Testing Density Forecasts, with Applications to Risk Management

By Berkowitz, Jeremy | Journal of Business & Economic Statistics, October 2001 | Go to article overview
Save to active project

Testing Density Forecasts, with Applications to Risk Management


Berkowitz, Jeremy, Journal of Business & Economic Statistics


The forecast evaluation literature has traditionally focused on methods of assessing point forecasts. However, in the context of many models of financial risk, interest centers on more than just a single point of the forecast distribution. For example, value-at-risk models that are currently in extremely wide use form interval forecasts. Many other important financial calculations also involve estimates not summarized by a point forecast. Although some techniques are currently available for assessing interval and density forecasts, existing methods tend to display low power in sample sizes typically available. This article suggests a new approach to evaluating such forecasts. It requires evaluation of the entire forecast distribution, rather than a scalar or interval. The information content of forecast distributions combined with ex post realizations is enough to construct a powerful test even with sample sizes as small as 100.

KEY WORDS: Densities; Evaluation; Forecasting; Risk Management.

Although the forecast evaluation literature has traditionally focused on point forecasts, many models in economics and finance produce forecasts that cannot be easily summarized by a point forecast. For example, the widely used value-at-risk (VaR) approach to quantify portfolio risk delivers interval forecasts (see Jorion 1997 for a recent survey). These models are used to assess corporate risk exposures and have received the official imprimatur of central banks and other regulatory authorities. In 1997, a Market Risk Amendment to the Basle Accord permitted banks to use VaR estimates for setting bank capital requirements related to trading activity.

Many other important financial calculations involve estimates not summarized by a single point on the forecast density. For example, the Standard Portfolio Analysis of Risk (SPAN) system, first implemented by the Chicago Mercantile Exchange in 1988, has become a very widely used approach to calculate margin requirements for customers and clearing-house members. SPAN is essentially a combination of stress tests performed on each underlying instrument in the relevant portfolio (see Artzner, Delbaen, Eber, and Heath 1999).

To evaluate the performance of such models and their forecasts, regulators and financial institutions must be able to compare the forecasts to subsequent outcomes. The most familiar such exercise is verifying that the model accurately delivers a given interval forecast. Early approaches to this problem were handled as if the upper end of the interval was a point forecast--the number of times the interval was exceeded is compared to the expected number.

However, Christoffersen (1998) emphasized that, being interval forecasts, there is more information in interval forecasts than a point. To see why, note that even if the model delivers the correct average coverage, it may not do so at every point in time. For example, in markets with persistent volatility, forecasts should be larger than average when volatility is above its long-run average and vice versa. Christoffersen proposed methods for evaluating interval forecasts. These methods go part of the way toward addressing the critique of Kupiec (1995), who argued that very large datasets are required to verify the accuracy of such models. Users typically set the VaR level deep in the tail of the distribution (the Accord stipulates a 99% level). Violations are thus expected to occur only once every 100 days. Simply counting the number of violations per year obviously uses very little of the data. Christoffersen suggested making greater use of the violations by noting that violations should occur 1% of the time and should not be bunched together--violations should be conditionally unpredictable. Nevertheless, interval evaluation methods remain quite data intensive since they only make use of whether or not a violation occurs (e.g., Lopez 1999).

If risk models are to be evaluated with small samples, a more general perspective on model performance is required.

The rest of this article is only available to active members of Questia

Sign up now for a free, 1-day trial and receive full access to:

  • Questia's entire collection
  • Automatic bibliography creation
  • More helpful research tools like notes, citations, and highlights
  • Ad-free environment

Already a member? Log in now.

Notes for this article

Add a new note
If you are trying to select text to create highlights or citations, remember that you must now click or tap on the first word, and then click or tap on the last word.
Loading One moment ...
Project items
Notes
Cite this article

Cited article

Style
Citations are available only to our active members.
Sign up now to cite pages or passages in MLA, APA and Chicago citation styles.

Cited article

Testing Density Forecasts, with Applications to Risk Management
Settings

Settings

Typeface
Text size Smaller Larger
Search within

Search within this article

Look up

Look up a word

  • Dictionary
  • Thesaurus
Please submit a word or phrase above.
Print this page

Print this page

Why can't I print more than one page at a time?

While we understand printed pages are helpful to our users, this limitation is necessary to help protect our publishers' copyrighted material and prevent its unlawful distribution. We are sorry for any inconvenience.
Full screen

matching results for page

Cited passage

Style
Citations are available only to our active members.
Sign up now to cite pages or passages in MLA, APA and Chicago citation styles.

Cited passage

Welcome to the new Questia Reader

The Questia Reader has been updated to provide you with an even better online reading experience.  It is now 100% Responsive, which means you can read our books and articles on any sized device you wish.  All of your favorite tools like notes, highlights, and citations are still here, but the way you select text has been updated to be easier to use, especially on touchscreen devices.  Here's how:

1. Click or tap the first word you want to select.
2. Click or tap the last word you want to select.

OK, got it!

Thanks for trying Questia!

Please continue trying out our research tools, but please note, full functionality is available only to our active members.

Your work will be lost once you leave this Web page.

For full access in an ad-free environment, sign up now for a FREE, 1-day trial.

Already a member? Log in now.

Are you sure you want to delete this highlight?