Non-liquid income streams do not trade on exchanges, meaning you cannot look up a ticker symbol to see a current price. The value of these assets can only be determined based on future projections, probabilities, and assumptions. As such, they are prime candidates for AI.
There are many different types of non-liquid income streams, including annuities, structured settlements, music royalties, and mineral rights. Although they are diverse asset types, one thing they all have in common is that they produce income over time. However, the exact time and amount of the payments vary greatly. The pricing of these assets requires more than just a simple formula. Here is how this works.
The Asset: What Is Being Priced?
Non-liquid income streams are simply an opportunity to receive cash in the future. Annuities are great examples of this. They pay the owner a fixed dollar amount periodically (usually at the end of a time period) either for a specified number of years or for life.
Structured settlements result from a legal claim and often have a custom payout schedule. Music royalties and mineral rights will produce income in the future, but will depend upon future sales of music and mineral rights. Companies acquire these future payments in the secondary markets at a discount to the full amount they’ll receive.
Many resources can clarify how active annuity buyers fit into the structure and evaluation of these transactions. However, just knowing what your asset is does not tell you how much it is worth. This is where AI comes into play.
Cash Flow Timing Models
The first thing you need to be able to do to properly price your asset is to accurately estimate the timing of your cash flow arrival.
With fixed annuities and guaranteed cash flow, you might think the timing would be straightforward. However, even in this case, AI can be used to model other factors, such as the likelihood of early buyouts or changes to the contractual status of those buyouts, and other behavioral characteristics.
Timing for royalties is more complex than timing for fixed annuities because of fluctuations in revenue due to seasonal factors, industry trends, and macroeconomic factors beyond the control of either party. AI models, through machine learning, can examine historical patterns of payments and find hidden relationships between payments and many variables. For instance, royalties associated with streaming revenue may be related to advertising expenditures, concerts, event dates, or social media activity.
Generally, AI model outputs are generated with the assumption that the estimates of future payment activity will remain the same throughout the duration of the payment. But as new data becomes available, AI models will dynamically update the estimates of future activities of the payments.
Calculating Discount Rates with Increased Accuracy
After estimating future cash flows, they are then discounted to a present value using a discount rate that represents the risk of an investment, the opportunity costs associated with that investment, and prevailing market conditions.
Discount rates have traditionally been calculated using broad benchmarks: the yield on a U.S. Treasury security plus a risk premium. However, with advances in technology and the ability of AI systems to analyze much more detailed data, significant advances are being made in calculating discount rates using the following:
- Credit risk of the borrower
- Previous default rates of similar borrowers
- Economic data on a macro level
- Liquidity constraints within the secondary market
- Individual transaction attributes
Using machine learning, AI can create more complex non-linear relationships among these variables. For instance, a minor change in the creditworthiness of a counterparty may have an impact on the value of that counterparty’s cash flows when combined with a long-duration payment stream.
Rather than producing one homogeneous discount rate, an AI system may produce many scenarios based on historical probability.
Longevity and Mortalities
In the case of life contingent annuities, the primary concern is the mortality risk associated with the underlying annuities. The annuity may pay for many years or may cease paying due to the death of the annuitant much earlier than what was projected.
AI can take traditional mortality tables and add greater detail by providing additional data, such as:
- Trends in demographics
- Health statistics broken down by geographical region
- Advancements in medical research and treatment
- Statistical factors regarding socioeconomic status.
While the insurance industry has used actuarial science for many years, AI will allow for the enhancement of mortality tables with machine learning, providing for greater datasets and the ability to identify patterns from larger datasets.
Counterparty Risk and Performance Risk
Structured settlements and many types of annuities depend on the strength of the insurer or corporate obligor. Royalty payments depend on publications, record labels, or operating businesses.
When evaluating counterparty risk using AI, the analysis will typically look at:
- Financial statementS
- Credit ratings
- The volatility of the capital markets
- The health of the industry
- The history of litigation against the counterparty
In the royalty market, predictive models can be used to predict decreases in performance based on technological advancements or audience changes. For example, a catalog that is heavily reliant on one specific platform will have more volatility in its future cash flows if the publishing, record label, or operating business of the platform is changing.
Using AI to Deal with Data Gaps
One challenge when it comes to pricing illiquid assets is that there is not enough data available. They are not traded on a public exchange like other types of assets, so you have fewer opportunities to look at past pricing history.
AI is one option that can help you overcome this limited data by providing you with additional insights based on the data that is available.
Even though AI helps with insight, there are still limitations. If the data you input into a model is biased or incomplete, the resulting output will reflect those issues. Illiquid markets are affected by selection bias since some of the transactions are not observable or publicly available. As a result, they will always need some level of human oversight.
Valuing illiquid income streams requires an understanding of finance, law, and probability, and the ability to forecast cash flows, quantify risk, and adhere to compliance regulations.
