Project Title: Reporting Frequency, Sample Structure, and the Effectiveness of Fair
Value Net Income as a Measure of Risk
This project examines the properties of Fair Value Net Income (FVNI) as a measure of
business risk. It also relates these properties to important implementation issues such as
measurability, reporting frequency, and cost. Our analysis is developed in the context of
bank trading portfolios, which are currently accounted for at market value and thus
represent an ideal framework for examining ‘fair value’ based risk metrics. While
explicit findings address the risks of bank trading portfolios, the issues addressed are also
relevant to proposed extensions of fair value reporting methods to broader classes of
assets and liabilities. The methodological contributions of the study include the
application of analytic and simulation methods to overcome sample limitations
encountered in conventional empirical and experimental approaches.
VAR and earlier portfolio risk measures such as ‘maturity gap’ and duration originated as
ways to reveal risks not apparent from conventional GAAP accounting methods. The utility
of these methods is evidenced by their widespread use by financial institutions, typically
going far beyond the measures now mandated by regulation. With the broad acceptance of
these methods, we can now speak of at least three distinct ‘dimensions’ of financial
information used by banking institutions: accrual method net income, cash flow, and risk
measures grounded in fair value methodology.
The introduction of fair value as an alternative basis for net income complicates the
distinctions between these three established dimensions of management information and
control. Fair value net income appears to compete directly with the accrual income concept
as the central indicator of firm performance. Risk information conveyed by fair value net
income also overlaps with results from established VAR methods. While VAR carries the
conceptual advantage of reflecting only current exposures, FVNI potentially conveys similar
information in a manner that is far less dependent on esoteric modeling processes.
The risk informativeness of FVNI, however, is highly dependent on the sample period over
which it is observed. Rate information can largely overcome limitations in FVNI sample
variability, since effects of small movements in the risk driver can be used to roughly
estimate exposure to larger changes. Unfortunately, outside the financial sector, we generally
lack explicit measures for primary drivers of business risk. For example, risks to the FVNI
of a software firm, broadly construed, have little to do with interest rates. Lacking explicit
information on risk drivers, FVNI must be observed over a sufficiently long, diverse and
commensurable historical period to reveal latent risks of the firm. Meanwhile, market
dynamics may change significantly, shortening the useful time period for measuring the
firm's current economic exposures.
Observation frequency is also a key contributor to the informativeness of FVNI. Frequent
observations allow us to infer risk from a shorter time period, helping assure our
measurements speak to the firm’s current position. In general, however, frequent
measurement and reporting is quite costly. Current GAAP net income, with its complex
rules and dependence on estimates such as depreciable life and loss reserves, probably ranks
among the more expensive of financial metrics to produce and audit. The fair value of
liquid trading portfolios is relatively easy to calculate and audit, but expansion of fair value
methods beyond their relatively narrow current domain will involve significant costs.
Consequently, the relationship between reporting frequency and the information content of
FVNI is highly relevant to the future shape and direction of accounting standards and risk
Our research will seek to address the following issues in turn:
1. Elucidate some fundamental principles regarding the relationship between sample
structure and information content, as well as the interactions between risk ‘drivers’
and stochastic outcomes.
2. Empirically assess the informativeness of FVNI streams, appropriately weighted for
contemporaneous rate movements, for risk assessment in the special case of bank
trading portfolios. We compare these results to alternative metrics such as VAR,
building on earlier work by Jorion, Liu, and Hirst.
3. Discuss some practical implications of our findings for the expansion of fair value
methods to broader areas of financial reporting and enterprise risk management.