The French are proud of their first-pillar state pension system, which by guaranteeing a defined benefit pension for private sector workers fits the country's image of social inclusion. Unfortunately, the pay-as-you-go system is forecast to plunge into deficit over the next few decades, with a shortfall of EUR37 billion expected by 2040, according to French government figures.
In 2001, the then socialist government devised a very French solution to the problem, when it created the Fonds de Reserve pour les Retraites (FRR). Shunning the private sector solutions generally favoured in the Anglo-Saxon world, the FRR was a so-called buffer fund, managed by the state solely for the purpose of filling a projected funding gap.
Funded by a mixture of social security contributions and privatisation receipts, by the end of 2004, the Paris-based FRR had assets of over EUR20 billion, of which 55% were invested in equities and the remainder in bonds, with mandates currently split between 20 different asset managers. Nicholas Dunbar met the chief investment officer of FRR, Jean-Louis Nakamura, in his office.
Life & Pensions: Basically, the French pension reserve fund is there to protect French taxpayers and pensioners ...
Nakamura: Exactly. Maybe thanks to this fund, the government might be in a position not to hike taxes as much as it thought, or not to decrease pension benefits so much. That's the buffer role of the French pension reserve fund. Of course it will be highly dependent on the total reserves accumulated during that period, and the demographics which determine the size of the funding gap. Currently, there are discussions taking place close to the prime minister's office on how the fund may be committed to contribute each year to a certain percentage of the financing gap from 2020 onwards.
Does this commitment amount to a liability?
We don't have specific liabilities. But the commitments will be quasi-liabilities, which at this stage have not been defined very precisely. This is an issue for us because as you can imagine we need this schedule to define our asset allocation. The only target we have at this stage is to accumulate as much money as we can. We know that between now and 2020, and I hope it will not be too late, we will be given a specific schedule and yearly target after 2020 to commit ourselves to the first-pillar general scheme. However it's quite difficult because we started with an initial asset allocation and we know it will have to be changed during this time in order to keep our objective of probability of loss. But we don't know how the fund will be used through 2020 and beyond.
Looking at the worst-case scenario, the more dependent the French state is on using FRR, then the more concerned it will be about your risk management in order that you meet those demands.
When we modelled our risk and our asset allocation, we were really cautious in all our categories of technical assumptions. For instance, we need an assumption of how the fund will be used after 2020. There are three scenarios: firstly, a very dramatic scenario in which the fund is used in one year, a more cautious scenario where the fund is used in 10 years, and a third scenario recommended by other people where the fund is drawn down over 20 years. We chose the 10-year scenario - we shortened the duration of the fund. We chose to keep the scenario where the average level of contributions was constant at around EUR1.5-2bn per year, depending on social and tax contributions. Had we made a bolder assumption about contributions, for example the average contribution we actually benefited from on a cumulative basis since the start of the fund, it would have reduced portfolio volatility. So, ironically, the biggest risk we faced was maybe to have been too conservative in our assumptions. That may be a paradox, but you have to be aware that it was an innovative debate within a supervisory body composed of social and labour representatives who decided unanimously to adopt this diversified asset allocation.
If you expect to pay a stream of fixed payments over 10 years after 2020, presumably you would use a bond-like discounting factor to work out the present value of this liability today. Is that your target for asset allocation?
Actually, we did not build our portfolio by fixing a target to which we could apply a discount rate for our post-2020 liabilities. We built our portfolio mix with different assumptions for each asset category, and we asked our supervisory body what they were most comfortable with. And they chose to express their risk aversion function by the objective of having a very low probability of making a capital loss by 2020 on the total amount likely to be transferred from the state to the fund.
You must have avoided a lot of interest rate risk by doing that.
That's why we are not in the same situation as many life insurance companies and pension funds who are obliged to buy very expensive bonds because they have to put a discount rate into their asset and liability models. We have a little morefreedom not to buy interest rate products if we think they are too expensive.
You have a probabilistic target of not losing capital?
Yes. And we model that by mixing together the various assumptions in terms of volatility, expected returns and correlations between different categories of assets. We use a mean-variance efficient frontier method. We could have used a more complicated method in order to escape from the criticism of the Normal distribution, but one of our priorities was to be as clear as possible and fully understood by a board composed of non-specialists. That was our governance challenge. We went into a lot of detail about different categories of products, about the way the asset mix has to be changed over time in order to keep the probability of loss constant. Of course, the closer you are to the liability date, the more you have to change your asset allocation, given the different properties of each asset category. We didn't set a return target, but as a result of the initial asset mix we have, given that we have associated return and volatility assumptions for each asset category, we have calculated an expected return on the initial portfolio.
What amount of capital are you applying your probability to?
It's the total contribution compounded at a real rate of 2.2%. The initial nominal return rate associated with the initial asset allocation was 6.3%, with anannual volatility of 12%.
What is the maximum probability of losing this capital that you are prepared to tolerate?
It's approximately 0.5%, or a 1 in 200 probability over 15 years. This was made very clear to the members of the board, and they accepted the risk. For each category of assumptions - risk premium of equities over bonds, nominal returns of bonds, correlation between equities and bonds - we were very conservative. For example if we had taken a negative correlation between equities and bonds, that would have prompted us to choose much morevolatile assets.
The simple approach is to take the historical performance of assets ...
At the beginning, there was a temptation to use traditional weights and take the 70-year average performance of equities and bonds, and to say to our supervisory board, if you believe in this, we can have an 80% equity allocation. But you have to keep in mind that we had this discussion at the beginning of 2003, after two years of bear markets, and people were afraid about the evolution of markets. The 20-year time horizon is not the 100-year horizon, and over a 70-year period there are quite significant probabilities that equities do not deliver more than 2% over bonds. So we put aside the methods relying on historical data, and we decided to be very simple and basic in order to be clear for the members of the board.
What did you tell them?
We drew a macroeconomic scenario in which the ageing population in Europe led to lower potential growth. The inflation may be slightly increased because transition countries entering the Eurozone go through a catch-up process in their purchasing power. It's also not impossible for an ageing population to change the balance between consumption and saving, with an upwards shock to the Eurozone inflation rate. This would result in lower real rates of bonds and a risk premium of equities over bonds at the lower range of estimates. We thought these were quite good macroeconomic assumptions. A shrinking working population could ask for a higher equilibrium wage, and there would be disputes over profit between shareholders and workers.
Have you entertained the possibility that your investment universe is insufficiently broad?
My personal view is that our investment universe is too strict and narrow, and we could improve the efficiency of our portfolio by widening the scope of our universe. There are various reasons why we did not do this from the start. The first reason arose from our regulations, which obliged us to limit our equity exposure outside the European economic area to 20%. In our macroeconomic scenario, the potential long-term growth of Europeis much lower than the emerging markets.
So you'd like to break the 20% limit?
It would be more efficient for the future French pensioner and saver to have public money invested outside the Eurozone, at least in a higher proportion than we are allowed. The second reason was mainly operational. You can't expect a fund, which is new in the French landscape and has to build everything from scratch, to do everything at the same time. We are building the portfolio progressively, and we plan to widen the scope of our investments. For example, we are planning to put private equity in our portfolio in a quite significant weight. Next year, we will do this for other assets. The decision has not been taken yet, but in my view emerging markets are a good candidate.
How about structured product investments?
That's interesting. If there is one thing that is very cheap right now, it's volatility. Right now, I'm unable to bet on volatility being more expensive in six months, and that would not be the philosophy of the fund. But over the long term, it's quite reasonable that volatility would be higher. The main problem for us is the fact that we are a public institution, and whenever we choose a new product for our needs, we need to launch a public tendering process, which lasts six to eight months. That's not the best way to accelerate the widening of our investment scope.
You've got your 99.5% risk over the next 15 years. How does it work over the next year?
We look at a way of keeping this risk constant by moving the portfolio each year to a less volatile mix. That's highly sensitive to the use of the fund after 2020, but as long as we stay within the 10-year drawdown scenario, we do not need to sharply increase the cash proportion in our portfolio. Each year we have to see whether we need to change our asset allocation in order to keep that probability of 99.5%. During the intervening periods, we use tactical asset allocation. Over a period of six to nine months we can move between the targeted weights of each broad category of assets, depending on where we see the equilibrium price of these assets.
Suppose the market makes you think that your asset allocation needed to be changed. What sort of signal would the market need to give you to tell you that?
For example, we have a debate at the moment about bonds. Have we seen a shift in the long-term equilibrium price of bonds, or is it purely an abnormal cyclical move that will be corrected in the coming months? Depending on the answer we get to this question, we will change our asset allocation, or we will decide to make a tactical bet where even if we lose money because interest rates keep declining, it's not so damaging: because we are a long-term fund and we are not in the position of a life insurer with annual obligations.
The week on Risk.net, July 14–20, 2017Receive this by email