As the sun sets on summer, we look back fondly on vacations, sunny afternoons, and all those wonderful distractions that got us through the work day. Hopefully you also look back sentimentally on The Summer of Attribution series!
Previous articles in the series addressed challenges and concerns in investment grade, high yield, and emerging market debt. For this piece, however, I wanted to focus on a topic with which institutional asset managers, consultants, and asset owners struggle mightily: liability-driven investments or LDI.
LDI has been buzzy for quite some time, so much so that we are dedicating an entire client-led panel to it at our Americas Symposium this October. The interesting thing I observe across clients and prospects is that there is no cohesive view for how to integrate LDI into an attribution or analytical framework. Because of the nuances associated with modelling liabilities and the (perceived) difficulty in combining them with asset portfolios, LDI analysis and attribution often feels like it is overly manual and behind the times. It requires a back-to-basics approach to do it right, much like a traditional summer barbecue.
In traditional benchmark-relative investing, the objective is simple; outperform the benchmark. But LDI tends to be a bit of a special cut. Certainly, the goal of outperforming the benchmark remains the same; however, the benchmark shifts from a passive, rules-based asset universe to one composed of projected liabilities. These liabilities can be either deterministic, as seen in defined benefit plans, or they can be stochastic, as seen in more complex life insurance and annuity products.
The difficulties in LDI are varied. Duration matching or constraints, the impact of inflation on real liabilities (and knock-on impacts on asset management decisions), cash flow matching in insurance products, and additional key risk and performance indicators to monitor, such as DV01, PV01, and IE01.
The challenge in viewing LDI in an attribution and analytics context is not specific to understanding an asset portfolio in a market standard model. The challenge is in the creation and maintenance of a liability benchmark and the application of this market standard model to the asset-liability benchmark combination.
When preparing for a barbecue, whether it’s at the beach, pool, or campsite, it’s helpful to have a checklist to make sure nothing gets left out. Chips, cheese, charcoal, ketchup, swap terms and conditions…just checking to see if you were paying attention! Similarly, with LDI, a list to define all the components for inputs and outputs is essential. Missing even a single item is a recipe for disaster. Asset portfolio? Projected Benefit Obligation? Custom Yield Curve? Calculation engine? Attribution model? Output mechanism? While these are all standard pieces of the LDI puzzle, every client and process varies.
Let us assume for a moment that our asset portfolio is easy to handle (because in my experience, it typically is). Coverage across treasuries, inflation-linked securities, and credits is high, and any terms and conditions required for derivatives like interest rate or inflation swaps, interest rate futures, or currency forwards are readily obtainable. The basic ingredients on the list are checked off out of the gate and life is beautiful. Now, on to the liabilities!
This is where difficulties tend to arise, but these can be avoided by breaking the overall task into three smaller steps:
First, start with the Projected Benefit Obligation. This is merely a stream of projected cash flows required to meet the liability, either nominal or real and is typically sourced from actuarial teams or third-party consultants. I most frequently see annual cash flows, but monthly and quarterly cash flows are common as well. When the cash flows are nominal, or not indexed to inflation, I treat them as simple zero-coupon bonds, where the issue date is a valuation date and the maturity date is the date of each cash flow. Payment occurs at maturity. When the cash flows are real, or indexed to inflation, I treat them similarly; however, I also incorporate the inflation index into the security modelling.
Second, the question of what curve to use to discount these cash flows must be addressed. There is not a one-size-fits-all approach here. Treasury curves and real curves are supported out of the box, but it is also quite common for clients to load a custom curve based on some sort of credit index or treasury plus assumption. Automation occurs under each of these, but it is important to note the extra level of complexity created when using a custom curve. Here I am referring to ensuring the validity and completeness of data across all frequencies and tenors desired.
Third, streamline the calculation of underlying analytics used in performance attribution and risk analysis of the asset and liability to avoid mismatches and breakpoints like:
It’s a mess, kind of like that weird gelatin dessert that shows up at least once every summer!
Once the liability is considered, extending the standard model to incorporate key rate durations, inflation impact, and any interest rate or currency overlays tells a meaningful story. It becomes possible to see how duration mismatches are headwinds or tailwinds in a given period, or how security selection and implementation can have a meaningful impact when real liabilities come into play.
We can enhance the attribution story by adding basic visualization around point-in-time and time series analytics. As shown in the figure below, it becomes easy to illustrate duration mismatches, intermediate and long-term cash flows, and core metrics like DV01 and IE01 alongside the performance attribution.
With LDI, there tend to be a lot more moving pieces than in traditional benchmark relative asset management. As I hope to have illustrated, though, these moving pieces, when properly managed, fit well into a standard fixed income attribution model. Understanding the goal for desired analysis up front, combined with taking all components into account during preparation, allows for the simplification and scalability of what can be a difficult concept.