EIOPA: A Dutch view on stress tests
Agnes Joseph, Niels Kortleve, Wilfried Mulder, Sibylle Reichert, Peter Vlaar and Siert Vos examine the relevance of EIOPA’s stress testing regime and argue the case for alternative methods of determining a fund’s resilience
At a glance
• Dutch IORPs do not pose a systemic risk to the financial sector.
• Although arguably too simplistic for modern DC IORPs, the DC stress test overall functions well.
• However, an asset-liability study would provide better insights into the impact of stress scenarios on defined benefit IORPs.
Following stress tests for insurance companies and banks by European supervisory authorities, the European Insurance and Occupational Pensions Authority (EIOPA) has launched a stress test for occupational pension funds (IORPs).
A selection of IORPs in 17 EU member states performed calculations from May to August 2015. The purpose of the stress test was to assess the resilience of IORPs in stress scenarios, to evaluate potential systemic risks of IORPs and to enhance consumer protection.
Below we outline some insights from the perspective of the Dutch pension sector.
Why stress test IORPs?
In the financial sector stress tests are used to assess the resilience of the balance sheet of financial institutions. It is important to assess risks and vulnerabilities because failure could disrupt financial markets. The potential impact of a financial player’s failure on other financial institutions and on public finances – if the government has to provide a bail-out – is indicated as ‘systemic risk’: the risk of the failure of one institution destabilising the whole system. We have seen that a domino effect can occur, particularly in the case of banks, once one important player runs into trouble.
IORPs, on the other hand, are different in nature. Dutch IORPs are institutions that do not have shareholders. They are unleveraged and, most importantly, cannot go into default or be the subject of a run – a rush of members and beneficiaries withdrawing their entitlements. Therefore, they do not pose a contagion risk to other financial institutions. On the contrary, Dutch IORPs absorb financial shocks without amplifying or transmitting these to other institutions, and thus they have a stabilising effect in financial markets, especially in times of financial market distress. Of course, Dutch IORPs do run risks, such as investment, longevity and mortality risks. These risks can be borne by
• The members and beneficiaries;
• The sponsoring employer(s);
• (Re)insurance companies.
If the risks are borne by the members and beneficiaries, the effect of financial distress is not instantaneous but evolves slowly. This is due to the long-term nature of the liabilities of Dutch IORPs and, in case of defined benefit (DB) schemes, also due to the associated recovery plans. In this case, the financial distress of a typical Dutch IORP is not transmitted to the financial system and therefore does not lead to instability of financial markets.
If the risks are borne by the sponsoring employer(s) and/or a (re)insurance company, financial distress on the IORP’s balance sheet could have a significant impact on the sponsoring employer(s) or the (re)insurance company and as such lead to systemic risk. In the case of the Netherlands, however, reinsurance is rather the exception.
EIOPA’s chairman, Gabriel Bernardino, has stated that it is important for the supervisory authorities to assess how financial shocks are transmitted through financial markets. However, as we have discussed above, financial shocks are largely absorbed within Dutch IORPs. As to the objective to protect consumer rights, we do not think the stress test as undertaken will provide the supervisor with useful insights because of the methodology used.
HBS leads to misleading conclusions
In the stress test, IORPs were asked to calculate the impact of three stress scenarios: two adverse market scenarios (with sharp declines in share prices and interest rates) and an increase in life expectancy. The stress test methodology used for defined contribution (DC) schemes is different from that used for DB/hybrid schemes.
DC schemes were asked to assess the impact of market stress on the pensionable income of members and beneficiaries by making future projections of pension benefits. Although the chosen method might be too simplistic for modern DC IORPs, the DC stress test overall functions well.
On the other hand, DB/hybrid schemes had to calculate the impact of the stress scenarios on both their national balance sheet and their holistic balance sheet (HBS). The HBS is a tool that combines a national balance sheet, taking into account the policy instruments of the IORP, such as sponsor support, (re)insurance, conditional indexation and potential benefit reductions.
In the Netherlands, most IORPs are DB or hybrid. The table shows that the impact of the stress scenarios on the HBS of a typical Dutch IORP. This IORP has neither material sponsor support nor (re)insurance. If the IORP is sufficiently funded, the nominal liabilities will be corrected for inflation by means of indexation. But if the financial position declines, indexation might be decreased and as a means of last resort nominal liabilities can even be reduced (benefit reductions).
Despite the sharp fall in assets and the rise in the best estimates of unconditional technical provisions (due to the decreases in interest rates) under the two market scenarios, the HBS funding ratio improves considerably in these scenarios. The decline in the funding ratio shown on a national balance sheet is to the contrary in the HBS more than compensated for by an increase in the value of the ex-post benefit reduction option and a decrease in the value of the indexation option.
An important reason, in addition to the rise in the best estimates of unconditional technical provisions, for the high value of the benefit reduction option, is that – also due to the aforementioned decreases in interest rates – the value of new accrual increases considerably whereas the contribution rates are not affected in the first five years. This is the case because under Dutch rules cost-covering contributions are calculated using an interest rate that has been fixed for five years at the level of interest rates at the end of 2014 and are therefore not affected by interest rate decreases.
Due to this negative total result on accrual the funding ratio further deteriorates over these five years, and as a consequence more benefit reductions are necessary. However, as contributions and new accruals are not included in the EIOPA version of the HBS (which can as a result not be considered as fully holistic), the funding deficit resulting from the negative total result on new accruals translates, via these benefit reductions, into lower liabilities and therefore a greater surplus. This is a technically valid result, but a strange one from a policy perspective, since benefit reductions should be the policy option of last resort and should not be viewed as a regular policy instrument.
Finally, the table demonstrates that, if the policy instruments were not taken into account, the funding ratios after (market) stress would be 77% and 83% respectively. Under the longevity scenario the funding ratio after stress would be 97%.
HBS-value of policy instruments
It is important to note that the option values of the policy instruments as shown on this HBS do not say anything about the expected values or the probability of future ex-post benefit reductions and indexations. Expected values of policy instruments are not part of the HBS, but can be calculated by making future projections of the assets and liabilities of the IORP. As the table shows, the expected value of indexation is 19.6% of the unconditional liabilities and the expected benefit reductions are only 1.8% of these liabilities. There is only a very small chance (1.4%) that actual reductions will be higher than the option value of benefit reductions of 20.2%. The likelihood of indexation being lower than the option value (8%) is only 16%.
What can we learn?
Obviously, as the stress scenarios consist of large negative shocks, the (national) balance sheets after stress look worse than before stress, although we have shown above that the HBS funding ratio can show an increase.
Since Dutch IORPs in principle cannot go into default, just looking at the impact on their balance sheets is not relevant. Furthermore, the option values of the policy instruments change, but these option values do not say anything about the expected values of these instruments, as shown in the figure. Therefore, this DB stress test will not produce results that provide useful information.
It would be more relevant to assess the impact of adverse scenarios on expected pension benefits of the IORP’s members and beneficiaries. This is something the stress test results do not show. The same is true for a stressed national balance sheet. A more relevant impact of stress scenarios would be how the IORP will recover, which could take years, and how this will affect members and beneficiaries. This possibly delayed impact is due to smoothing mechanisms that exist in Dutch pension regulation, such as a 10 year recovery plan. The impact on expected pension benefits could then be analysed by looking at future projections.
DC test: more insights
The results of the stress tests as calculated for DB and DC IORPs can be confusing for two reasons.
First, looking at the DB stress test the general public might interpret the option values on the HBS as expected values. As we have shown in the figure, there is a big difference between these two values. If the option value is mistaken for an expected value, this will certainly result in a decrease of confidence in the Dutch pension sector. As a consequence, the willingness to participate in collective pension schemes could decline.
Second, the results of the DC stress test cannot be compared to those of the DB test as they are based on a completely different methodology. EIOPA recognised that the balance sheet approach would not have been informative for DC IORPs. The DC stress test therefore did not assess the resilience of the IORP, but rather looked at future projections, in particular at the potential outcomes for members and beneficiaries under various defined scenarios. We therefore propose two alternatives.
DB projections or ALM?
Other tools are available that can provide insight into the financial perspectives of the IORP and the benefits it will be able to pay.
One option is to use the methodology for the DC stress test. Another alternative is an asset liability management (ALM) study. An ALM study gives better insights into the impact of stress scenarios, since it includes future projections and thereby provides information about the possible impacts on the participants.
An ALM study includes the effects of policy instruments and the future returns above the risk-free interest rate. It also explicitly shows possible future scenarios of pension benefits. It allows for analysis of what happens after a shock in both high-return and low-return scenarios. Moreover, contrary to the DB stress test approach, it provides metrics – such as expected impact and impact in a ‘bad weather’ scenario over multiple time horizons – that provide insights into the consequences of a shock.
EIOPA’s motivation for stress testing IORPs is to assess the resilience of IORPs, to evaluate systemic risk of IORPs and to enhance consumer protection. We have some critical observations with regard to the stress test and these objectives.
Dutch IORPs do not pose a systemic risk for the financial sector. During the financial crisis IORPs have worked as a stabilising factor for the financial sector. Therefore, performing a stress test that assesses systemic risk of IORPs is questionable.
With respect to resilience and consumer protection, EIOPA’s stress test for DB IORPs – based on the impacts on both the HBS and the national balance sheets – does not provide useful information. In our view it is more useful to assess the impact of different stress scenarios on the pension benefits for beneficiaries and members, and as such provide information on the macroeconomic consequences of shocks. EIOPA’s stress test for DC IORPs is better suited in this respect.
We are also concerned about possible unintended consequences of the publication of the results. Without knowing the relevant background of IORPs and the methods used for the stress tests, the results may lead to false conclusions. For example, there is a risk that values presented in the HBS are incorrectly interpreted as expected values instead of option values. In addition, the differences in outcomes of the stress tests for DC schemes (using projections) and DB schemes (based on valuation of balance sheets) will be difficult to explain.
In case of future stress tests for DB/hybrid IORPs, we think that the expected values of policy instruments, combined with results for ‘bad weather’ scenarios, are more informative than the current option values of these instruments. In our view, an ALM study, which includes projections and as such provides information about the impact on the members and beneficiaries, provides better insights into the impact of stress scenarios on DB IORPs.
Agnes Joseph is senior actuary at Syntrus Achmea Pensioenbeheer, Niels Kortleve is innovation manager at PGGM, Wilfried Mulder is senior policy adviser at APG, Sibylle Reichert is head of the Brussels office of the Pension Federation, Peter Vlaar is head of ALM modelling at APG and Siert Vos is senior policy adviser at MN