How to improve EIOPA’s stress test

Following the stress tests in 2015 and 2017, EIOPA organised a new stress test for IORPs in 2019 and published its final report on this test in December2. This 2019 stress test consisted of three parts: stressing balance sheets, stressing cash flows and a questionnaire on ESG (see panel: ‘Main characteristics of IORP stress test 2019). Compared with previous stress tests the Cash Flow Analysis (CFA) has been further developed. In this article, we will focus on the CFA. In our view the CFA module has several advantages for stress testing.

We will discuss and illustrate these advantages by using CFA results for a ‘representative IORP’ based on the Netherlands. Subsequently, we propose some possible improvements to the CFA, as well as the way in which these c  stress tests.

Advantages of cash flow analysis
A CFA approach has several advantages. First, it produces results that are easy to interpret and understand for various stakeholders, and it also provides insight into consequences for sponsors and members and beneficiaries. The CFA results show a change in cash flows in a stress scenario compared with a baseline scenario. This is a straightforward comparison that can be related to the economic scenarios and the policy instruments used by the IORP. 

Moreover, the expected returns used for calculating the expected cash flows are comparable with the parameters that are embedded in the economic scenarios IORPs use themselves, which further facilitates interpretation and consistency. Assuming risk-free returns – the ‘alternative scenario’ EIOPA prescribed – does not provide insights in the real economic situation and creates volatility. 

Second, the CFA reflects upon the long-term nature of IORP’s assets and liabilities by exploring the impact of a stress scenario on the projected cash flows for the whole time-horizon under consideration and not only at the impact on the balance sheet now. With the CFA approach, we can assess not only the size of the stress impacts (for example benefit cuts), but also the timing (see figure 1 for an example). 

Also, the definition and calculation of cash flow is clearer and less open to model risk (or to discussion) than that of the value of a balance sheet item that summarises future cash flows. This may provide important additional insights in the effects of a shock: does the impact occur entirely in one year or is this spread out over multiple years? From EIOPA’s stress test we learn that the impact on sponsor support can be relatively higher and prevail especially in the initial years, whereas lower indexation and benefit reductions are relatively smaller but can be sustained for longer.

From a macroprudential point of view and the impact on the real economy, this distinction may be crucial. This is an added value of the CFA to the common balance sheet (CBS) approach. In the CBS approach the balance sheet item “benefit cuts”  represents the current option value of all future benefit cuts, but this item does not indicate anything about the possible timing and gives no more than a rough indication of the expected size of these cuts. From this option value it is not clear when the benefit cuts are expected to occur, what their probability is and what size they will be. Contrary to this, the CFA also gives insights into the timing and size over different years of these impacts. 

Third, the CFA takes into account all conditional/discretionary items and allows for good insight in the cash flows based on both unconditional benefits and discretionary elements in the pension contract, such as sponsor support, conditional indexation and benefit reductions. The CFA provides insight in how a stress scenario influences both the size and – as explained as second advantage – the distribution over  time of both unconditional and discretionary elements (see figure 2). 

Fourth, as the CFA is consistent with other existing tools – for instance used for asset-liability management (ALM) and continuity tests – it can also be used in own risk (solvency) assessments (ORAs) for IORPs (under the IORP II Directive). Using the same methodology for ORAs and stress tests would lower the burden and costs, particularly for smaller IORPs, and this would lead to more recognisable and useful stress test results.

Fifth, the CFA could improve the convergence between the DB and DC parts of the stress tests, as the same kind of CFA methodology could be used in both. This would also facilitate horizontal assessments including both DB and DC, which EIOPA also mentions as a desired improvement. Importantly, this would also make the ESG parts (for example, exploring climate change-related risks) of the DB and DC stress tests comparable, and would facilitiate incorporating ESG risks into the CFA framework. By integrating ESG into the CFA, the stress test could be simplified. This is important, as it is expected that sustainable finance will remain a key priority for EIOPA in coming years. EIOPA also notes in its Stress Test report that simplification is an important goal to increase the feasibility of conducting the stress test for smaller IORPs. 

Finally, the CFA approach facilitates (in the form of a voluntary module in the stress test 2019, as earlier explained) an analysis based on going concern IORPs, in which new accrual by both existing members and new entrants can be taken into account, as well as the related contributions. As a result, this approach, which was used by the Dutch IORPs participating in the 2019 stress tests (and also by IORPs in a few other EU member states), produces the most complete picture of the impact of shocks for IORPs that are open for new accruals3. These capture the full impact of a stress scenario, whereas closed modelling excludes new accrual and does not fully assess the impact on the pension system. 

Some insights from CFA for the Netherlands: Figure 1 makes clear that in the stress/adverse scenario the cash inflows – contributions and sponsor support – are hardly affected by the shock, but that cash outflows of our ‘representative IORP’ would become substantially lower. This decrease would be caused by (not giving) conditional indexation and even benefit cutting. Figure 2 demonstrates that in the stress scenario there would be some benefit cuts in the future, but that the benefits paid out will be especially lower due to granting less indexation over time4. Conditional indexation is higher in the baseline scenario (dark green line) than in the adverse scenario (dark grey line). In our calculations we used a wider range of outputs than EIOPA which show net outcomes (by summing inflows and outflows). By splitting out these cash flows, we get more insight on the impact.

Possible improvements to the CFA: As discussed above, the CFA approach has several advantages. At least three further improvements in addition to the CFA, as specified in EIOPA’s 2019 stress test for IORPs, are possible. 

Relate cash flows to economic indicators. A first improvement would be to relate the national results of the CFA to relevant macroeconomic factors, such as GDP, consumption, the aggregate wage sum of employees or aggregate profits of companies (see figure 3 for example). This gives insight in the impacts, for example, of benefit cuts on consumption, or of, for example, sponsor support on corporate profits; benefit reductions of about €1bn (figure 2) are relatively small compared with €341bn consumption. This provides an easy method for assessing the impact of a stress on the nationwide economy. This can be useful in identifying macroprudential issues, which is an important purpose of the stress test.

Use required excess return as an indicator of shifting underfunding forward to the future: The CFA, as currently specified in the stress test 2019, lacks an indicator to signal whether deficits are left for future generations to absorb. Although in most cases the stress test results are relatively clear on this issue, even without a separate indicator, in some cases this may raise questions. We propose to use the internal rate of return (IRR) method (see panel), as in the 2017 stress test and the required risk premium (over the risk-free rate) as indicators. One can calculate the IRR before the stress is applied (at t=0) and decompose this into the risk-free interest rate and a required risk premium above this rate. At the end of the projection horizon (in the 20 years stress test), the same exercise is performed (figure 4). 

In the case of our Dutch calculations, it can be observed that in the baseline scenario the required excess return at t=0 is -0.2%. This implies that the IORP is in good financial health: even a return of 20bps lower than the risk-free rate is sufficient to be able to pay out future pension benefits. At t=20, the required excess return is even -1.1%, (110bps), so over time the financial health of this IORP further improves and its required (excess) return drops. 

This means that no deficits are transferred to future generations. On the contrary, if the required risk premium would have increased, a deficit is transferred to the next generation. But for our representative IORP this is not the case, even not after the stress scenario: immediately after the stress is applied at t=0 our fund needs an excess return of 60bps, but at t=20 the required excess return needed has turned again negative at minus 90bps. In other words, the interventions undertaken by the board of the IORP – in particular benefit reductions (see figure 2) – are more than sufficient to prevent deficits from being shifted forward.

Introduction in the DB stress test of replacement rates as indicators of the impact of stress: Another improvement to the CFA would be the introduction in the DB stress test of replacement rates (pension income at retirement as percentage of average/last wage) for representative participants and pensioners of IORPs, in a similar fashion to the DC stress test (see figure 5 for participants and pensioners of our representative Dutch IORP). The impact of the stress scenario on the replacement rate then is an indicator of the impact of the stress. This also enables assessments of the impact on the macro disposable income by multiplying the impact on the replacement rate by the number of members of the IORP. 

The cash flows themselves (as presented in figures 1 and 2) show the impact on the economy itself, and in addition the replacement rates provide for a clear insight in the intergenerational fairness of the impact of shocks. Looking at different generations is a methodology used in the Netherlands to assess this fairness and the sustainability of pension deals.

In addition, the introduction of replacement rates in the DB stress test could bring forward the integration of DB and DC stress tests into a single framework. This was not possible in the specifications of the stress test 2019, owing to the fact that in these specifications the set-ups of the DB and the DC stress testing methods were incomparable. By means of the introduction of replacement rates in the DB stress test, EIOPA could bridge the current big gap between the methodologies used for DB and DC stress tests. 

This would make it possible to apply the CFA in both tests, which would entail the significant additional advantage that the results of both tests could both be compared and summed up on a European level. This will improve the comparison of the results of both types of tests and will increase the usefulness of the results when assessing the macro impact and the impact of shocks. 

Furthermore, this could offer important added value in case of a final Brexit, because in that case the DB stress test would otherwise be (too) heavily dominated by the Dutch results without integration with the DC stress test. If and when Dutch IORPs transform into one of the two new pension contracts, both being DC, then the current DB stress test would become obsolete. 

Conclusions/recommendations
The stress test 2019 already included substantial improvements, in particular in providing a realistic picture of various risks (to the economy, financial stability, members and beneficiaries, and sponsoring undertakings) when compared with the previous stress tests of 2015 and 2017. In particular, the improved CFA, including the open modelling approach introduced as a voluntary module therein, can be considered as an important step forward. But in our view the stress test and the CFA could even further be improved. 

Summarising, we would recommend continuing using the CFA in future stress tests. Based on the advantages of the CFA, this seems to be a suitable model for conducting (macroprudential) stress tests by being able to relate outcomes to the European or national economy, also looking at the size and the timing of the consequences of stress scenarios. Furthermore, the open (going concern) modelling module which was introduced in the 2019 stress test – although this might to a certain extent be country-specific – provides for the most complete picture of impact of shocks. 

For future stress tests and CFAs, we would therefore in the first place recommend more member states to conduct this module, as it provides for very interesting additional information and insights, while the additional workload imposed on both IORPs and national supervisory authorities is limited.

For future CFAs, we furthermore recommend including the additional possible improvements to the CFA that we have discussed in this article. In particular, we suggest to include the extensions of: 

• relating the CFA results to macroeconomic factors (GDP, consumption, wage sum, profits);

• the application of the IRR and required excess return approach as indicators of the sustainability of the (funding of) pension deals. These two extensions allow clearer insights in the potential macroprudential risks stemming from IORPs.

Adding replacement rates for participants and pensioners of DB IORPs could also accelerate the integration of the DB and the DC stress tests into a single framework. This will improve the comparison of the results of both tests as well as the usefulness of the results for assessing the macro impact of stress scenarios. This allows for (joint) horizontal assessments of the impact on both DB and DC schemes (which EIOPA is also seeking) and will give more and clearer insights in the fairness and the sustainability of the IORP pension deals.

1 EIOPA, Frankfurt, 2 April 2019, Occupational Pensions Stress Test 2019.
2 EIOPA-19/673, Frankfurt, 17 December 2019, 2019 IORPs Stress Test Report.
3 This is of course dependent on the contract boundaries of the pension deal.
4 De Nederlandsche Bank, 17 December 2019, DNBulletin: Stresstest: Schock op financiële markten werkt nog lang door op pensioen en economie.

Agnes Joseph is an actuary at Achmea; Niels Kortleve is innovation manager at PGGM; Wilfried Mulder is senior policy adviser at APG; Siert Vos is senior policy officer, balance sheet management, PGB