Private equity: opportunities and challenges in internal bank investment risk measurement

Participations as an asset class for development banks

State funds are not sufficient to finance the urgently needed investments in important fields of action such as energy and transportation networks, security and digital infrastructure in Germany and Europe. Mobilizing private capital for this is a contribution that promotional banks should and can make. KfW can effectively leverage this with equity investments and as an anchor investor in funds, among other things [see Börsenzeitung 31.12.2024, p. 6-7]. However, increased investment in the equity investment business also requires adequate risk assessment procedures in order to take account of the nature of the equity investment business (increased risk/return profile with a direct P&L impact) in overall bank management.

In terms of methodology, the management of investment risk requires a paradigm shift from models that predict the probability of default on a one-year horizon focused on the lending business to models that appropriately reflect the return and risk over the respective investment horizon.

What can such a model with a value-based perspective look like in practice in a (development) bank and what challenges arise in its methodological design? This is illustrated below using the scoring procedure for investment funds, which was newly introduced in 2020 and further developed in 2024.

What is the scope of the model?

The scoring focuses on closed-end funds that are set up and managed by fund managers and in which the investors participate on the liabilities side with equity (unstructured/structured). The funds raised are invested on the assets side in equity investments in various, mostly unlisted portfolio companies, whereby the fund can also provide mezzanine financing.   The contractual term of 10 years is often supplemented by optional extensions, so that the life of the fund can be up to 15 years until its final settlement.

The model supports the investment decision and is used to differentiate risk when determining the economic capital.

What does the model structure look like?

At the heart of the model are a quantitative module and an early phase module. The latter was the focus of the 2024 further development. Both modules forecast the fund return at the end of the fund term using the key figure DPI (distributions to paid-in capital): The quantitative module developed on external data uses realized cash flows and the early-stage module developed on external and internal data is based on fund master data (e.g. fund size) and a structured assessment of management quality using a list of questions on structure and stability, team (quality and quantity), identity of interests and track record. The interaction of the individual modules is shown in Figure 1, whereby the management quality (MQ) score in the early phase module is converted into a DPI (T) forecast using a linear function. The derivation of the linear function is based on the assumption that the DPI(T) forecast for mature funds is sufficiently accurate and can therefore serve as an approximation for the true, i.e. final DPI(T) of the funds. The continuous DPI(T) forecasts are mapped into discrete intervals after taking country risks into account and assigned scores. This increases comprehensibility and enables further use (consistent with the lending business) in other business processes (such as investment decisions or limits). As a result, a scoring system was developed with seven performing characteristics and a score for default.

Why was DPI chosen as the target size?

In the literature and in practice, performance measures that are based on the expected cash flows (payments and receipts), in some cases without taking into account any transaction costs over the term of the fund, are primarily used for valuation models.  These include the IRR (Internal Rate of Return) and its modifications MIRR (Modified Internal Rate of Return) and MIC (Multiple of Invested Capital). These have weaknesses, so their use as a target figure was rejected. The background to this is that the IRR can be significantly distorted by strategic cash flow management, e.g. the IRR can be artificially increased by delayed investments, premature distributions or borrowing [see MSCI, 2024]. For risk modeling, on the other hand, the use of measures such as DPI or TVPI (Total Value to Paid-In Capital) with regard to the actual cash flows has the advantage of objectivity and consideration of costs. Accordingly, the realized cash flows of finally settled or liquidated funds are used as the target figure for modelling. This means that the DPI corresponds to the TVPI.

How can unrepresentative external data be handled?

The model was developed on the basis of external data, as too few finally calculated internal funds were available. The data from the external data providers are predominantly from North America or have older vintages. Their distributions differ from the internal data.  Possible options for dealing with non-representativeness are 1) stratified sampling generates a sample that is representative for the application, but reduces the amount of data. 2) Checking the relevance of the affected factor in a multivariate model: If it has no influence on the target variable and the value range of the development data covers that of the application data, non-representativeness is acceptable. 3) Appropriate modeling: e.g. linear integration of metric factors with a linear relationship over the entire range of values or segmentation in the case of categorical variables.

The current model handles this as follows: Outliers are specifically removed (option 2). Sampling (option 1) was tested but discarded in favor of appropriate mapping. Weaknesses (e.g. investment focus) are documented and flanked by analyses or model risk buffers (option 3). However, the aim is to calibrate the model on the internal data basis with increasing quality and quantity.

Paradigm shift completed, integration into other control processes is progressing

With the value-based model, KfW has completed the paradigm shift from a one-year default view to a value-based view and, with equity fund scoring, uses a lifecycle-based, value and return-oriented view with modular forecasting and a clear governance framework. This results in a large number of opportunities, in particular:

> Supporting the investment decision for the market or market sequence at individual transaction level

> Significant improvement in adequate investment risk measurement for overall bank management (e.g. portfolio limitation)

> Closer integration with other financial and performance indicators as well as synergy potential for possible further developments in the individual risk measurement of investments

What is next for the next model cycle? The feedback from the investment experts from the first rating cycle was very positive and the opportunities for improvement identified were successfully implemented in the subsequent further development. Further impetus is expected from the internal data generated over time.