How to design efficient defined contribution pension systems?

Defined contribution pension systems offer solutions to the problem of an ageing population, but they also present many challenges. The oldest of these, in Chile, is a relevant example.

Few subjects touch upon everyone’s lives and generate so much controversy as pension funds. Developed and developing nations alike have witnessed massive – and at times, violent – protests by the population that oppose changes to the system or want to fully replace the one in place.

Pension systems come in different varieties. Defined Benefit (DB) and Defined Contribution (DC) are the two extreme cases, and most systems in the world use one or another, or some type of hybridization between the two. The former, which is in place in countries like France and Brazil, fixes workers’ salaries after retirement. The latter, exemplified by Chile and the 401k accounts in the United States, sets the amount the workers must save every month, e.g., a fixed percentage of their salary. As population pyramids in the world change, it becomes harder to fund the retirement of an aging population, which is a key cause of the massive shift from DB to DC systems in the last decades.

The 70% mark

DC systems are a response to such challenge in the sense that in such systems each worker is responsible for his/her own retirement savings. The monthly deposits go to an account exclusively dedicated to increasing the lump sum the worker will have access to at retirement age, usually in the form of monthly payments until death (the so-called annuities). Despite a global preference for DC systems throughout the world, they are not without challenges of their own. A key premise of those systems is that retirement savings must be invested in order to generate enough money for a comfortable retirement, which is estimated to be around 70% of the workers’ last salary. This value is referred to as the replacement rate.

Let us use the Chilean system as an example since it is the oldest DC system in the world (it started in 1981), and it is mandatory for all workers with very few exceptions (the armed forces and the police for instance). The worker can freely choose between 7 pension fund administrators (AFP, for their Spanish acronym), which are companies with an exclusive mandate to manage pension savings until retirement. In each AFP, the worker can choose how to invest the money among five funds: A, B, C, D, and E. Their difference is their risk-return profile: fund A is heavily invested in stocks, and therefore has higher expected returns and risks. On the other hand, fund E is mostly comprised of bonds: it has lower returns, but very small risk. The other intermediate funds can be seen as a smooth transition between funds A and E.

A pension system full of challenges

Therefore, to achieve a replacement rate of around 70% the AFPs must obtain attractive returns for each of the funds they manage. The regulator in Chile defined the funds by portfolio limits: fund A can have up to 80% of its value invested in stocks, fund B 60%, fund C 40%, fund D 20%, and fund E can have at most 5%. Much like the QI is a proxy for intelligence and GDP for economic growth, those limits are supposed to be a proxy for the risk of each multifund. Unfortunately, several complications arise from such a strategy.

First, how can those limits be defined? If they are too stringent, a fund might have to sell a position in a profitable investment to satisfy the requirements of the law. If they are too loose, the desired risk-return profiles might not be ordered according to the design. With the recent addition of investment instruments such as alternatives and derivatives, the current legislation requires the definition of 50 of such limits! Given that an additional 1% of returns per year can result in a pension up to 25% higher, those limits might have a dire effect on the final pension of a worker.

Limits and risk measures

A proposal that has been discussed in academic and practitioner circles is to replace those limits with risk measures. Risk measures are a way to directly measure the risk of a portfolio without imposing specific limits on every financial instrument allowed by the regulator. Each of the funds A through E would be characterized by a single number: the amount of risk tolerated by each of the funds. The tolerance parameter would be higher for fund A, allowing for exposure to instruments with higher returns, and smaller for fund E, limiting the choices to safer instruments. Having risk limits would automatically differentiate between instruments belonging to the same class, such as bonds. With portfolio limits, all bonds are treated the same since the percentage they represent in the portfolio is what is being measured, not the overall risk of the positions.

What is the appropriate risk measure for defining the multifunds? That is not a simple question, but there are different alternatives in the literature. Classical measures that focus on volatility, such as the standard deviation, are not recommended because in long-term planning the volatility of the journey does not matter. A possible candidate is the Conditional Value-at-Risk (CVaR), which is a risk measure that has been extensively used in financial problems and it is relatively easy to compute. The regulator in Mexico already uses the CVaR, in combination with asset class limits.

In summary, the movement to DC pension systems, or at least hybrid systems with a DC component seems to be irreversible. It is important to continue discussing how to design such a system, with special attention to the investment rules in the accumulation phase, to be able to provide a satisfactory retirement income for the workers during their retirement phase.