 0,475 0,328 0,367 0,474 0,514 0,934 1,439 0,542 0,799 0,454 0,470 0,478 0,600 0,391 0,467 0,537 0,615 1,082 1,512 0,629 0,932 0,522 0,554 0,549 0,670 0,449 0,567 0,611 0,746 1,273 1,621 0,740 0,988 0,603 0,688 0,646 0,714 0,529 0,817
T ABLE 3: C OMPARISONS OF SOME QUANTILES OF THE MAXIMUM DRAWDOWN DISTRIBUTION CALCULATED FOR 500 SIMULATIONS ( LEFT COLUMN ) AND THE ORIGINAL HISTORIC VALUES ( RIGHT COLUMN ) OF VARIOUS MSCI COUNTRY INDICES
15
3.3.
Indicator analysis in the equity market 2
The choice of the quantile used as a threshold influences on the one hand the number of activation events and on the other hand the duration of an activation period. For the quantiles of historic values as well as simulated drawdowns, different threshold quantiles are analyzed for different time periods: the whole data set (29.12.1988 – 08.02.2013 for the MSCI EMU price index), the last 10 years and the last 5 years. Data is available only on trading days, so a year is assumed to consist of 261 trading days. The results are aggregated in the following tables: •
the number of activation events in the analyzed period,
•
the percentage of time where the adaptation was activated,
•
the mean and median length of the observed activation periods in days.
In this set of tables, the quantiles for activation and deactivation of the adaptation are the same, and no minimum period for the activation is applied – this may lead to a bulk of short activation/deactivation events instead of some longer activation events for some quantile combinations. Note that only the indicator “% active trigger time” is a (decreasing) function of the quantile while “number trigger moments” and “mean/median active trigger” can vary significantly and can be both higher or lower for higher quantiles.
all data Quantile
Nb trigger moments % active Trigger time mean active trigger (d) median active trigger (d) historic simulated historic simulated historic simulated historic simulated 17 22 12 30 32 172 181 70% 11 23 68 322 8 244 75% 5 25 31 22 20 28 66 98 4 13 80% 19 85% 30 15 15 22 31 100 5 8 90% 12 28 10 16 52 33 3 5 8 5 10 63 52 3 3 95% 5
T ABLE 4: A CTIVATION ANALYSIS FOR THE WHOLE DATA SET (29.12.1988 – 08.02.2013), HISTORIC AND SIMULATED QUANTILES ARE USED AS ACTIVATING AND DEACTIVATING THRESHOLDS (MSCI EMU)
last 10 years Nb trigger moments % active Trigger time mean active trigger (d) median active trigger (d) Quantile historic simulated historic simulated historic simulated historic simulated 70% 6 8 36 39 156 144 108 22 10 3 31 37 80 242 11 246 75% 80% 8 13 26 33 86 78 3 13 85% 22 8 20 28 23 123 4 81 90% 2 19 14 21 183 26 183 5 95% 4 2 12 14 78 182 34 182
16
2
Note that tables in this chapter depict information like “number of trigger moments”, “% active trigger time” and “mean/median active trigger”. This does not imply that the activation/deactivation decision is mechanistically relying on a fully-automated trigger mechanism. Instead, activation and deactivation decisions are taken by EIOPA on a discretionary basis.
T ABLE 5: A CTIVATION ANALYSIS FOR THE LAST 10 YEARS, HISTORIC AND SIMULATED QUANTILES ARE USED AS ACTIVATING AND DEACTIVATING THRESHOLDS (MSCI EMU)
last 5 years Nb trigger moments % active Trigger time mean active trigger (d) median active trigger (d) Quantile historic simulated historic simulated historic simulated simulated historic 55 70% 2 58 358 151 359 10 5 75% 7 2 47 56 87 244 9 253 51 80% 4 9 41 134 94 91 13 85% 18 4 29 44 21 191 3 214 90% 1 15 20 32 264 22 264 4 95% 3 1 19 20 82 262 3 262 T ABLE 6: A CTIVATION ANALYSIS FOR THE LAST 5 YEARS, HISTORIC AND SIMULATED QUANTILES ARE USED AS ACTIVATING AND DEACTIVATING THRESHOLDS (MSCI EMU)
Where mean and median values of the active trigger periods differ significantly, there are more activation events which last only for a few days. It is considered useful to use different quantiles for activation and deactivation. While the activation quantile should refer to a situation which is “unforeseen” (thereby implying a relatively high quantile), the deactivation quantile should depict a situation which has (at least somewhat) normalized. Using exactly the same quantile for both activation and deactivation might however result in a mechanism which switches the CCP on and off quite often and potentially multiple times within a short period of time. In this case only the condition to keep the CCP activated for at least 12 months, would add some continuity. Table 7 and Table 8 show the activation analysis for different activation and deactivation quantiles of the simulated maximum drawdown distribution, again for the whole dataset, the last 10 years and the last 5 years. In the second table, the adaptation is assumed to be activated for at least one year after initial activation, i.e. 261 trading days. This means, once the adaptation is activated, it is deactivated only after at least 261 days.
Especially for higher quantiles and larger differences between activation quantile and deactivation quantile, the effects on the number of activation moments and the length of the active adaptation periods are negligible. However, the lower the activation quantile and the closer activation and deactivation quantile, the greater effect of a minimum activation period. Then Table 9 provides an overview of various activation/deactivation quantile combinations for all EU countries for which an MSCI equity index is available, again for the whole dataset, the last 10 years and the last 5 years. Dashes (“—”) indicate that no complete time series is available for the specific time period (e.g. in the case of Lithuania, the full time series has a length of only 4.5 years, so no figures are provided in the columns for 5 years and 10 years).
17
activation quantile /deactivation quantile 70/50 70/55 70/60 70/65 70/70
all data
Nb trigger moments last 10 Y last 5Y 5 3 5 3 5 3 5 3 11 7
all data 2 2 2 2 5
% active Trigger time last 10 Y last 5Y 35 46 33 40 32 40 32 40 32 39
all data 70 61 60 60 58
mean active trigger (d) median active trigger (d) last 10 Y last 5Y all data last 10 Y last 5Y 441 397 458 391 391 410 348 394 262 262 407 346 391 258 258 406 344 389 257 257 180 144 151 22 22
458 394 391 389 11
75/50 75/55 75/60 75/65 75/70 75/75
5 5 5 5 5 6
3 3 3 3 3 4
2 2 2 2 2 3
34 32 32 31 31 31
44 39 38 38 38 37
69 59 58 58 57 56
430 400 396 395 389 322
384 336 333 332 325 242
446 382 379 378 369 243
391 262 258 257 256 244
391 262 258 257 256 246
446 382 379 378 369 253
80/50 80/55 80/60 80/65 80/70 80/75
5 5 5 5 5 5
3 3 3 3 3 3
2 2 2 2 2 2
33 31 31 31 30 30
43 37 37 37 36 36
65 55 55 55 53 53
420 389 385 385 378 376
369 320 318 316 310 308
423 359 356 355 346 343
389 260 256 255 254 253
389 260 256 255 254 253
423 359 356 355 346 343
85/50 85/55 85/60 85/65 85/70 85/75
5 5 5 5 5 5
3 3 3 3 3 3
2 2 2 2 2 2
30 27 27 27 26 26
40 35 35 34 34 33
61 51 51 50 49 49
373 342 339 338 331 329
351 302 300 298 291 290
396 331 329 327 318 316
366 237 233 232 231 230
366 244 241 240 238 238
396 331 329 327 318 316
90/50 90/55 90/60 90/65 90/70 90/75
4 4 4 4 4 4
3 3 3 3 3 3
2 2 2 2 2 2
24 22 22 22 21 21
39 33 33 33 32 32
58 48 48 48 46 46
379 343 340 339 331 330
339 290 287 286 279 278
378 313 311 309 300 298
377 313 311 309 300 298
338 244 241 240 238 238
378 313 311 309 300 298
95/50 95/55 95/60 95/65 95/70 95/75
2 2 2 2 2 2
2 2 2 2 2 2
1 1 1 1 1 1
16 16 15 15 15 15
24 23 23 23 22 22
27 27 27 27 26 26
497 488 486 485 475 473
309 301 299 297 288 286
357 357 356 354 337 333
497 488 486 485 475 473
309 301 299 297 288 286
357 357 356 354 337 333
T ABLE 7: A CTIVATION ANALYSIS FOR DIFFERENT ACTIVATING / DEACTIVATING QUANTILES OF SIMULATED MAXIMUM DRAWDOWN DISTRIBUTION (MSCI EMU) MSCI EMU Price Index Start 31.12.1988 Nb trigger moments activation quantile /deactivation quantile all data last 10 Y last 5Y 70/50 5 3 70/55 5 3 70/60 5 3 70/65 5 3 70/70 6 4 75/50 5 3 75/55 5 3 75/60 5 3 75/65 5 3 75/70 5 3 75/75 6 3 80/50 5 3 80/55 5 3 80/60 5 3 80/65 5 3 80/70 5 3 80/75 5 3 85/50 5 3 85/55 5 3 85/60 5 3 85/65 5 3 85/70 5 3 5 3 85/75 3 90/50 4 90/55 4 3 90/60 4 3 90/65 4 3 90/70 4 3 90/75 4 3 95/50 2 2 95/55 2 2 95/60 2 2 95/65 2 2 95/70 2 2 95/75 2 2
18
% active Trigger time all data 2 2 2 2 3 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1
last 10 Y 35 32 32 32 36 35 32 32 32 32 36 33 31 31 31 31 30 33 31 31 31 31 30 25 25 25 25 24 24 16 16 15 15 15 15
mean active trigger (d)
last 5Y 44 39 39 39 49 44 39 39 39 38 38 41 37 36 36 36 36 41 36 36 36 36 36 37 37 36 36 35 35 24 24 24 24 23 23
all data 69 59 58 58 77 68 58 58 58 57 56 61 53 53 53 51 51 61 53 53 52 51 51 53 53 52 52 50 50 27 27 27 27 26 26
last 10 Y 437 408 407 406 379 434 405 404 403 399 374 416 391 390 389 385 382 414 389 388 388 384 380 399 395 389 388 383 379 497 488 486 484 474 466
median active trigger (d)
last 5Y 385 342 341 340 316 383 340 339 339 332 332 354 317 316 316 309 309 351 316 315 315 308 308 319 319 313 312 306 305 310 310 308 307 298 297
all data 446 382 380 379 333 443 380 378 377 368 367 400 345 344 343 333 333 396 344 342 341 332 331 348 348 338 337 328 327 357 357 354 352 333 332
last 10 Y 390 261 261 261 261 388 261 261 261 261 261 370 261 261 261 261 261 365 261 261 261 261 261 348 348 338 337 328 327 497 488 486 484 474 466
last 5Y 390 262 262 262 262 388 262 262 262 262 262 370 262 262 262 262 262 365 262 262 262 262 262 277 277 262 262 262 262 310 310 308 307 298 297
446 382 380 379 261 443 380 378 377 368 367 400 345 344 343 333 333 396 344 342 341 332 331 348 348 338 337 328 327 357 357 354 352 333 332
T ABLE 8 : ACTIVATION ANALYSIS FOR DIFFERENT ACTIVATING /DEACTIVATING QUANTILES OF SIMULATED MAXIMUM DRAWDOWN DISTRIBUTION (MSCI EMU PRICE INDEX ), WHEN THE ADAPTATION IS ASSUMED TO BE ACTIVE FOR AT LEAST A YEAR , I . E . 261 TRADING DAYS
3.4.
Crisis periods in the equity market
The following tables summarize the information of the previous tables by presenting, for one specific activation/deactivation quantile combination, the actual periods during which the adaptation was activated in the equity market. The depicted activation/deactivation quantile combination of 95% and 75%, respectively, is for illustration purposes only and should not prejudge any decision on the final calibration.
Maximum Drawdowns and trigger activation at 95% quantile, deactivation at 75%, and a minimum of 261 trading days for trigger activation for MSCI EMU log returns m axdd Log MSCI EMU Trigger 95%/75% 1,2
1
0,8
0,6
0,4
0,2
0
F IGURE 4: MAXIMUM D RAWDOWNS CALCULATED OVER A PERIOD OF 1 YEAR FOR MSCI EMU PRICE INDEX, AND THRESHOLDS BASED ON 95% ACTIVATION QUANTILE AND 75% DEACTIVATION QUANTILE . T HE ADAPTATION IS ASSUMED TO BE ACTIVE FOR AT LEAST 1 YEAR (261 TRADING DAYS )
19
MSCI EMU
Switch on Switch off 95%
75%
Period 1
10.09.2001 16.01.2004
Period 2
02.10.2008 13.01.2010
Period 3
--
--
Period 4
--
--
Period 5
--
--
Period 6
--
--
T ABLE 9: C RISIS PERIODS (= ACTIVATED ADAPTATION ) IN EQUITY MARKETS (MSCI EMU PRICE INDEX, ASSUMING THRESHOLDS AT THE 95% QUANTILE AND A DEACTIVATION AT THE 75% QUANTILE )
It can be seen that the proposed methodology covers EU-wide and worldwide crisis situations like e.g. the burst of the dotcom bubble, aggravated by the terrorist attacks in September 2001, or the financial crisis erupting in Autumn 2008. In addition, the methodology is sensitive to country-/region-specific developments, as can be seen e.g. in the case of the Nordic banking crisis in 1991.
20
MSCI Austria Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
MSCI Bulgaria Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
MSCI Denmark Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
MSCI Finland Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on 95% 06.02.1991 18.09.2008 21.11.2011 ----
Switch off 75% 05.02.1992 11.01.2010 19.11.2012 ----
Switch on
Switch off
95%
75%
07.10.2008 ------
Switch on 95% 19.10.1984 05.10.1992 21.09.2001 06.10.2008 ---
Switch on 95% 10.01.1991 09.02.2001 23.09.2008 ----
13.11.2009 ------
Switch off 75% 21.10.1985 05.10.1993 18.09.2003 17.02.2010 ---
Switch off 75% 09.01.1992 23.06.2003 12.01.2010 ----
MSCI Belgium Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
MSCI Czech Republic Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
MSCI Estonia Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
MSCI France Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on 95% 09.10.2002 23.06.2008 -----
Switch on 95% 09.10.1998 10.10.2008 -----
Switch on 95% 05.02.2008 ------
Switch on 95% 05.06.1981 23.11.1987 11.09.2001 06.10.2008 ---
Switch off 75% 19.02.2004 11.11.2009 -----
Switch off 75% 11.10.1999 16.10.2009 -----
Switch off 75% 22.10.2009 ------
Switch off 75% 04.06.1982 21.11.1988 14.01.2004 18.11.2009 ---
21
MSCI Germany Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
MSCI Hungary Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
MSCI Italy Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
MSCI Netherlands Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
MSCI Portugal Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
22
Switch on 95% 10.11.1987 11.09.2001 08.10.2008 ----
Switch on 95% 02.04.1999 09.10.2008 -----
Switch on 95% 20.05.1982 03.02.1988 19.09.2001 06.10.2008 23.01.2012 --
Switch on 95% 06.04.1988 20.09.2001 29.09.2008 ----
Switch on 95% 05.01.1989 11.10.1990 14.09.2001 30.06.2008 23.05.2012 --
Switch off 75% 08.11.1988 16.01.2004 20.11.2009 ----
Switch off 75% 31.03.2000 19.11.2009 -----
Switch off 75% 19.05.1983 01.02.1989 18.09.2002 16.02.2010 21.01.2013 --
Switch off 75% 05.04.1989 03.02.2004 09.02.2010 ----
Switch off 75% 04.01.1990 10.10.1991 14.07.2003 01.10.2009 --
MSCI Greece Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
MSCI Ireland Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
MSCI Lithuania Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
MSCI Poland Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
MSCI Romania Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on 95% 04.07.1991 16.10.2008 26.08.2010 ----
Switch on 95% 10.06.2008 ------
Switch on 95% ------
Switch on 95% 16.11.1994 16.10.2008 -----
Switch on 95% 22.10.2008 ------
Switch off 75% 02.07.1992 22.10.2009 ----
Switch off 75% 13.01.2010 ------
Switch off 75% 31.05.2010 ------
Switch off 75% 15.11.1995 15.10.2009 -----
Switch off 75% 19.01.2010 ------
MSCI Slovenia Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
MSCI Sweden Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on 95% 17.09.2008 ------
Switch on 95% 07.10.1988 07.01.1991 14.04.1999 23.02.2001 07.07.2008 --
Switch off 75% 18.11.2009 ------
Switch off 75% 09.10.1989 07.01.1992 13.04.2000 28.08.2003 06.10.2009 --
MSCI Spain Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
MSCI United Kingdom Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on 95% 24.09.1990 21.09.2001 10.10.2008 19.04.2012 ---
Switch on 95% 06.12.1973 09.05.1988 19.11.2002 08.10.2008 ---
Switch off 75% 23.09.1991 23.06.2003 13.01.2010 ----
Switch off 75% 12.11.1975 08.05.1989 02.12.2003 13.01.2010 ---
T ABLE 10: C RISIS PERIODS (=ACTIVATED ADAPTATION ) IN EQUITY MARKETS (MSCI COUNTRY INDICES , ASSUMING THRESHOLDS AT THE 95% QUANTILE AND A DEACTIVATION AT THE 75% QUANTILE )
3.5.
Indicators for other asset classes
While for the analysis in the equity market a maximum drawdown approach has been used, there are different approaches available for government bond and corporate bond spreads. As there is no index readily available for which the maximum drawdown could be derived, one approach could be to simply base the activation/deactivation quantile on the distribution of absolute spread values; another approach would be to build an index based on spread changes. In this chapter, the first proposal (distribution of absolute spread values) is presented. The spreads are derived by comparing the redemption yield of a benchmark bond (3-year maturity 3, provided by DataStream for most EU currencies and countries) with the 1month overnight interest rate swap rate; for corporate bonds, Barclay’s indices are used. The difference in the maturity of bonds and the base rate (OIS) can be seen as problematic, as it includes the “normal” steepness of the interest rate term structure in addition to the credit spread. As mentioned above, it is suggested to use the risk-free interest rate term structure along the specifications of the LTGA framework instead of OIS rates for future analyses.
3
The 3-year maturity has been chosen solely for the purpose of data availability in order to minimize the maturity gap between the sovereign bond index and the proxy for the risk-free interest rate. For the future implementation, when the risk-free rate is derived based on the principles set out in the LTGA framework (matching the maturity of the sovereign bond index), more common maturities like 5 years or 10 years should be used.
23
3.5.1 Government bonds The following tables show the activation/deactivation moments based on absolute government bond spreads and the 95%/75% activation/deactivation quantile combination for all EU countries with available data. Government Bonds Austria Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Government Bonds Czech Republic Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Government Bonds Finland Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Government Bonds Germany Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Government Bonds Ireland
24
Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on 95% 09.08.1999 26.01.2009 16.11.2011 ----
Switch on
Switch off 75% 07.08.2000 17.05.2010 14.11.2012 ----
Switch off
95% 23.02.2009 01.03.2011
75% 22.02.2010 28.02.2012
----
----
Switch on 95% 20.10.1999 01.04.2009 01.04.2011 ----
Switch on 95% 03.08.1999 15.07.2009 -----
Switch on 95% 15.03.2011 ------
Switch off 75% 18.10.2000 06.04.2010 30.03.2012 ----
Switch off 75% 01.08.2000 14.07.2010 -----
Switch off 75% 11.10.2012 ------
Government Bonds Belgium Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Government Bonds Denmark Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Government Bonds France Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Government Bonds Greece Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Government Bonds Italy Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on 95% 20.10.1999 05.08.2009 14.12.2010 ----
Switch on 95% 28.08.2009 10.02.2011 -----
Switch on 95% 05.08.1999 05.06.2009 08.03.2011 ----
Switch on 95% 06.12.2011 ------
Switch on 95% 18.07.2011 ------
Switch off 75% 18.10.2000 04.08.2010 02.05.2012 ----
Switch off 75% 27.08.2010 09.02.2012 -----
Switch off 75% 03.08.2000 04.06.2010 06.03.2012 ----
Switch off 75% -------
Switch off 75% -------
Government Bonds Malta Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Government Bonds Poland Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Government Bonds Slovenia Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Government Bonds United Kingdom Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on 95% 14.07.2009 04.03.2011 -----
Switch on 95% -20.02.2009 -----
Switch on 95% 30.11.2011 ------
Switch on 95% 16.03.2009 02.02.2011 -----
Switch off 75% 13.07.2010 13.04.2012 -----
Switch off 75% 06.07.2005 18.08.2010 -----
Switch off 75% -------
Government Bonds Netherlands Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Government Bonds Portugal Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Government Bonds Spain Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on 95% 03.08.1999 26.01.2009 -----
Switch on 95% 23.06.2011 ------
Switch on 95% 29.11.2010 ------
Switch off 75% 01.08.2000 26.04.2010 -----
Switch off 75% -------
Switch off 75% -------
Switch off 75% 23.06.2010 01.02.2012 -----
T ABLE 11: C RISIS PERIODS (= ACTIVATED ADAPTATION ) IN GOVERNMENT BOND MARKETS (3- YEAR D ATASTREAM BENCHMARK BONDS, ASSUMING THRESHOLDS AT THE 95% QUANTILE AND A DEACTIVATION AT THE 75% QUANTILE )
25
3.5.2 Corporate bonds The following tables show the activation/deactivation moments based on absolute corporate bond spreads for all EU countries with available data.
Corporate Bond Spreads AT Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on
Switch off
95%
75%
08.12.2008 ------
07.10.2010 ------
Corporate Bond Spreads BE Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on Switch off 95% 75% 30.09.2008 31.08.2010 -----------
Corporate Bond Spreads DE Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on Switch off 95% 75% 24.10.2008 19.08.2010 -----------
Corporate Bond Spreads ES Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on Switch off 95% 75% 26.01.2009 25.01.2010 13.09.2011 30.11.2012 ---------
Corporate Bond Spreads FI Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on Switch off 95% 75% 28.09.2001 27.09.2002 04.12.2008 04.05.2010 ---------
Corporate Bond Spreads FR Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on Switch off 95% 75% 08.12.2008 23.08.2010 23.11.2011 21.11.2012 ---------
Corporate Bond Spreads IT Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on Switch off 95% 75% 22.01.2009 21.01.2010 12.09.2011 05.12.2012 ---------
Corporate Bond Spreads NL Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on Switch off 95% 75% 24.11.2008 11.08.2010 -----------
T ABLE 12: C RISIS PERIODS (= ACTIVATED ADAPTATION ) IN CORPORATE BOND MARKETS (BARCLAYS INDEX FAMILY , ASSUMING THRESHOLDS AT THE 95% QUANTILE AND A DEACTIVATION AT THE 75% QUANTILE )
26
4.
Combination of individual indicators
4.1.
Portfolio approach
Instead of single indicators, the maximum drawdown method may be applied to an insurance portfolio comprising the three main asset classes, namely government bonds, corporate bonds and equity. An advantage of such an approach would be that it would allow for a higher degree of consistency with the quantification of the adaptation, which is based on the observation of spreads in a representative asset portfolio. Some methodological questions occur in connection with this approach: •
What data are used to approximate the single components of the insurance portfolio, as for the maximum drawdown approach returns (relative returns or log returns of portfolio performance) have to be calculated? In a first attempt, we have taken the price indices for equity (MSCI country indices as above), while for government bonds and corporate bonds and equivalent indicator had to be constructed. Therefore changes in absolute spreads have been translated into an indexed time series by multiplying the spread changes by the durations (8.8 years and 6.6 years for government bonds and corporate bonds, respectively, based on data provided by NSAs).
•
The average portfolio of the insurance industry in each country forms the basis for the analysis in this chapter. Portfolio weights are based on information provided by NSAs. It is noted that these weights can shift substantially over time, so without historic data on portfolio weights any back-testing can only be an approximation.
With regard to the future implementation of the adaptation activation mechanism, it is therefore recommended to assess the primary indicators as calculated for the whole (representative) portfolio per country/currency. Portfolio weights should be re-adjusted on a regular basis. 4
27 4 In practical terms, this means that at each adjustment date the historical distribution of maximum drawdowns would need to be simulated so that the thresholds can be derived.
•
In the following examples, the following portfolio weights have been used:
Government bonds 18,0%
Corporate bonds 56,2%
Belgium
57,2%
36,6%
6,2%
Bulgaria
71,6%
13,9%
14,5%
Austria
Cyprus
Equity 25,8%
n.a.
n.a.
n.a.
Czech Rep
65,1%
28,5%
6,4%
Denmark
21,3%
60,2%
18,4%
Estonia
45,0%
48,2%
6,8%
Finland
13,3%
62,8%
23,8%
France
35,3%
48,8%
15,9%
Germany
12,5%
78,8%
8,6%
n.a.
n.a.
n.a.
Hungary
86,9%
10,2%
2,9%
Ireland
81,9%
16,2%
1,9%
Italy
67,0%
26,7%
6,3%
Latvia
67,2%
27,3%
5,5%
Lithuania
98,1%
0,4%
1,6%
Luxemburg
60,4%
23,2%
16,5%
Malta
70,3%
21,9%
7,8%
Netherlands
62,2%
32,8%
5,0%
Poland
94,9%
3,4%
1,7%
Portugal
38,6%
55,5%
5,9%
Greece
Romania
n.a.
n.a.
n.a.
Slovakia
52,9%
45,0%
2,2%
Slovenia
44,9%
49,0%
6,1%
Spain
41,0%
53,0%
6,1%
Sweden
19,5%
52,2%
28,3%
United Kingdom
32,3%
53,9%
13,9%
T ABLE 13: A SSET ALLOCATION OF INSURANCE UNDERTAKINGS ( BASED ON INFORMATION PROVIDED BY NSA’ S )
The Table 13 shows that the weights can differ significantly between countries. Percentages in Table 13 above are calculated considering as 100 per cent the sum of bonds and equities. Annex 6 in this appendix shows the activation analysis for different activating/deactivating quantiles of simulated maximum drawdown distribution (portfolios), when the adaptation is assumed to be activated for at least a year, i.e. 261 trading days
28
Quantiles 50% 55% 60% 65% 70% 75% 80% 85% 90% 95% 97,5% 99%
Quantiles 50% 55% 60% 65% 70% 75% 80% 85% 90% 95% 97,5% 99%
Quantiles
Austria
Belgium
Bulgaria
Czech Republic
Denmark
Estonia
Finland
France
Germany
simulated historic simulated historic simulated historic simulated historic simulated historic simulated historic simulated historic simulated historic simulated historic 0,079 0,086 0,073 0,066 0,072 0,068 0,068 0,070 0,054 0,044 0,059 0,058 0,098 0,097 0,064 0,051 0,055 0,049 0,086 0,089 0,079 0,074 0,079 0,069 0,073 0,073 0,059 0,045 0,064 0,065 0,107 0,106 0,070 0,054 0,059 0,050 0,095 0,093 0,085 0,091 0,086 0,074 0,078 0,074 0,066 0,047 0,070 0,068 0,118 0,124 0,076 0,058 0,064 0,051 0,105 0,101 0,092 0,094 0,096 0,099 0,085 0,075 0,074 0,055 0,077 0,081 0,129 0,140 0,083 0,069 0,069 0,054 0,116 0,117 0,101 0,108 0,108 0,100 0,094 0,076 0,084 0,092 0,086 0,092 0,143 0,158 0,090 0,092 0,075 0,070 0,131 0,127 0,113 0,113 0,121 0,105 0,104 0,078 0,096 0,116 0,096 0,096 0,159 0,201 0,097 0,111 0,083 0,081 0,150 0,180 0,127 0,129 0,140 0,141 0,116 0,091 0,111 0,126 0,109 0,097 0,178 0,229 0,106 0,125 0,092 0,088 0,174 0,215 0,146 0,162 0,166 0,236 0,130 0,171 0,130 0,131 0,130 0,149 0,201 0,316 0,119 0,140 0,104 0,104 0,209 0,269 0,170 0,224 0,205 0,351 0,152 0,215 0,162 0,310 0,160 0,245 0,231 0,349 0,136 0,145 0,128 0,129 0,275 0,572 0,202 0,231 0,256 0,375 0,182 0,249 0,214 0,365 0,218 0,337 0,280 0,380 0,164 0,240 0,170 0,239 0,340 0,629 0,231 0,259 0,297 0,406 0,208 0,252 0,254 0,394 0,258 0,364 0,332 0,437 0,191 0,262 0,199 0,258 0,421 0,656 0,272 0,284 0,346 0,418 0,241 0,254 0,308 0,405 0,302 0,376 0,412 0,498 0,225 0,283 0,230 0,272
Hungary
Ireland
Italy
Latvia
Lithuania
Luxembourg
Malta
Netherlands
Poland
simulated historic simulated historic simulated historic simulated historic simulated historic simulated historic simulated historic simulated historic simulated historic 0,062 0,069 0,148 0,097 0,097 0,071 0,057 0,047 0,083 0,090 0,057 0,047 0,079 0,090 0,065 0,059 0,125 0,095 0,068 0,076 0,165 0,110 0,106 0,087 0,061 0,057 0,092 0,091 0,061 0,057 0,087 0,091 0,069 0,064 0,154 0,106 0,073 0,083 0,184 0,124 0,115 0,091 0,066 0,060 0,099 0,096 0,066 0,060 0,095 0,093 0,074 0,066 0,200 0,107 0,078 0,086 0,208 0,140 0,124 0,109 0,072 0,062 0,108 0,098 0,072 0,062 0,105 0,095 0,080 0,077 0,225 0,109 0,085 0,088 0,233 0,176 0,135 0,120 0,078 0,077 0,121 0,098 0,078 0,077 0,115 0,096 0,087 0,080 0,268 0,118 0,094 0,096 0,264 0,195 0,149 0,138 0,085 0,085 0,137 0,099 0,085 0,085 0,126 0,098 0,095 0,089 0,323 0,143 0,104 0,106 0,298 0,357 0,166 0,155 0,095 0,101 0,156 0,100 0,095 0,101 0,138 0,131 0,104 0,108 0,375 0,204 0,119 0,126 0,354 0,415 0,187 0,217 0,106 0,112 0,179 0,122 0,106 0,112 0,154 0,218 0,115 0,116 0,416 0,271 0,142 0,146 0,435 0,445 0,221 0,246 0,125 0,132 0,201 0,238 0,125 0,132 0,172 0,225 0,130 0,142 0,476 0,330 0,183 0,272 0,553 0,779 0,263 0,354 0,157 0,235 0,239 0,281 0,157 0,235 0,202 0,245 0,152 0,204 0,545 0,342 0,215 0,287 0,635 0,911 0,305 0,396 0,185 0,258 0,272 0,286 0,185 0,258 0,227 0,255 0,174 0,214 0,573 0,438 0,249 0,305 0,734 0,979 0,360 0,399 0,220 0,279 0,322 0,289 0,220 0,279 0,263 0,262 0,202 0,218 0,631 0,476
Portugal
Slovakia
Slovenia
Spain
Sweden
United Kingdom
Eurozone
simulated historic simulated historic simulated historic simulated historic simulated historic simulated historic simulated historic 50% 0,099 0,067 0,057 0,047 0,101 0,110 0,088 0,073 0,083 0,052 0,052 0,052 0,057 0,047 55% 0,110 0,069 0,061 0,057 0,113 0,127 0,096 0,082 0,090 0,062 0,057 0,057 0,061 0,057 60% 0,123 0,104 0,066 0,060 0,126 0,167 0,106 0,096 0,097 0,079 0,061 0,060 0,066 0,060 65% 0,139 0,116 0,072 0,062 0,142 0,194 0,116 0,112 0,105 0,109 0,067 0,062 0,072 0,062 70% 0,158 0,131 0,078 0,077 0,158 0,204 0,126 0,125 0,115 0,122 0,073 0,069 0,078 0,077 75% 0,179 0,165 0,085 0,085 0,176 0,208 0,138 0,135 0,125 0,129 0,081 0,072 0,085 0,085 80% 0,203 0,186 0,095 0,101 0,198 0,227 0,149 0,144 0,138 0,135 0,090 0,095 0,095 0,101 85% 0,233 0,259 0,106 0,112 0,223 0,320 0,166 0,199 0,157 0,158 0,104 0,117 0,106 0,112 90% 0,269 0,284 0,125 0,132 0,254 0,337 0,190 0,216 0,187 0,174 0,127 0,128 0,125 0,132 95% 0,326 0,372 0,157 0,235 0,304 0,346 0,223 0,250 0,236 0,317 0,180 0,291 0,157 0,235 97,5% 0,388 0,407 0,185 0,258 0,359 0,353 0,256 0,282 0,283 0,355 0,222 0,303 0,185 0,258 99% 0,452 0,425 0,220 0,279 0,416 0,367 0,294 0,289 0,332 0,359 0,258 0,326 0,220 0,279
T ABLE 14: COMPARISONS OF SOME QUANTILES OF THE MAXIMUM DRAWDOWN DISTRIBUTION CALCULATED FOR 500 SIMULATIONS ( LEFT COLUMN ) AND THE ORIGINAL HISTORIC VALUES ( RIGHT COLUMN ) OF VARIOUS COUNTRY PORTFOLIOS
29
The following graphs depict the “composition” of the portfolio indicator by presenting the time series of maximum drawdowns for the whole portfolio and each of the three asset classes, scaled by the quantiles of the simulated maximum drawdown distribution. The individual lines need to be interpreted independent from each other, in particular drawdown quantiles of the three asset classes do not add up to the portfolio drawdown quantile. By construction, maximum drawdowns are auto-correlated which results in these typical shapes of the graphs. 1,0 0,9 0,8
Quantiles
0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0
Gov_Eurozone
Corp_Eurozone
MSCI_EMU
Portf olio Eurozone
F IGURE 5: MAXIMUM DRAWDOWNS OF PORTFOLIOS AND INDIVIDUAL ASSET CLASSES CALCULATED OVER A PERIOD OF 1 YEAR FOR VARIOUS COUNTRIES
Annex 9 shows the same graphs but referred to each Member State Portfolio Austria
30
Period 1 Period 2 Period 3 Period 4
Switch on Switch off 95% 75% 18.09.2008 13.01.2010 29.12.2011 27.12.2012 -----
Portfolio Belgium Period 1 Period 2 Period 3 Period 4
Switch on Switch off 95% 75% 31.10.2008 30.10.2009 23.11.2011 21.11.2012 -----
Portfolio Bulgaria Period 1 Period 2 Period 3 Period 4
Switch on Switch off 75% 95% 31.10.2008 27.11.2009 -------
Portfolio Czech Republic Period 1 Period 2 Period 3 Period 4
Switch on Switch off 95% 75% 07.01.2009 21.01.2010 -------
Portfolio Denmark Period 1 Period 2 Period 3 Period 4
Switch on Switch off 95% 75% 06.10.2008 18.01.2010 -------
Portfolio Estonia
Switch on Switch off 75% 95% 10.10.2008 11.12.2009 -------
Period 1 Period 2 Period 3 Period 4
Portfolio Finland Switch on Switch off 95% 75% Period 1 09.02.2001 18.03.2003 Period 2 09.10.2008 30.10.2009 Period 3 --Period 4 ---
Portfolio France Period 1 Period 2 Period 3 Period 4
Switch on Switch off 95% 75% 31.07.2002 30.07.2003 10.10.2008 04.12.2009 -----
Portfolio Germany Period 1 Period 2 Period 3 Period 4
Switch on Switch off 95% 75% 15.10.2008 11.01.2010 -------
Portfolio Hungary Period 1 Period 2 Period 3 Period 4
Switch on Switch off 95% 75% 12.11.2008 04.01.2010 -------
Portfolio Ireland
Portfolio Italy
Period 1 Period 2 Period 3 Period 4
Switch on Switch off 95% 75% 21.04.2011 06.07.2012 -------
Switch on Switch off 95% 75% 26.01.2009 25.01.2010 01.11.2011 07.02.2013 -----
Portfolio Lithuania Period 1 Period 2 Period 3 Period 4
Switch on Switch off 95% 75% -- 31.05.2010 -------
Portfolio Malta
Portfolio Netherlands Period 1 Period 2 Period 3 Period 4
Switch on Switch off 95% 75% 25.03.2002 24.03.2003 08.12.2008 14.01.2010 -----
Period 1 Period 2 Period 3 Period 4
Switch on Switch off 95% 75% -- 06.07.2006 -------
Portfolio Portugal Period 1 Period 2 Period 3 Period 4
Switch on Switch off 95% 75% 04.02.2009 03.02.2010 06.07.2011 03.09.2012 -----
Portfolio Slovenia Period 1 Period 2 Period 3 Period 4
Switch on Switch off 95% 75% 03.12.2008 02.12.2009 -------
Period 1 Period 2 Period 3 Period 4
Period 1 Period 2 Period 3 Period 4 Portfolio Poland
Switch on Switch off 95% 75% -- 02.02.2010 -------
31
Portfolio Spain Period 1 Period 2 Period 3 Period 4
Switch on Switch off 95% 75% 26.01.2009 25.01.2010 18.06.2012 08.02.2013 -----
Portfolio Sweden Period 1 Period 2 Period 3 Period 4
Switch on Switch off 95% 75% 08.10.2008 10.11.2009 -------
Portfolio United Kingdom Period 1 Period 2 Period 3 Period 4
Switch on Switch off 95% 75% 10.10.2008 11.12.2009 -------
Portfolio Eurozone Period 1 Period 2 Period 3 Period 4
Switch on Switch off 95% 75% 15.10.2008 12.01.2010 -------
T ABLE 15: C RISIS PERIODS (= ACTIVATED ADAPTATION ) IN COUNTRY PORTFOLIOS ( ASSUMING THRESHOLDS AT THE 95% QUANTILE AND A DEACTIVATION AT THE 75% QUANTILE )
4.2. “X-out-of-Y” approach Alternatively, an “x out of y” approach could be used for combining indicators in different asset classes. With indicators for three asset classes (sovereign bonds, corporate bonds, equity) currently proposed, the first choice to test would be a “2 out of 3” approach, which means that an activation of the adaptation requires the mechanism to be activated in (at least) two asset classes. Consequently, a deactivation could be considered when only one or no indicator indicates a crisis period.
Based on the tables above, a combined mechanism of equity index drawdowns, government bond spreads and corporate bond spreads, the following periods with an activated adaptation can be identified for those countries for which all three indicators are currently available (again, assuming a 95%/75% activation/deactivation quantile combination):
-
-
32
-
Austria: o 08.12.2008 o 21.11.2011 Belgium: o 30.09.2008 Finland: o 04.12.2008 France: o 08.12.2008 o 23.11.2011 Germany: o 24.10.2008 Italy: o 22.01.2009
–17.05.2010 – 19.11.2012
-
– 04.08.2010 – 06.04.2010 – 04.06.2010 – 21.11.2012 – 14.07.2010 – 21.01.2010
o 12.09.2011 Netherlands: o 24.11.2008 Spain: o 26.01.2009 o 13.09.2011
– 21.01.2013 – 26.04.2010 – 25.01.2010 – ongoing
1999
AT
BE
DE
ES
FR
IT
NL
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
Gov bonds Corp bonds Equity Portfolio FX Volatility Interbank Rates Gov bonds Corp bonds Equity Portfolio FX Volatility Interbank Rates Gov bonds Corp bonds Equity Portfolio FX Volatility Interbank Rates Gov bonds Corp bonds Equity Portfolio FX Volatility Interbank Rates Gov bonds Corp bonds Equity Portfolio FX Volatility Interbank Rates Gov bonds Corp bonds Equity Portfolio FX Volatility Interbank Rates Gov bonds Corp bonds Equity Portfolio FX Volatility Interbank Rates
F IGURE 6: T IMING OF ADAPTATION ACTIVATIONS IN DIFFERENT ASSET CLASSES AND COUNTRIES ACTIVATION / DEACTIVATION )
(95%/75%
33
5.
Additional indicators
This chapter summarizes briefly the analyses performed with the additional indicators proposed in chapter 2.
5.1.
Supportive indicators
5.1.1. Volatility of foreign exchange rates The volatility of foreign exchange rates is a common indicator of financial markets stresses, especially when it depicts sudden portfolio shifts into or out of a country. Usually it would be necessary not to consider only one bilateral exchange rate per currency as the heightened volatility could also be caused by the reference currency; instead exchange rates versus e.g. EUR and USD or the nominal value of a currency against a basket of other currencies are more meaningful. Given the better availability of bilateral exchange rates, the former approach has been followed in the analysis.
FX BGN EUR Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
FX DKK EUR Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
FX LVL EUR
34
Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on Switch off 95% 75% 04.01.2001 01.10.2002 09.01.2003 20.05.2004 17.06.2004 01.07.2005 -------
FX CZK EUR
Switch on Switch off 95% 75% 01.03.1982 13.09.1983 21.08.1987 04.10.1988 29.09.1992 21.02.1994 07.04.1995 05.04.1996 -----
FX HUF EUR
Switch on Switch off 95% 75% 27.12.1999 20.03.2001 28.06.2002 24.07.2003 ---------
FX LTL EUR
Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on Switch off 95% 75% -- 31.03.2000 09.07.2008 10.09.2009 ---------
Switch on Switch off 95% 75% 28.06.2001 27.06.2002 07.07.2003 05.07.2004 17.10.2008 16.10.2009 05.10.2011 03.10.2012 -----
Switch on Switch off 95% 75% -- 30.04.2002 -----------
FX PLN EUR Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
FX SEK EUR Period 1 Period 2 Period 3 Period 4 Period 5 Period 6 Period 7 Period 8 Period 9
Switch on Switch off 95% 75% 08.08.2001 11.09.2002 24.10.2008 23.10.2009 ---------
FX RON EUR
Switch on Switch off 95% 75% 17.09.1981 16.09.1982 11.10.1982 10.10.1983 25.11.1992 24.11.1993 11.02.1994 10.02.1995 29.11.1995 27.11.1996 13.10.1998 12.10.1999 21.11.2008 10.11.2009 02.08.2012 ----
FX GBP EUR
Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Period 1 Period 2 Period 3 Period 4 Period 5 Period 6 Period 7 Period 8 Period 9
Switch on Switch off 95% 75% -- 31.03.2000 30.06.2000 29.06.2001 29.08.2001 28.08.2002 12.01.2009 11.01.2010 -----
Switch on Switch off 95% 75% 18.11.1980 12.01.1982 17.12.1982 16.12.1983 31.01.1986 30.01.1987 29.09.1992 28.09.1993 29.07.1997 28.07.1998 21.06.2000 20.06.2001 28.11.2008 02.12.2009 28.10.2010 27.10.2011 ---
T ABLE 16: C RISIS PERIODS BASED ON OBSERVED VOLATILITY OF FOREIGN EXCHANGE RATES AGAINST EUR ( ASSUMING THRESHOLDS AT THE 95% QUANTILE AND A DEACTIVATION AT THE 75% QUANTILE )
FX BGN USD Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
FX DKK USD Period 1 Period 2 Period 3 Period 4 Period 5 Period 6 Period 7 Period 8 Period 9 Period 10 Period 11
Switch on Switch off 95% 75% 11.01.2001 10.01.2002 10.06.2003 08.06.2004 12.09.2008 11.09.2009 01.06.2010 31.05.2011 -----
FX CZK USD
Switch on Switch off 95% 75% 20.04.1995 18.04.1996 20.10.1998 19.10.1999 08.01.2001 07.01.2002 18.07.2002 17.07.2003 11.09.2008 10.09.2009 28.05.2010 27.05.2011 -----------
FX EUR USD
Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Period 1 Period 2 Period 3 Period 4 Period 5 Period 6 Period 7 Period 8 Period 9 Period 10 Period 11
Switch on Switch off 95% 75% 02.11.1998 01.01.1999 12.07.2002 11.07.2003 04.09.2008 03.09.2009 04.06.2010 03.06.2011 -----
Switch on Switch off 95% 75% 07.04.1980 15.04.1981 27.05.1981 26.05.1982 19.07.1982 18.07.1983 21.05.1985 20.05.1986 18.12.1987 28.12.1988 27.03.1991 31.03.1992 26.10.1992 25.10.1993 08.01.2001 07.01.2002 09.06.2003 07.06.2004 12.09.2008 11.09.2009 31.05.2010 30.05.2011
35
FX HUF USD Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
FX LTL USD Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
FX RON USD Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
FX GBP USD Period 1 Period 2 Period 3 Period 4 Period 5 Period 6 Period 7 Period 8 Period 9 Period 10 Period 11
Switch on Switch off 95% 75% 16.09.2008 15.09.2009 24.05.2010 26.05.2011 03.10.2011 01.10.2012 -------
FX LVL USD
Switch on Switch off 95% 75% 17.07.2002 18.07.2003 11.09.2008 10.09.2009 28.05.2010 27.05.2011 -------
FX PLN USD
Switch on Switch off 95% 75% 06.01.1999 19.01.2000 21.12.2004 20.12.2005 02.10.2008 01.10.2009 11.06.2010 10.06.2011 -----
FX SEK USD
Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on Switch off 95% 75% 09.09.2008 08.09.2009 25.05.2010 25.05.2011 ---------
Switch on Switch off 95% 75% 10.09.2008 06.10.2009 21.05.2010 20.05.2011 23.09.2011 21.09.2012 -------
Switch on Switch off 95% 75% 14.03.1997 13.03.1998 11.07.2002 15.07.2003 15.09.2008 19.09.2009 04.06.2010 03.06.2011 -----
Switch on Switch off 95% 75% 09.06.1981 08.06.1982 25.01.1985 24.01.1986 11.12.1987 04.01.1989 20.06.1989 19.06.1990 17.08.1990 19.08.1991 25.09.1992 24.09.1993 10.09.2008 09.09.2009 ---------
T ABLE 17: C RISIS PERIODS BASED ON OBSERVED VOLATILITY OF FOREIGN EXCHANGE RATES AGAINST USD ( ASSUMING THRESHOLDS AT THE 95% QUANTILE AND A DEACTIVATION AT THE 75% QUANTILE )
36
5.1.2. Volatility of short-term interbank rates In a similar manner as the volatility of foreign exchange rates also the volatility of shortterm interbank rates could be used as an indicator for a financial crisis since they provide information on the perceived riskiness in the respective banking sector. For this historic
analysis, 3-months interbank rates have been used – their use is also recommended for the future implementation.
3 months interbank rate EUR Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on Switch off 75% 95% 18.12.2008 17.12.2009 -03.04.2012 ---------
3 months interbank rate BGN Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on Switch off 75% 95% -- 03.09.2004 10.08.2012 ----------
3 months interbank rate CZK Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on Switch off 95% 75% 21.04.1993 20.04.1994 21.05.1997 20.05.1998 29.12.1998 28.12.1999 23.01.2009 17.02.2010 -05.10.2012 ---
3 months interbank rate GBP Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on Switch off 95% 75% 26.07.1984 25.07.1985 20.07.1988 19.07.1989 23.10.1992 22.10.1993 17.11.2008 06.01.2010 -24.08.2012 ---
3 months interbank rate HUF Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on Switch off 95% 75% 03.03.2000 02.03.2001 17.01.2003 24.02.2004 30.10.2008 02.02.2010 -------
3 months interbank rate LTL Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on Switch off 95% 75% 11.10.1999 09.10.2000 12.01.2010 11.01.2011 ---------
3 months interbank rate LVL Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on Switch off 95% 75% 29.03.2007 10.04.2008 24.06.2009 23.06.2010 02.09.2010 01.09.2011 30.11.2011 28.11.2012 -----
3 months interbank rate PLN Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on Switch off 95% 75% 01.02.1999 22.03.2000 20.12.2001 09.01.2003 27.07.2005 26.07.2006 19.01.2009 18.01.2010 04.01.2013 ----
3 months interbank rate RON Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on Switch off 95% 75% 24.02.1997 16.03.1998 02.06.1998 01.06.1999 20.10.2008 19.10.2009 -------
3 months interbank rate SEK Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Switch on Switch off 95% 75% 08.12.2008 09.12.2009 01.07.2010 30.06.2011 ---------
T ABLE 18: C RISIS PERIODS BASED ON OBSERVED VOLATILITY OF INTERBANK RATES (ASSUMING THRESHOLDS 95% QUANTILE AND A DEACTIVATION AT THE 75% QUANTILE )
AT THE
37
5.2.
Complementary indicators
As complementary indices it is suggested assessing the development of composite indices of systemic stress, like e.g. the CISS index of the European Central Bank or the Financial Stress Index of the Federal Reserve Bank of St. Louis. For further details on these indices, refer to Annex 3.
5.3.
Real-economy indicators
Financial market indicators tend to be rather volatile and, while sometimes over-shooting in a crisis situation, might recover very quickly and well before the general economic environment (including consumer and business confidence) has stabilized. Therefore it is deemed beneficial to include real-economy indicators in the analysis in order to interpret financial market time series correctly. The set of real-economy indicators to be considered has been set up based on the following considerations: •
Indicators should be rather frequent (e.g. available on a monthly basis) and available shortly after the reference period.
•
Indicators should be (tail-)correlated with financial market time series and, ideally, be a driving factor for causing a sharp, steep and unforeseen fall in financial markets.
•
Indicators should be available on a harmonized basis for a large number of currencies and countries.
Based on these considerations, the following indicators have been selected: •
Annual change of industrial production (source: Eurostat);
•
Consumer confidence (source: European Commission);
•
Annual change of consumer prices (source: Eurostat).
Macroeconomic theory suggests that it is unlikely that all indicators will, at the same time, indicate a crisis situation, as these situations might differ in nature. As the most relevant real-economy crisis situations, resulting in sharp declines in financial markets, demand shocks and inflation shocks have been identified. As opposed to these exogenous shocks, a crisis might emerge endogenously in the financial market – these situations would be covered by the indicators set out in chapters 5.1 and 5.2.
38
For the case of Austria, the graph below shows the interplay between the maximum drawdowns of the portfolio (as described in chapter 4.1) and the consumer confidence indicator (quantiles of the distribution of observed values). While the overall correlation is not particularly high, the correlation in times of stressed financial markets (i.e. 2008 and 2011) is much higher. In the activation event the consumer confidence slightly lags the portfolio drawdown, while in the deactivation moment the consumer confidence has already improved so the deactivation signal of the portfolio approach is confirmed.
1,0 0,9 0,8 0,7 0,6 0,5
Portfolio
0,4
Consumer confidence inverted
0,3 0,2 0,1 0,0
F IGURE 7: M AXIMUM DRAWDOWNS OF PORTFOLIOS AND CONSUMER CONFIDENCE (A USTRIA )
6.
Interplay between the activation/deactivation adaptation and its size calibration
6.1.
of
the
General considerations
•
When deriving the indicator set for the CCP activation mechanism, one should aim for a strong correlation between the indicators used and the methodology for deriving the actual size of the adaptation. This is taken into account by deriving the indicators from data which is as close as possible to the data which is used for calculating the size of the adaptation – i.e. the focus is on those asset classes which make up the biggest part of the representative portfolio.
•
With regard to the deactivation, the interplay is less obvious: In theory, a deactivation decision could be warranted while the calculation of the actual size of the adaptation still yields a positive number.
6.2.
Deactivation of the CCP
The following chapter includes a summary of issues to consider when determining 1) a deactivation of the CCP, 2) a floating CCP over time.
39
1) Deactivation of CCP in a Floating Phase In the discussion on the CCP activation mechanism the following phases have been identified (see Figure 8): 1. Monitoring Phase 2. Activation / CCP Active Phase a) „stable“ CCP b) „floating“ CCP 3. Deactivation / CCP Deactive Phase
F IGURE 8: CCP ACTIVATION /DEACTIVATION CYCLE 5
The term “stable” CCP of phase 2a is somewhat misleading. More specifically phase 2a currently describes the premise that after activation the CCP stays active for at least 12 months. Within this time span the CCP may be adjusted upwards (i.e. premium coverage can be increased), but not downwards (i.e. premium coverage can be decreased). After the stable 12-months CCP period a deactivation decision needs to be taken. Given the suggested approach to use two different thresholds for respectively activating and deactivating various activation and deactivation scenarios can be identified.
40 5
Note that this graph (as well as the following one) shows the actual time series (e.g. spread or equity index) as opposed to the maximum drawdown. Therefore the vertical scale appears to be inverted when comparing to most other graphs in this report.
Figure 9 illustrates such a deactivation scenario. Basically the discussion has shown that Figure 9 is an illustration of a possible floating stage scenario. A premise for this particular floating CCP scenario would be the requirement that a mechanistic deactivation of the CCP requires a market stabilization of a certain amount (i.e. reflected through a decreasing CCP coverage under the deactivation threshold as determined by the activation mechanism). Given this premise, Figure 9 illustrates a floating CCP between the end of the stable CCP period (marked with “1”) and the mechanistically determined deactivation moment (marked with “2”). For clarification it should be mentioned that it is the current working assumption that the implementation of mechanistically determined activation moments are always subject to expert judgment (i.e. activation and deactivation of CCP will never be purely mechanistically determined).
60 50 40
2
1
30
A 20 10 1 year
0 Deactivation quantile
Activation quantile
F IGURE 9: D EACTIVATION OF THE CCP
Such a scenario approach activation/deactivation issues.
could
be
used
to
investigate
other
possible
For instance, the line movement in Figure 9 (i.e. illustrating a negative index development with an activation of the adaptation at point “A”) could also cross the activation threshold after the 12-month stable period. In the discussions it has been shown that in such a situation a continuation of a floating CCP would be favorable (i.e. passing the activation threshold does not trigger a new 12-months stable period). Also, Figure 9 does not account for the time span between moment “1” and “2”. This leads to another issue worth discussing and possibly investigating further:
41
2) Deactivation with a floating CCP over time With a continuing floating CCP one needs to consider the aspect of time when determining a suitable deactivation moment. As the CCP should give institutions the possibility to avoid fire sales in case of sharp, steep and unforeseen market movements, it is questionable if it still fulfills this purpose over an “unlimited” (i.e. undefined) time span (e.g. market prices/values of assets stay low for years). Already in the 12-months stable CCP phase it can be expected that institutions marginally and constantly adapt their asset allocation to the new market situation. The CCP should just buy institutions time to avoid drastic portfolio changes with possible vicious market effects. Nevertheless, institutions need to prepare for a deactivation of the CCP, which will always be out of their control and therefore unforeseeable. Somehow the deactivation decision needs to account for this time aspect, which becomes very substantial in a floating CCP phase.
7.
42
Conclusion •
If finally it is decided to set out the framework tested for the adaptation in LTGA exercise, the following recommendations would be applicable:
•
1. A broad range of indicators should be used to inform any decision on the activation or deactivation of the adaptation. This set of indicators includes primary indicators (government bond spreads, corporate bond spreads and equity indices) as well supportive indicators.
•
2. Maximum drawdowns are a suitable measure for determining a situation in which financial markets experience a sharp, steep and unforeseen fall.
•
3. The maximum drawdown should be calculated for a portfolio instead of calculating it for each asset class independently.
•
4. Activation should be based on a quantile threshold which should be relatively high (to meet the “unforeseen” criterion).
•
5. Deactivation should also be based on a quantile threshold which should be lower than the activation quantile, i.e. situation on financial market should have normalized.
•
6. The current set of indicators have not been designed as part of a purely mechanistic approach to trigger the adaptation. Instead they have been developed in line with the current framework, hence under the assumption that it should be supervisory discretionary decision in order to prevent insurance undertakings from pricing-in the mitigating effect of the adaptation.
ANNEX 1: COMPOSITE SYSTEMIC STRESS INDICATORS •
Various composite indices of systemic stress produced by public authorities could be used to double-check the results derived with the main set of primary indices. Amongst those, this annex outlines the CISS index by the ECB and the Financial Stress Index by the Federal Reserve Bank of St. Louis.
ECB CISS Index •
Composite index on a weekly basis, comprised of:
•
Realised volatility of the 3-month Euribor rate
•
Interest rate spread between 3-month Euribor and 3-month French T-bills
•
Monetary Financial Institution’s (MFI) emergency lending at Eurosystem central banks: MFI’s recourse to the marginal lending facility, divided by their total reserve requirements
•
Realised volatility of the German 10-year benchmark government bond index
•
Yield spread between A-rated non-financial corporations and government bonds (7-year maturity bracket)
•
10-year interest rate swap spread: weekly average of daily data
•
Realised volatility of the DataStream non-financial sector stock market index
•
CMAX for the DataStream non-financial sector stock market index: maximum cumulated index losses over a moving 2-year window
•
Stock-bond correlation
•
Realised volatility of the idiosyncratic equity return of the DataStream bank sector stock market index over the total market index
•
Yield spread between A-rated financial and non-financial corporations (7-year maturity)
•
CMAX as defined above interacted with the inverse price-book ratio (book-price ratio) for the financial sector equity market index
•
Realised volatility of the euro exchange rate vis-à-vis the US dollar, the Japanese Yen and the British Pound, respectively
•
For further details see: http://www.ecb.int/pub/pdf/scpwps/ecbwp1426.pdf
43
• F IGURE 10: ECB CISS I NDEX (08.01.1999 – 22.02.2013), TAKEN FROM THE ESRB R ISK D ASHBOARD M ARCH 2013
St. Louis Fed’s Financial Stress Index • • • • • • • • • • • • • • • • • •
44
• •
Composite index on a weekly basis, comprised of: Effective federal funds rate 2-year Treasury 10-year Treasury 30-year Treasury Baa-rated corporate Merrill Lynch High-Yield Corporate Master II Index Merrill Lynch Asset-Backed Master BBB-rated Yield curve: 10-year Treasury minus 3-month Treasury Corporate Baa-rated bond minus 10-year Treasury Merrill Lynch High-Yield Corporate Master II Index minus 10-year Treasury 3-month London Interbank Offering Rate-Overnight Index Swap (LIBOR-OIS) spread 3-month Treasury-Eurodollar (TED) spread 3-month commercial paper minus 3-month Treasury bill J.P. Morgan Emerging Markets Bond Index Plus Chicago Board Options Exchange Market Volatility Index (VIX) Merrill Lynch Bond Market Volatility Index (1-month) 10-year nominal Treasury yield minus 10-year Treasury Inflation Protected Security yield (breakeven inflation rate) Vanguard Financials Exchange-Traded Fund (equities) For further details see: http://research.stlouisfed.org/publications/net/NETJan2010Appendix.pdf
7,0 6,0 5,0 4,0 3,0 2,0 1,0 0,0 -1,0 -2,0
F IGURE 11: S T . L OUIS F ED ’ S F INANCIAL S TRESS I NDEX (31.12.1993 – 08.02.2013)
45
ANNEX 2: DIFFERENCES BETWEEN PRICE AND RETURN INDICES
46
•
It has been discussed whether the drawdown calculations for equity indices should be based on price indices or return indices. While the latter would better reflect economic reality by including both capital gains and recurring dividends, price indices are usually the more widely used indices. Out of the major indices in Europe, only the “Deutscher Aktien-Index” (DAX) employs a return concept. For that reason, one might argue that the development of price indices is the more relevant one as portfolio managers are focusing on these benchmarks and would potentially be inclined to start a fire sale in case the price indices depict a sharp and steep fall. From a statistical point of view, the use of price indices is seen as advantageous as indices are more widely available (and even existing for countries where return indices are more common) and have a longer history.
•
This annex presents a quick comparison of price and return indices for two exemplary countries for which the respective MSCI country index has been chosen – the time series of the price and return index have been extracted from DataStream.
•
It can be concluded that drawdowns of return indices are generally lower – this holds true for both observed and simulated drawdowns. The activation/deactivation dates as such do not vary largely.
•
Estonia
MSCI Estonia Price index (02.06.2003 - 08.02.2013)
MSCI Estonia Return index (02.06.2003 - 08.02.2013)
Quantiles 50% 55% 60% 65% 70% 75% 80% 85% 90% 95% 97,5% 99%
Quantiles 50% 55% 60% 65% 70% 75% 80% 85% 90% 95% 97,5% 99%
Period 1 Period 2 Period 3
historic simulated 0,21129 0,23330 0,23622 0,24836 0,25616 0,26548 0,28351 0,28679 0,29813 0,31296 0,35536 0,34187 0,42661 0,37984 0,55158 0,43028 0,78479 0,50346 0,92861 0,61564 1,00324 0,71549 1,09212 0,83926 Switch on Switch off 95% 75% 05.02.2008 22.10.2009 -----
Period 1 Period 2 Period 3
historic simulated 0,19479 0,21748 0,20826 0,23330 0,23562 0,25094 0,26263 0,27153 0,28886 0,29562 0,34513 0,32344 0,41629 0,36014 0,54246 0,41033 0,76729 0,48130 0,89436 0,58978 0,96908 0,68820 1,05761 0,82174 Switch on Switch off 95% 75% 29.01.2008 23.10.2009 -----
T ABLE 19: D RAWDOWN DISTRIBUTION AND ACTIVATION PERIODS FOR THE MSCI E STONIA (BOTH PRICE AND RETURN INDEX )
•
Slovenia
MSCI Slovenia Price index (02.06.2003 - 08.02.2013)
MSCI Slovenia Return index (02.06.2003 - 08.02.2013)
Quantiles 50% 55% 60% 65% 70% 75% 80% 85% 90% 95% 97,5% 99%
Quantiles 50% 55% 60% 65% 70% 75% 80% 85% 90% 95% 97,5% 99%
Period 1 Period 2 Period 3
historic simulated 0,17312 0,18212 0,18828 0,19585 0,19919 0,21048 0,20839 0,23211 0,23223 0,25843 0,26163 0,29194 0,28924 0,32864 0,38706 0,37110 0,60113 0,43285 0,79873 0,54152 0,93222 0,62891 0,98754 0,74030 Switch on Switch off 95% 75% 17.09.2008 18.11.2009 -----
Period 1 Period 2 Period 3
historic simulated 0,16879 0,17006 0,18014 0,18232 0,18899 0,19566 0,20051 0,21278 0,21312 0,23791 0,23526 0,26900 0,26249 0,30894 0,38484 0,35507 0,58929 0,41749 0,78557 0,51640 0,91869 0,61139 0,97349 0,71909 Switch on Switch off 95% 75% 15.09.2008 18.11.2009 -----
T ABLE 20: D RAWDOWN DISTRIBUTION AND ACTIVATION PERIODS FOR THE MSCI S LOVENIA ( BOTH PRICE AND RETURN INDEX )
47
ANNEX 3: SENSITIVITY OF RESULTS TO CHANGES IN THE ASSUMED AVERAGE DURATION OF BOND HOLDINGS
48
•
For the portfolio approach, it was necessary to transform the time series of spreads into an index which is sensitive only to the spread changes. For this purpose, spreads have been multiplied with the average duration collected from NSAs, namely 8.8 years for government bonds and 6.6 years for corporate bonds. However, it is acknowledged that durations can vary substantially across countries and over time.
•
This annex presents a sensitivity analysis by showing the differences in the drawdown distributions and the activation/deactivation dates when assuming the modified duration to be 4, 5 and 6 years, respectively.
•
The results show a quite heterogeneous pattern with regard to the duration assumption. It cannot be said that in general a longer duration results in more or less activation moments or that the activation periods are either longer or shorter. In some 90% of the analysed trigger moments we see, however, a coincidence in timing in a sense that the adaptation is activated within a two-weeks period for all assumed bond durations (4, 5 and 6 years).
Duration 4 years
Duration 6 years
Portfolio Austria Period 1 Period 2 Period 3
Switch on Switch off 95% 75% 18.09.2008 19.11.2009 -----
Portfolio Austria Period 1 Period 2 Period 3
Switch on Switch off 95% 75% 18.09.2008 20.11.2009 16.02.2012 08.02.2013 ---
Portfolio Austria Period 1 Period 2 Period 3
Switch on Switch off 95% 75% 18.09.2008 08.01.2010 07.02.2012 05.02.2013 ---
Portfolio Belgium Period 1 Period 2 Period 3
Switch on Switch off 95% 75% 04.04.2003 02.04.2004 29.09.2008 14.10.2009 ---
Portfolio Belgium Period 1 Period 2 Period 3
Switch on Switch off 95% 75% 15.04.2003 13.04.2004 29.09.2008 14.10.2009 ---
Portfolio Belgium Period 1 Period 2 Period 3
Switch on Switch off 95% 75% 03.04.2003 01.04.2004 29.09.2008 14.10.2009 ---
Portfolio Germany Period 1 Period 2 Period 3
Switch on Switch off 95% 75% 10.10.2008 14.10.2009 -----
Portfolio Germany Period 1 Period 2 Period 3
Switch on Switch off 95% 75% 10.10.2008 15.10.2009 -----
Portfolio Germany Period 1 Period 2 Period 3
Switch on Switch off 95% 75% 10.10.2008 23.10.2009 -----
Portfolio Spain Period 1 Period 2 Period 3
Switch on Switch off 95% 75% 10.10.2008 23.10.2009 -----
Portfolio Spain Period 1 Period 2 Period 3
Switch on Switch off 95% 75% 10.10.2008 29.10.2009 30.05.2012 ----
Portfolio Spain Period 1 Period 2 Period 3
Switch on Switch off 95% 75% 10.10.2008 23.10.2009 30.05.2012 ----
Portfolio Finland Period 1 Period 2 Period 3
Switch on Switch off 95% 75% -- 18.04.2003 22.10.2008 21.10.2009 ---
Portfolio Finland Period 1 Period 2 Period 3
Switch on Switch off 95% 75% -- 18.04.2003 10.10.2008 09.10.2009 ---
Portfolio Finland Period 1 Period 2 Period 3
Switch on Switch off 95% 75% -- 17.04.2003 10.10.2008 09.10.2009 ---
Portfolio France Period 1 Period 2 Period 3
Switch on Switch off 95% 75% 05.12.2002 04.12.2003 09.10.2008 09.10.2009 ---
Portfolio France Period 1 Period 2 Period 3
Switch on Switch off 95% 75% 05.12.2002 04.12.2003 09.10.2008 26.10.2009 ---
Portfolio France Period 1 Period 2 Period 3
Switch on Switch off 95% 75% 27.02.2003 26.02.2004 09.10.2008 26.10.2009 ---
Portfolio Italy
•
Duration 5 years
Period 1 Period 2 Period 3
Switch on Switch off 95% 75% 08.10.2008 12.01.2010 -----
Portfolio Netherlands Period 1 Period 2 Period 3
Switch on Switch off 95% 75% 25.02.2003 24.02.2004 08.10.2008 09.10.2009 ---
T ABLE 21: DURATIONS
Portfolio Italy Period 1 Period 2 Period 3
Switch on Switch off 95% 75% 08.10.2008 12.01.2010 01.02.2012 30.01.2013 ---
Portfolio Netherlands Period 1 Period 2 Period 3
Switch on Switch off 95% 75% 25.02.2003 24.02.2004 08.10.2008 09.10.2009 ---
Portfolio Italy Period 1 Period 2 Period 3
Switch on Switch off 95% 75% 09.10.2008 12.01.2010 18.01.2012 16.01.2013 ---
Portfolio Netherlands Period 1 Period 2 Period 3
Switch on Switch off 95% 75% 26.02.2003 25.02.2004 08.10.2008 15.10.2009 ---
A CTIVATION EVENTS IN THE PORTFOLIO APPROACH UNDER VARIOUS ASSUMPTIONS FOR BOND (4, 5 AND 6 YEARS )
49
ANNEX 4: CORRELATION ANALYSIS
•
A preliminary analysis on correlations of indicators has been performed with rolling windows of varying lengths. It turned out that correlations are quite unstable both within asset classes across countries and across asset classes within one country.
•
This gives rise to some conclusions:
1. It will be difficult to fill up missing or short time series with proxy data for the whole EU or Eurozone, therefore this should be done on an exceptional basis only. At least it should be ensured that any decision on activating the adaptation by EIOPA is taken on an informed basis so that the exceptional use of proxies is fully understood and evaluated based on the actual circumstances. 2. It will be difficult to use breaks in the correlation structure (between asset classes or countries) as a crisis indicator per se. Therefore the indicators presented in chapter 2 should be analysed in isolation per currency and per country.
50
ANNEX 5: ACTIVATION ANALYSIS FOR DIFFERENT ACTIVATING/DEACTIVATING QUANTILES OF SIMULATED MAXIMUM DRAWDOWN DISTRIBUTION (MSCI COUNTRY INDICES, PRICE INDEX), WHEN THE ADAPTATION IS ASSUMED TO BE ACTIVE FOR AT LEAST A YEAR, I.E. 261 TR MSCI Austria Start 31.12.1980 activation quantile Nb trigger moments /deactivation all data last 10 Y last 5Y quantile 70/50 70/70 75/55 80/60 85/65 90/70 95/75
8 11 7 6 5 4 3
3 4 3 3 2 2 2
MSCI Belgium Start 31.12.1980 activation quantile Nb trigger moments all data last 10 Y last 5Y /deactivation 70/50 7 3 70/70 7 3 75/55 7 3 80/60 7 3 85/65 5 2 90/70 4 2 95/75 2 2
MSCI Bulgaria Start 31.05.2006 activation quantile Nb trigger moments /deactivation all data last 10 Y last 5Y 70/50 2 0 70/70 3 0 75/55 2 0 80/60 2 0 85/65 2 0 90/70 1 0 95/75 1 0
MSCI Czech Republic Start 01.01.1996 activation quantile Nb trigger moments all data last 10 Y last 5Y /deactivation 70/50 5 2 70/70 7 3 75/55 5 2 80/60 4 1 85/65 4 1 90/70 3 1 95/75 2 1
MSCI Denmark Start 31.12.1980 activation quantile Nb trigger moments /deactivation all data last 10 Y last 5Y 70/50 10 3 70/70 10 3 75/55 10 3 80/60 8 3 85/65 7 3 6 3 90/70 95/75 4 2
MSCI Estonia Start 2.6.2003 activation quantile Nb trigger moments /deactivation all data last 10 Y last 5Y 70/50 3 0 70/70 4 0 75/55 2 0 80/60 2 0 85/65 2 0 90/70 1 0 95/75 1 0
% active Trigger time all data 2 3 2 2 2 2 2
40 41 33 25 19 15 11
all data 2 2 2 2 1 1 1
all data 2 3 2 2 2 1 1
all data 1 2 1 1 1 1 1
all data 2 2 2 2 2 2 1
all data 2 3 2 2 2 1 1
last 10 Y
mean active trigger (d)
last 5Y 56 55 44 42 30 28 23
% active Trigger time last 10 Y last 5Y 36 43 34 40 33 43 32 43 23 29 17 28 9 24
% active Trigger time last 10 Y last 5Y 40 0 37 0 35 0 32 0 29 0 17 0 16 0
% active Trigger time last 10 Y last 5Y 38 24 48 31 34 23 27 11 27 11 20 10 12 10
% active Trigger time last 10 Y last 5Y 42 41 40 41 41 41 32 38 26 34 23 34 16 23
% active Trigger time last 10 Y last 5Y 41 0 51 0 31 0 30 0 29 0 18 0 18 0
all data 89 88 66 63 59 55 46
last 10 Y 423 313 396 346 313 311 307
median active trigger (d)
last 5Y 483 356 380 365 392 361 302
60 55 60 60 35 35 28
mean active trigger (d) all data last 10 Y last 5Y 430 373 408 349 401 370 381 369 379 371 353 368 360 313
54 50 47 44 39 22 22
mean active trigger (d) all data last 10 Y last 5Y 351 0 217 0 308 0 283 0 255 0 292 0 287 0
27 42 26 21 21 20 20
mean active trigger (d) all data last 10 Y last 5Y 341 306 305 269 308 298 305 275 297 274 296 266 261 261
61 60 61 55 49 48 26
mean active trigger (d) all data last 10 Y last 5Y 349 358 337 355 343 356 333 327 317 298 322 295 345 301
55 74 55 55 54 34 26
mean active trigger (d) all data last 10 Y last 5Y 348 0 322 0 389 0 384 0 369 0 459 0 447 0
all data 582 380 428 407 384 361 302
last 10 Y 330 261 335 287 261 262 318
last 5Y 544 392 335 312 392 361 302
582 498 428 407 384 361 302
392 358 392 391 459 455 363
median active trigger (d) all data last 10 Y last 5Y 357 268 307 262 353 264 261 263 261 371 339 368 360 313
392 358 392 391 459 455 363
351 217 308 283 255 292 287
median active trigger (d) all data last 10 Y last 5Y 351 0 261 0 308 0 283 0 255 0 292 0 287 0
351 261 308 283 255 292 287
351 273 334 275 274 266 261
median active trigger (d) all data last 10 Y last 5Y 351 306 262 261 262 298 269 275 268 274 266 266 261 261
351 273 334 275 274 266 261
396 393 395 360 317 312 340
median active trigger (d) all data last 10 Y last 5Y 262 262 261 262 261 262 263 262 261 262 261 262 301 301
396 393 395 360 317 312 340
361 321 361 357 352 441 344
median active trigger (d) all data last 10 Y last 5Y 261 0 261 0 389 0 384 0 369 0 459 0 447 0
361 261 361 357 352 441 344
51
MSCI Finland Start 31.12.1982 Nb trigger moments activation quantile /deactivation all data last 10 Y last 5Y 70/50 8 4 70/70 9 4 75/55 8 4 80/60 8 4 85/65 8 4 90/70 4 2 95/75 3 2
MSCI France Start 31.12.1980 activation quantile Nb trigger moments /deactivation all data last 10 Y last 5Y 70/50 8 3 70/70 8 3 75/55 8 3 80/60 7 3 85/65 7 3 90/70 6 3 95/75 4 2
MSCI Germany Start 31.12.1980 activation quantile Nb trigger moments all data last 10 Y last 5Y /deactivation 70/50 7 3 70/70 9 4 75/55 6 3 80/60 6 3 85/65 6 3 90/70 5 3 95/75 3 2
MSCI Greece Start 30.12.1988 activation quantile Nb trigger moments all data last 10 Y last 5Y /deactivation 70/50 7 3 70/70 8 4 75/55 7 3 80/60 7 4 85/65 6 3 90/70 5 3 95/75 3 2
MSCI Hungary Start 01.01.1996 Nb trigger moments activation quantile last 5Y /deactivation all data last 10 Y 70/50 6 4 6 4 70/70 75/55 4 2 2 80/60 4 85/65 4 2 90/70 4 2 2 1 95/75
MSCI Ireland Start 30.12.1988 activation quantile Nb trigger moments last 10 Y last 5Y /deactivation all data 70/50 6 3 7 4 70/70 75/55 6 3 6 3 80/60 85/65 6 3 3 1 90/70 95/75 1 1
MSCI Italy Start 31.12.1980 activation quantile Nb trigger moments all data last 10 Y last 5Y /deactivation 70/50 9 3 70/70 12 4 75/55 9 3 80/60 8 3 85/65 8 3 90/70 7 3 95/75 5 2
52
all data 2 2 2 2 2 1 1
all data 2 2 2 2 2 2 1
all data 2 3 2 2 2 2 1
all data 2 3 2 3 2 3 2
all data 2 2 2 2 2 2 1
all data 2 3 2 2 2 1 1
all data 2 3 2 2 2 2 2
% active Trigger time last 10 Y last 5Y 40 47 41 46 39 47 38 46 35 44 19 23 15 23
% active Trigger time last 10 Y last 5Y 35 41 34 40 35 41 30 39 29 36 25 35 17 21
% active Trigger time last 10 Y last 5Y 34 39 39 48 29 36 29 36 27 34 21 31 14 21
% active Trigger time last 10 Y last 5Y 43 53 45 56 42 52 36 48 34 49 23 36 19 35
% active Trigger time last 10 Y last 5Y 49 44 42 48 31 27 29 26 25 28 26 23 12 11
% active Trigger time last 5Y last 10 Y 34 43 55 38 43 33 31 43 42 30 17 20 7 17
% active Trigger time last 10 Y last 5Y 41 46 41 45 39 45 34 44 31 39 24 36 17 24
54 52 53 52 48 27 26
mean active trigger (d) all data last 10 Y last 5Y 390 306 355 300 380 304 371 301 340 286 378 304 405 300
60 58 60 57 51 50 23
mean active trigger (d) all data last 10 Y last 5Y 365 355 355 347 361 353 363 335 349 308 342 304 357 278
58 76 53 52 48 41 23
mean active trigger (d) all data last 10 Y last 5Y 405 338 366 311 412 316 399 312 382 293 360 266 390 278
86 92 84 75 77 71 70
mean active trigger (d) all data last 10 Y last 5Y 384 459 351 365 376 450 319 309 359 421 290 310 389 454
58 55 53 52 50 46 22
mean active trigger (d) last 10 Y last 5Y all data 329 318 311 311 347 343 319 336 309 327 297 286 291 276
60 85 60 60 60 39 33
mean active trigger (d) all data last 10 Y last 5Y 357 372 345 359 372 346 322 370 368 315 518 347 430 430
71 70 69 69 57 52 48
mean active trigger (d) all data last 10 Y last 5Y 378 396 287 294 363 388 354 386 323 334 288 313 280 309
351 338 347 341 311 346 337
median active trigger (d) all data last 10 Y last 5Y 322 263 267 262 313 262 301 262 261 262 304 304 337 300
351 338 347 341 311 346 337
393 380 390 372 331 325 293
median active trigger (d) all data last 10 Y last 5Y 263 263 261 262 262 262 261 262 261 262 261 262 277 278
393 380 390 372 331 325 293
377 327 343 337 309 269 293
median active trigger (d) all data last 10 Y last 5Y 268 268 261 262 343 262 337 262 309 262 261 262 293 278
377 261 343 337 309 269 293
557 399 545 325 501 310 454
median active trigger (d) all data last 10 Y last 5Y 308 385 335 366 283 362 271 283 275 289 261 273 266 454
557 388 545 305 501 273 454
375 360 347 336 327 297 291
median active trigger (d) all data last 10 Y last 5Y 265 277 274 261 347 339 336 302 291 327 275 297 291 276
375 360 347 336 327 297 291
392 367 392 392 391 502 430
median active trigger (d) all data last 10 Y last 5Y 262 261 261 301 262 261 262 261 261 261 261 518 430 430
392 339 392 392 391 502 430
463 305 451 448 370 340 309
median active trigger (d) all data last 10 Y last 5Y 393 397 261 263 380 396 320 393 271 280 261 261 261 309
463 263 451 448 370 340 309
MSCI Lithuania Start 01.06.2009 activation quantile Nb trigger moments all data last 10 Y last 5Y /deactivation 70/50 2 1 70/70 2 0 75/55 1 0 80/60 1 0 85/65 1 0 90/70 1 0 95/75 1 0
MSCI Netherlands Start 31.12.1980 activation quantile Nb trigger moments all data last 10 Y last 5Y /deactivation 70/50 7 3 70/70 7 3 75/55 6 3 80/60 6 3 85/65 5 3 90/70 4 2 95/75 3 2
MSCI Poland Start 31.12.1993 activation quantile Nb trigger moments all data last 10 Y last 5Y /deactivation 70/50 5 1 70/70 7 2 75/55 4 1 80/60 4 1 85/65 4 1 90/70 4 1 95/75 2 1
MSCI Portugal Start 30.12.1988 activation quantile Nb trigger moments /deactivation all data last 10 Y last 5Y 70/50 7 3 70/70 10 4 75/55 9 3 80/60 9 3 85/65 7 3 90/70 6 3 95/75 5 3
MSCI Romania Start 30.11.2006 activation quantile Nb trigger moments /deactivation all data last 10 Y last 5Y 70/50 1 0 70/70 1 0 75/55 1 0 80/60 1 0 85/65 1 0 90/70 1 0 95/75 1 0
MSCI Slovenia Start 02.06.2003 activation quantile Nb trigger moments all data last 10 Y last 5Y /deactivation 70/50 2 0 70/70 3 0 75/55 2 0 80/60 2 0 85/65 2 0 90/70 1 0 95/75 1 0
MSCI Spain Start 31.12.1980 -- Price Index activation quantile Nb trigger moments /deactivation all data last 10 Y last 5Y 70/50 10 4 70/70 10 4 75/55 10 4 80/60 9 4 85/65 8 3 90/70 7 3 95/75 4 3
all data 0 0 0 0 0 0 0
all data 2 2 2 2 2 1 1
% active Trigger time last 10 Y last 5Y 31 41 31 41 28 41 26 39 22 37 19 26 15 23
all data 1 2 1 1 1 1 1
all data 2 3 2 2 2 2 2
all data 1 1 1 1 1 1 1
all data 2 3 2 2 2 1 1
all data 3 3 3 3 2 2 2
% active Trigger time last 10 Y last 5Y 54 0 54 0 27 0 27 0 27 0 27 0 27 0
% active Trigger time last 10 Y last 5Y 39 16 47 24 32 16 30 15 25 13 21 10 10 10
% active Trigger time last 10 Y last 5Y 46 50 45 48 42 39 41 38 35 36 29 30 24 30
% active Trigger time last 10 Y last 5Y 25 0 25 0 25 0 22 0 21 0 21 0 20 0
% active Trigger time last 10 Y last 5Y 41 0 39 0 28 0 24 0 22 0 14 0 12 0
% active Trigger time last 10 Y last 5Y 40 52 39 51 39 51 35 51 33 48 26 37 15 31
0 0 0 0 0 0 0
60 60 60 59 53 32 27
mean active trigger (d) all data last 10 Y last 5Y 262 0 262 0 262 0 262 0 262 0 262 0 262 0
mean active trigger (d) all data last 10 Y last 5Y 376 358 373 354 388 358 365 341 370 317 394 339 408 304
33 48 31 31 25 21 20
mean active trigger (d) all data last 10 Y last 5Y 392 424 334 311 405 406 380 399 318 329 267 271 261 261
79 77 58 55 51 41 40
mean active trigger (d) all data last 10 Y last 5Y 411 430 283 314 290 338 285 327 315 310 300 264 303 260
31 31 31 27 26 26 25
mean active trigger (d) all data last 10 Y last 5Y 404 0 397 0 398 0 354 0 343 0 337 0 329 0
79 75 54 46 44 27 24
mean active trigger (d) all data last 10 Y last 5Y 517 0 325 0 349 0 301 0 284 0 357 0 306 0
83 82 83 82 76 54 42
mean active trigger (d) all data last 10 Y last 5Y 337 337 324 331 328 335 325 332 346 415 317 321 315 267
0 0 0 0 0 0 0
394 391 394 381 345 416 345
median active trigger (d) all data last 10 Y last 5Y 262 0 262 0 262 0 262 0 262 0 262 0 262 0
median active trigger (d) all data last 10 Y last 5Y 261 268 261 262 266 268 261 262 261 262 339 339 345 304
0 0 0 0 0 0 0
394 391 394 381 345 416 345
424 311 406 399 329 271 261
median active trigger (d) all data last 10 Y last 5Y 424 424 361 311 405 406 385 399 334 329 266 271 261 261
424 311 406 399 329 271 261
514 332 376 360 334 265 259
median active trigger (d) all data last 10 Y last 5Y 420 420 261 335 262 343 261 319 262 316 262 262 261 261
514 408 376 360 334 265 259
404 397 398 354 343 337 329
median active trigger (d) all data last 10 Y last 5Y 404 0 397 0 398 0 354 0 343 0 337 0 329 0
404 397 398 354 343 337 329
517 325 349 301 284 357 306
median active trigger (d) all data last 10 Y last 5Y 517 0 287 0 349 0 301 0 284 0 357 0 306 0
517 287 349 301 284 357 306
362 354 359 355 492 350 271
median active trigger (d) all data last 10 Y last 5Y 282 330 261 330 261 328 261 328 261 357 261 331 295 261
397 397 394 393 492 350 271
53
MSCI Sweden Start 30.12.1980 activation quantile Nb trigger moments /deactivation all data last 10 Y last 5Y 70/50 8 3 70/70 8 3 75/55 8 3 80/60 8 3 85/65 6 2 90/70 6 2 95/75 5 2
MSCI United Kingdom Start 31.12.1970 Nb trigger moments activation quantile all data last 10 Y last 5Y /deactivation 70/50 10 2 70/70 11 2 75/55 10 2 80/60 7 2 85/65 6 2 90/70 5 2 95/75 4 2
all data 2 2 2 2 1 1 1
all data 1 1 1 1 1 1 1
% active Trigger time last 10 Y last 5Y 36 39 40 38 35 38 33 37 26 27 25 23 21 23
% active Trigger time last 10 Y last 5Y 35 29 37 28 32 26 24 26 20 24 18 23 12 23
55 53 54 53 33 26 25
mean active trigger (d) all data last 10 Y last 5Y 381 334 424 326 371 329 350 324 366 346 346 298 353 295
38 37 32 32 28 26 25
mean active trigger (d) all data last 10 Y last 5Y 384 381 366 369 347 342 379 340 375 314 388 300 342 297
357 344 352 346 429 333 327
median active trigger (d) all data last 10 Y last 5Y 274 262 261 262 270 262 261 262 261 346 261 298 261 295
357 344 352 346 429 333 327
500 476 422 417 365 338 331
median active trigger (d) all data last 10 Y last 5Y 262 381 261 369 261 342 262 340 314 314 338 300 301 297
500 476 422 417 365 338 331
T ABLE 22: ACTIVATION ANALYSIS FOR DIFFERENT ACTIVATING / DEACTIVATING QUANTILES OF SIMULATED MAXIMUM DRAWDOWN DISTRIBUTION ( MSCI COUNTRY INDICES , PRICE INDEX ), WHEN THE ADAPTATION IS ASSUMED TO BE ACTIVE FOR AT LEAST A YEAR , I . E . 261 TR
54
ANNEX 6: ACTIVATION ANALYSIS FOR DIFFERENT ACTIVATING/DEACTIVATING QUANTILES OF SIMULATED MAXIMUM DRAWDOWN DISTRIBUTION (PORTFOLIOS), WHEN THE ADAPTATION IS ASSUMED TO BE ACTIVATED FOR AT LEAST A YEAR, I.E. 261 TRADING DAYS Portfolio Austria Start 03.01.2000 Nb trigger moments activation quantile last 10 Y last 5Y /deactivation quantile all data 70/50 4 2 70/70 5 3 75/55 2 2 80/60 2 2 85/65 2 2 90/70 2 2 95/75 2 2
Portfolio Belgium Start 03.01.2000 Nb trigger moments activation quantile last 10 Y last 5Y /deactivation quantile all data 70/50 4 2 70/70 4 2 75/55 4 2 80/60 4 2 85/65 2 2 90/70 2 2 95/75 2 2
Portfolio Bulgaria Start 31.05.2006 Nb trigger moments activation quantile last 10 Y last 5Y /deactivation quantile all data 70/50 1 70/70 1 75/55 1 80/60 1 85/65 1 90/70 1 95/75 1 -
Portfolio Czech Republic Start 10.02.2006 Nb trigger moments activation quantile last 10 Y last 5Y /deactivation quantile all data 70/50 2 70/70 2 75/55 1 80/60 1 85/65 1 90/70 1 95/75 1 -
Portfolio Denmark Start 07.01.2004 Nb trigger moments activation quantile last 10 Y last 5Y /deactivation quantile all data 70/50 2 70/70 2 75/55 2 80/60 2 85/65 2 90/70 1 95/75 1 -
Portfolio Estonia Start 02.06.2003 Nb trigger moments activation quantile last 10 Y last 5Y /deactivation quantile all data 70/50 3 70/70 3 75/55 2 80/60 1 85/65 1 90/70 1 95/75 1 -
all data 2 3 2 2 2 2 2
all data 2 2 2 2 2 2 2
all data 1 1 1 1 1 1 1
all data 2 2 1 1 1 1 1
all data 2 2 2 2 2 1 1
all data 2 2 2 1 1 1 1
% active Trigger time last 10 Y last 5Y 42 35 43 36 25 33 24 32 24 31 22 28 18 23
% active Trigger time last 10 Y last 5Y 40 32 39 31 35 26 32 22 16 21 16 21 15 20
% active Trigger time last 10 Y last 5Y 25 23 24 23 19 17 16 -
% active Trigger time last 10 Y last 5Y 46 32 19 18 18 18 15 -
% active Trigger time last 10 Y last 5Y 34 33 30 29 29 15 14 -
% active Trigger time last 10 Y last 5Y 43 42 32 20 18 17 12 -
65 67 63 63 61 57 47
mean active trigger (d) all data last 10 Y last 5Y 359 457 291 311 425 425 419 419 406 406 370 370 303 303
64 63 52 44 43 42 40
mean active trigger (d) all data last 10 Y last 5Y 339 416 335 409 301 340 272 283 279 279 274 274 261 261
33 31 32 31 26 23 22
mean active trigger (d) all data last 10 Y last 5Y 429 398 421 402 336 304 281 -
65 46 27 25 25 25 21
mean active trigger (d) all data last 10 Y last 5Y 424 297 356 328 327 326 272 -
61 60 54 53 52 28 26
mean active trigger (d) all data last 10 Y last 5Y 400 397 353 346 341 363 336 -
mean active trigger (d) all data last 10 Y last 5Y 60 364 58 356 57 404 36 517 35 458 32 421 24 306 -
425 290 412 408 396 370 303
median active trigger (d) all data last 10 Y last 5Y 295 457 262 292 425 425 419 419 406 406 370 370 303 303
425 292 412 408 396 370 303
416 409 340 283 279 274 261
median active trigger (d) all data last 10 Y last 5Y 326 416 319 409 289 340 264 283 279 279 274 274 261 261
416 409 340 283 279 274 261
429 398 421 402 336 304 281
median active trigger (d) all data last 10 Y last 5Y 429 398 421 402 336 304 281 -
429 398 421 402 336 304 281
424 297 356 328 327 326 272
median active trigger (d) all data last 10 Y last 5Y 424 297 356 328 327 326 272 -
424 297 356 328 327 326 272
395 393 353 346 341 363 336
median active trigger (d) all data last 10 Y last 5Y 400 397 353 346 341 363 336 -
395 393 353 346 341 363 336
median active trigger (d) all data last 10 Y last 5Y 392 278 381 261 372 404 472 517 458 458 421 421 306 306 -
392 381 372 472 458 421 306
55
Portfolio Finland Start 03.01.2000 Nb trigger moments activation quantile last 10 Y last 5Y /deactivation quantile all data 70/50 3 3 70/70 3 3 75/55 2 2 80/60 2 2 85/65 2 1 90/70 2 1 95/75 2 1
all data 1 1 1 1 1 1 1
Portfolio France Start 03.01.2000 Nb trigger moments activation quantile /deactivation quantile all data last 10 Y last 5Y 70/50 4 2 70/70 5 3 75/55 4 2 80/60 3 2 85/65 3 2 90/70 3 2 95/75 2 1
all data 2 3 2 2 2 2 1
Portfolio Germany Start 03.01.2000 Nb trigger moments activation quantile last 10 Y last 5Y /deactivation quantile all data 70/50 4 2 70/70 4 2 75/55 4 2 80/60 3 1 85/65 3 1 90/70 2 1 95/75 1 1
all data 2 2 2 1 1 1 1
Portfolio Hungary Start 03.01.2000 Nb trigger moments activation quantile /deactivation quantile all data last 10 Y last 5Y 70/50 4 3 70/70 5 4 75/55 3 2 80/60 2 1 1 85/65 2 2 1 90/70 95/75 1 1
all data 2 3 2 1 1 1 1
Portfolio Ireland Start 30.03.2001 Nb trigger moments activation quantile last 10 Y last 5Y /deactivation quantile all data 70/50 2 2 2 70/70 2 75/55 2 2 2 2 80/60 85/65 2 2 2 90/70 2 95/75 1 1
all data 2 2 2 2 2 2 1
% active Trigger time last 10 Y last 5Y 40 37 39 36 31 25 30 24 29 14 26 11 24 11
% active Trigger time last 10 Y last 5Y 40 28 40 30 39 27 30 26 27 23 25 23 16 12
% active Trigger time last 10 Y last 5Y 38 27 36 26 34 22 27 13 25 13 17 13 9 12
% active Trigger time last 10 Y last 5Y 46 43 48 45 31 26 22 16 19 13 18 12 10 12
% active Trigger time last 10 Y last 5Y 24 29 27 23 24 28 23 27 22 27 20 23 10 12
35 31 30 29 28 23 21
mean active trigger (d) all data last 10 Y last 5Y 458 325 443 311 526 327 514 318 502 365 451 294 413 277
55 60 54 52 46 45 23
mean active trigger (d) all data last 10 Y last 5Y 341 359 272 259 330 352 340 336 310 297 284 296 281 301
54 52 44 26 26 26 25
mean active trigger (d) all data last 10 Y last 5Y 329 353 312 341 288 287 305 340 285 332 297 332 324 324
66 70 52 32 25 24 23
mean active trigger (d) last 5Y all data last 10 Y 372 358 294 294 315 341 339 417 325 293 285 308 299 299
58 54 57 54 53 47 24
mean active trigger (d) last 10 Y last 5Y all data 376 376 349 349 368 368 351 351 348 348 305 305 317 317
451 409 392 374 365 294 277
median active trigger (d) all data last 10 Y last 5Y 451 262 409 262 526 327 514 318 502 365 451 294 413 277
451 409 392 374 365 294 277
359 259 352 336 297 296 301
median active trigger (d) all data last 10 Y last 5Y 339 359 262 261 319 352 348 336 332 297 261 296 281 301
359 261 352 336 297 296 301
353 341 287 340 332 332 324
median active trigger (d) all data last 10 Y last 5Y 306 353 283 341 290 287 314 340 262 332 297 332 324 324
353 341 287 340 332 332 324
428 305 341 417 325 308 299
median active trigger (d) all data last 10 Y last 5Y 369 424 296 328 263 341 339 417 293 325 308 285 299 299
428 395 341 417 325 308 299
376 349 368 351 348 305 317
median active trigger (d) last 5Y all data last 10 Y 376 376 349 349 368 368 351 351 348 348 305 305 317 317
376 349 368 351 348 305 317
Portfolio Italy Start 03.01.2000 Nb trigger moments activation quantile last 10 Y last 5Y /deactivation quantile all data 70/50 4 2 70/70 4 2 75/55 4 2 80/60 2 2 85/65 2 2 90/70 2 2 95/75 2 2
56
Portfolio Lithuania Start 01.06.2009 Nb trigger moments activation quantile last 10 Y last 5Y /deactivation quantile all data 70/50 1 70/70 1 75/55 1 80/60 1 85/65 1 90/70 1 95/75 1 -
all data 2 2 2 2 2 2 2
all data -
% active Trigger time last 10 Y last 5Y 36 27 36 27 36 27 20 26 19 25 18 24 17 23
% active Trigger time last 10 Y last 5Y 27 27 27 27 27 27 27 -
54 53 53 52 51 49 46
-
mean active trigger (d) all data last 10 Y last 5Y 307 353 304 347 304 347 336 336 330 330 316 316 297 297
mean active trigger (d) all data last 10 Y last 5Y 262 262 262 262 262 262 262 -
353 347 347 336 330 316 297
median active trigger (d) all data last 10 Y last 5Y 276 353 270 347 273 347 336 336 330 330 316 316 297 297
353 347 347 336 330 316 297
-
median active trigger (d) all data last 10 Y last 5Y 262 262 262 262 262 262 262 -
-
Portfolio Malta Start 03.02.2009 Nb trigger moments activation quantile /deactivation quantile all data last 10 Y last 5Y 70/50 1 70/70 1 75/55 1 80/60 1 85/65 1 90/70 1 1 95/75
all data -
Portfolio Netherlands Start 03.01.2000 Nb trigger moments activation quantile last 10 Y last 5Y /deactivation quantile all data 70/50 4 2 70/70 4 2 75/55 3 1 80/60 3 1 85/65 3 1 90/70 3 1 95/75 2 1
Portfolio Poland Start 07.07.2005 Nb trigger moments activation quantile last 10 Y last 5Y /deactivation quantile all data 70/50 2 70/70 2 75/55 2 80/60 1 85/65 1 90/70 1 95/75 1 -
all data 1 1 1 1 1 1 1
all data 1 1 1 -
Portfolio Portugal Start 30.03.2001 Nb trigger moments activation quantile /deactivation quantile all data last 10 Y last 5Y 70/50 2 2 70/70 2 2 75/55 2 2 80/60 2 2 85/65 2 2 90/70 2 2 95/75 2 2
Portfolio Slovenia Start 01.04.2008 Nb trigger moments activation quantile /deactivation quantile all data last 10 Y last 5Y 70/50 2 70/70 2 75/55 2 80/60 2 85/65 1 90/70 1 95/75 1 -
all data 2 2 2 2 2 2 2
all data -
% active Trigger time last 10 Y last 5Y 25 25 25 25 25 25 25 -
-
% active Trigger time last 10 Y last 5Y 36 26 35 26 28 16 27 16 24 12 24 12 16 11
33 33 33 32 24 23 22
% active Trigger time last 10 Y last 5Y 30 26 26 13 13 13 13 -
mean active trigger (d) all data last 10 Y last 5Y 307 343 302 342 316 424 314 418 276 306 275 301 275 289
mean active trigger (d) all data last 10 Y last 5Y 20 298 20 262 20 262 262 262 262 262 -
% active Trigger time last 10 Y last 5Y 29 35 27 33 22 26 21 25 21 25 20 24 18 22
% active Trigger time last 5Y last 10 Y 45 43 44 43 22 21 21 -
mean active trigger (d) last 5Y all data last 10 Y 262 262 262 262 262 262 262 -
70 65 52 51 49 48 43
mean active trigger (d) all data last 10 Y last 5Y 454 454 424 424 340 340 329 329 319 319 312 312 283 283
mean active trigger (d) all data last 10 Y last 5Y 284 274 282 273 278 261 261 -
-
-
median active trigger (d) all data last 10 Y last 5Y 262 262 262 262 262 262 262
-
425 423 424 418 306 301 289
median active trigger (d) all data last 10 Y last 5Y 271 343 262 342 262 424 262 418 262 306 262 301 275 289
425 423 424 418 306 301 289
median active trigger (d) all data last 10 Y last 5Y 261 298 261 262 261 262 262 262 262 262 -
261 261 261 -
454 424 340 329 319 312 283
median active trigger (d) all data last 10 Y last 5Y 454 454 424 424 340 340 329 329 319 319 312 312 283 283
454 424 340 329 319 312 283
-
median active trigger (d) all data last 10 Y last 5Y 284 274 282 273 278 261 261
-
447 422 430 410 292 290 216
median active trigger (d) all data last 10 Y last 5Y 290 447 265 422 274 430 275 410 292 292 290 290 216 216
447 422 430 410 292 290 216
median active trigger (d) all data last 10 Y last 5Y 343 424 343 337 322 337 337 412 337 379 386 379 370 345 370 300 300 300 285 285 285
343 337 337 379 370 300 285
Portfolio Spain Start 04.01.2000 Nb trigger moments activation quantile last 10 Y last 5Y /deactivation quantile all data 70/50 4 2 70/70 4 2 75/55 4 2 80/60 3 2 85/65 2 2 90/70 2 2 95/75 2 2
Portfolio Sweden Start 30.03.2001 Nb trigger moments activation quantile last 10 Y last 5Y /deactivation quantile all data 70/50 3 2 70/70 4 2 75/55 3 2 80/60 2 1 85/65 2 1 90/70 1 1 95/75 1 1
all data 2 2 2 2 2 2 2
all data 2 2 2 1 1 1 1
% active Trigger time last 10 Y last 5Y 41 34 40 32 40 33 32 32 17 22 17 22 13 17
mean active trigger (d) all data last 10 Y last 5Y 354 447 342 422 346 430 360 410 292 292 290 290 216 216
69 65 66 63 45 45 33
% active Trigger time last 10 Y last 5Y 37 26 43 26 36 26 25 15 22 14 10 12 9 11
mean active trigger (d) all data last 10 Y last 5Y 53 385 343 52 329 337 52 374 337 29 386 379 28 345 370 23 300 300 22 285 285
57
Portfolio United Kingdom Start 30.03.2001 Nb trigger moments activation quantile /deactivation quantile all data last 10 Y last 5Y 70/50 3 2 70/70 3 2 75/55 2 1 2 1 80/60 1 85/65 2 2 1 90/70 95/75 1 1
Portfolio Eurozone Start 30.03.2001 Nb trigger moments activation quantile last 10 Y last 5Y /deactivation quantile all data 70/50 3 2 70/70 3 2 75/55 3 2 80/60 2 1 85/65 2 1 90/70 1 1 95/75 1 1
all data 1 1 1 1 1 1 1
all data 2 2 2 1 1 1 1
% active Trigger time last 10 Y last 5Y 28 34 27 32 25 17 25 17 22 16 20 13 10 12
mean active trigger (d) last 10 Y last 5Y all data 359 347 35 347 33 332 35 385 452 34 381 447 32 341 421 344 26 303 306 306 24
median active trigger (d) all data last 10 Y last 5Y 456 323 359 432 302 347 452 385 452 447 381 447 421 421 341 344 303 344 306 306 306
456 432 452 447 421 344 306
% active Trigger time last 10 Y last 5Y 33 28 32 27 32 27 23 16 22 16 11 13 10 13
mean active trigger (d) all data last 10 Y last 5Y 343 362 331 347 331 346 361 423 338 414 335 335 325 325
median active trigger (d) all data last 10 Y last 5Y 304 362 299 347 302 346 361 423 338 414 335 335 325 325
362 347 346 423 414 335 325
56 53 53 33 32 26 25
362 347 346 423 414 335 325
T ABLE 23: ACTIVATION ANALYSIS FOR DIFFERENT ACTIVATING / DEACTIVATING QUANTILES OF SIMULATED MAXIMUM DRAWDOWN DISTRIBUTION (PORTFOLIOS ), WHEN THE ADAPTATION IS ASSUMED TO BE ACTIVATED FOR AT LEAST A YEAR , I . E . 261 TRADING DAYS .
58
ANNEX 7: MAXIMUM DRAWDOWNS OF PORTFOLIOS AND INDIVIDUAL ASSET CLASSES CALCULATED OVER A PERIOD OF 1 YEAR FOR VARIOUS COUNTRIES 1,0 0,9 0,8
Quantiles
0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0
Gov_EUR_AT
Corp_EUR_AT
Equ_EUR_AT
Portf olio Austria
1,0 0,9 0,8
Quantiles
0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0
59 Gov_EUR_BE
Corp_EUR_BE
Equ_EUR_BE
Portf olio Belgium
1,0 0,9 0,8
Quantiles
0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0
Gov_Europe
Corp_Europe
Equ_BGN
Portf olio Bulgaria
1,0 0,9 0,8
Quantiles
0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0
Gov_CZK
60
Corp_Europe
Equ_CZK
Portf olio Czech Republic
1,0 0,9 0,8
Quantiles
0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0
Gov_DKK
Corp_Europe
Equ_DKK
Gov_Europe
Corp_Europe
Equ_EUR_EE
Portf olio Denmark
1,0 0,9 0,8
Quantiles
0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0
Portf olio Estonia
61
1,0 0,9 0,8
Quantiles
0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0
Gov_EUR_FI
Corp_EUR_FI
Equ_EUR_FI
Gov_EUR_FR
Corp_EUR_FR
Equ_EUR_FR
Portf olio Finland
1,0 0,9 0,8
Quantiles
0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0
62
Portf olio France
1,0 0,9 0,8
Quantiles
0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0
Gov_EUR_DE
Corp_EUR_DE
Equ_EUR_DE
Portf olio Germany
1,0 0,9 0,8
Quantiles
0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0
Gov_Europe
Corp_Europe
Equ_HUF
Portf olio Hungary
63
1,0 0,9 0,8
Quantiles
0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0
Gov_EUR_IE
Corp_Europe
Equ_EUR_IE
Portf olio Ireland
1,0 0,9 0,8
Quantiles
0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0
Gov_EUR_IT
64
Corp_EUR_IT
Equ_EUR_IT
Portf olio Italy
1,0 0,9 0,8
Quantiles
0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0
Gov_Europe
Corp_Europe
Equ_LTL
Gov_EUR_MT
Corp_Europe
MSCI_Europe
Portf olio Lithuania
1,0 0,9 0,8
Quantiles
0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0
Portf olio Malta
65
1,0 0,9 0,8
Quantiles
0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0
Gov_EUR_NL
Corp_EUR_NL
Equ_EUR_NL
Portf olio Netherlands
1,0 0,9 0,8
Quantiles
0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0
Gov_PLN
66
Corp_Europe
Equ_PLN
Portf olio Poland
1,0 0,9 0,8
Quantiles
0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0
Gov_EUR_PT
Corp_Europe
Equ_EUR_PT
Portf olio Portugal
Gov_EUR_SI
Corp_Europe
Equ_EUR_SI
Portf olio Slovenia
1,0 0,9 0,8
Quantiles
0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0
67
1,0 0,9 0,8
Quantiles
0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0
Gov_EUR_ES
Corp_EUR_ES
Equ_EUR_ES
Portf olio Spain
1,0 0,9 0,8
Quantiles
0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0
Gov_EUR_SE
68
Corp_Europe
Equ_EUR_SE
Portf olio Sweden
1,0 0,9 0,8
Quantiles
0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0
Gov_EUR_UK
Corp_Europe
Equ_EUR_UK
Portf olio UK
F IGURE 12: MAXIMUM DRAWDOWNS OF PORTFOLIOS AND INDIVIDUAL ASSET CLASSES CALCULATED OVER A PERIOD OF 1 YEAR FOR VARIOUS COUNTRIES
69
