Historical simulations are heavily used by limited partners like pension plans and secondaries. They are helpful to model the risks, forecast the returns, and mitigate the J-curve effects associated with private equity investments.
These techniques rely on raw historical data to produce aggregate cash flow trajectories for a portfolio. The generic method consists of randomly selecting liquidated funds from a data base that have similar characteristics to the funds in the portfolio. Then, the historical cash flows of these funds are used as a proxy for a possible scenario.
For example, assume a portfolio with 1 North American VC fund and 1 European Mega buyout fund. For every simulation, the algorithm will randomly select liquidated funds that match the region and the strategy profile of the portfolio (i.e. 1 VC in North America and 1 Mega Buyout in Europe). Their historical cash flows will then be combine to produce a scenario for the whole portfolio. This will be repeated many times to produce a distribution of outcomes.
Despite its simplicity, this technique suffers from major drawbacks that reduce its usefulness for modelling the cash flows of private equity funds. We list here our top 3 reasons not to use historical simulations.
1. You are missing changes in the market and the macro-economic environment
The scenarios are constructed independently from the state of the economy. This basically means that the model completely ignores the changes in the economic environment.
For instance, the forecasts and risk measures made by the algorithm for a portfolio on the 31 of December 2019 and on the 31 of March 2020 will be extremely similar. Yet, the pandemic crisis has changed the level of uncertainty in the market and a good model should adjust to this new information to reflect this change in the environment.
2. You are missing correlations between funds
By design, the algorithm selects funds randomly from the data base to create scenarios. Therefore, the correlations between the funds are ignored although cash flows can in practice display very similar patterns and strong co-movement (e.g. no more distributions during a crisis).
Ignoring correlations has dramatic consequences from a risk management standpoint as it deflates the risk-measures of your portfolio and can create a false sense of diversification.
3. You are missing intrinsic GP and fund characteristics
The level of information granularity that can be exploited through historical simulations is limited. In its simplest form, the algorithm will only use information about the region, strategy, and vintage of the fund.
As a result, the algorithm completely misses essential data points such as the intrinsic quality of the GP or the current performance of the fund. How is the fund performing? Has it been written-up / down? When was the last distribution? Is the GP focusing its resources on the fund, e.g. is the GP fundraising? Is any of the other funds of the GP in distress?
This information carries explanatory power that can improve the precision of the forecasts.
Bonus: You are missing the use of credit facilities
Credit facilities are short-term subscription lines used to ensure smooth funding and closing of transactions. They also allow GPs to provide more visibility on the schedule of capital calls to their LPs.
Nonetheless they can have an impact on performance as they inflate the internal rate of return (IRR) of the fund and potentially allow the GP to receive carried interest more quickly. Moreover, despite being cheap, they are not free and have a negative impact on the final net money multiple of the fund.
This must be taken into consideration when modelling private equity investments.
At RockSling Analytics, we have developed alternative cutting-edge algorithms based on extensive data sets and powerful computational methods to capture these patterns and enhance the decision-making process of LPs.
Our tools can mine granular data of different types and from various sources very efficiently, allowing LPs to get deep level of insights about their portfolio. If you wish more information about RockSling Analytics or about this article, do not hesitate to contact us at harry@rockslinganalytics.com