The risk models within them are built from the bottom-up. Each and every security is organised, categorised and linked to factors that underpin a range of models and tools for analysis.
Despite that level of rigour, bottom-up systems do not provide all of the answers and fundamental investment questions are often left unanswered:
1 - How does a security or portfolio link to a client's long term portfolio objective? Most bottom-up systems comprehensively assess the risk contribution of say adding a single security into a portfolio of stocks. But how does adding that same security impact a total multi-asset portfolio's chances of delivering on a risk and return target over a 10 year horizon? This is a different investment problem to solve.
2 - What is the risk of a portfolio in a multi-year context? Most portfolio risk systems estimate and predict the tracking error, volatility and Value-at-Risk of a portfolio. Often that risk is calibrated over short term time horizons to match turnover and the holding time of an individual position in a portfolio. In contrast, the investment problem for a multi-asset class portfolio is not necessarily tracking error or volatility - it is typically the failure to fund a retirement, or failure to meet liability payments. These types of multi- year objectives are typically better modelled with a top-down framework that requires a different set of assumptions for risk and the interactions between various parts of a portfolio.
3 - Can one size fit all? Not all securities are created equal. Multi-asset class portfolios are increasingly including investments that take the form of listed securities, unlisted securities, derivatives, direct assets, limited partnership interests and even cryptocurrency. Bottom-up systems need to classify and analyse each one of these investments to provide an aggregate view of risk. But one size doesn't always fit. Often bottom-up risk models had been initially built with liquid assets in mind. Shoe-horning each and every extra security into such a model may not be the best approach.
4 - Can the client get a window into their portfolio? Portfolio management systems are designed predominantly with portfolio managers in mind. Rarely do the clients of an investment manager get a seat in the cockpit. Yet to allocate capital effectively, those decision makers increasingly need a thorough understanding of the risk in each part of their portfolio - both in terms of how it is calibrated and managed from the bottom-up, as well as in the context of the total portfolio objective and design.
It is not simply a case of choosing one or the other. Bottom-up systems help to manage risk within a portfolio (usually within a single or confined mix of asset classes). In contrast, top-down technologies typically focus on the allocation of risk across portfolios and asset classes to deliver on a strategic objective. Extending on the differences:
Time horizon - a top-down system typically looks further into the future by running asset simulations and scenarios over multi-year horizons. This a different investment problem and modelling challenge to that required to manage risk over shorter horizons.
Width of asset classes - bottom-up systems often run into challenges as exposures widen into ever more esoteric and illiquid asset classes. Risk models are not additive - each new asset type requires new model inputs and a recalibration of assumptions. In contrast, a top-down framework starts from the investment objectives and then connects all asset types that a portfolio will be exposed to. An investment team will maintain assumptions for each and every type of exposure or asset class. The challenge is deciding how granular those exposures should be and how the multi-asset model gets calibrated.
Risk and return are equally important - unlike bottom-up systems where risk is the predominant consideration, expected returns are equally important in a top-down technology. This allows users to design and construct portfolios to meet long term objectives that include both risk and return aspects.
Considerations for connecting bottom-up and top-down systems
For investment groups wanting to reap the benefits of combining top-down and bottom-up technologies, there are five important considerations to make:
1 -A clearly defined asset class schema is a necessary starting point - bottom-up classifications of securities and portfolio types often dictate how an asset class framework is defined. Yet investment groups must also approach the problem from the other direction - i.e. having a clear framework to shape how capital gets allocated to meet longer term portfolio objectives. Too often it's a case of the tail wagging the dog where the legacy approach to implementing portfolios of securities dictates the top- down structure.
2 -Don't over-fit the model - the heart of a top-down system is a multi-asset risk model that connects all asset classes. Most studies show that very few factors explain the lion's share of investment risk and return over longer time horizons. Security-level precision gets quickly spurious in portfolios that are multi asset, truly diversified and managed to deliver strategic outcomes. When looking over medium to long term horizons, more important than any single model is assessing the impact of different model assumptions, such as shifts in correlation, risk and return regimes.
3 -Aggregation will take place at some point in the portfolio chain, the question is where - in any top-down system, underlying investments will roll up into categories of risk factors, asset classes or portfolios. That decision on aggregation will always be unique to each investment team; generally speaking, the roll-up should align to the asset class categories and/or any risk factors used in the multi-asset modelling framework.
4 - A top-down system doesn't need to recreate the bottom-up - a bottom-up approach loses its shine when portfolios become sufficiently diversified across asset classes. At that point, broad (top-down) risk drivers become most important for determining risk and return in a portfolio. Whilst there is a tendency to want to bring in each and every security into the system, it rarely helps to inform top-down decision making.
5 - Reconciling bottom-up and top-down models provides important insight comparing the different measures of risk between bottom-up and top-down systems is a powerful exercise. However it is rare for a capital allocator to access risk measures from bottom-up tools used to manage exposures within an asset category and compare them to risk assumptions used in their top-down models. There is no reason why they can't better connect. For example, bottom-up outputs can be used as inputs into assumptions used in a top-down framework. With a better understanding, this can impact how capital is allocated and then how risk is allocated within a sleeve, such as the alpha target and discretion provided to a manager.
A challenge of investment philosophy, not technology
When it comes to bottom-up and top-down investment systems, it's not a case of preferring one or the other. Each plays a complementary role in the investment process. But bringing them together requires important considerations for an investment team that are as much a question of investment philosophy, not technology.
