Scenario Weighting in VaR: How It Works and When to Use It

Scenario weighting in VaR means keeping your historical scenarios but giving them unequal probabilities, usually so that recent observations count more than old ones. You then sort the weighted losses and read the VaR from the point where the cumulative weight reaches your tail probability. This guide shows the mechanics, a worked example, how to choose the decay factor, and what UK rules say about weighting schemes. It is written for finance teams that carry currency exposure.

TL;DR

• Plain historical simulation gives each of N scenarios a probability of 1/N. Scenario weighting replaces that with unequal probabilities that still sum to 1.
• The most common scheme is exponential decay: each older scenario gets λ times the weight of the one after it, so recent days dominate.
• VaR is read by sorting losses from worst to best and accumulating weights until they reach 1 minus the confidence level.
• A lower λ makes VaR react faster to volatility, but it shrinks the effective sample and adds noise. Backtest the choice.
• Weighted VaR answers "how much is a percentile loss". Stress tests answer "what happens in this specific event". Use both.

Table of Contents

• What scenario weighting means in historical VaR
• How the weights are assigned
• Choosing the decay factor
• Scenario weighting vs. filtered simulation, stress tests and stressed VaR
• What UK rules say about weighting schemes
• Applying it to a corporate FX book
• Our perspective
• See your weighted VaR on live positions
• FAQ
• Sources

What scenario weighting means in historical VaR

Historical simulation revalues today's portfolio under each past market move and collects the resulting profit and loss figures. In its plain form, each of the N observations gets probability 1/N, and VaR is the empirical tail quantile. The Federal Reserve's study of the method notes that this is equivalent to assuming returns are i.i.d. over time: a move from three years ago counts exactly as much as one from last week.

Scenario weighting keeps the same observations but changes their probabilities. Actuarial material on the technique describes choosing weights so the reweighted output meets a target VaR or tail-VaR stress while distorting the original distribution as little as possible. That is the general idea. In day-to-day treasury and risk work, the weights are almost always a simple decay schedule.

Why bother? Volatility clusters. The Federal Reserve paper makes the point that, because of this, recent returns can be more informative about today's risk than distant ones. If EUR/GBP has been jumpy for three weeks, an equal-weight window of 500 days can dilute that signal. Weighting lets the model respond without throwing away the long history that gives the tail its shape.

How the weights are assigned

The best-known scheme is the age-weighted approach from Boudoukh, Richardson and Whitelaw, often called BRW or the hybrid method. Their paper describes it as estimating percentiles directly with exponentially declining weights, which keeps the non-normal tail shape of historical simulation while giving recent data more influence.

The mechanics, as the Federal Reserve paper lays them out:

1. Choose a decay factor λ between 0 and 1.
2. Give the most recent return the largest weight. Each next older return gets λ times the weight of the one before it.
3. Scale the weights so they sum to 1.
4. Order the returns (or the portfolio P/L) and accumulate their weights up to the target percentile.

In practice, with a portfolio of FX forwards, options and cash balances, the sequence is:

1. Build your daily scenarios from the risk-factor moves (spot rates, forward points, volatilities).
2. Revalue the current portfolio under each scenario.
3. Attach each scenario's weight.
4. Sort the P/Ls from worst loss to best.
5. Add up weights from the worst loss until the total first reaches 1 minus the confidence level. The loss at that point is the VaR.

A worked example

Here is a deliberately tiny illustration, calculated from the scheme above. We have four scenarios, a decay factor of 0.5 and a 70% confidence level, so the tail probability is 30%. The scenario list starts with the most recent day.

With four observations, the normalized weights are 0.5333, 0.2667, 0.1333 and 0.0667.

Sort from worst to best and accumulate:

The cumulative weight first passes 30% at the £5m loss, so the 70% VaR is £5m. Notice what weighting did: the most recent day was a mild one with the heaviest weight, yet the worst loss still sits at the top of the ranking and sets the tail. The method does not hide history, it reorders how much each piece counts. A published walk-through of the same approach, GlobalCapital's exponentially weighted historical simulation example, follows the same steps: pick a decay factor, generate scenarios, revalue, rank the P/Ls with their weights, and read off the percentile.

Real models use hundreds of scenarios and a λ much closer to 1, such as 0.98 or 0.99. The mechanics are identical.

Choosing the decay factor

The decay factor is the one real decision in the model. It sets how fast the past fades.

• A lower λ puts most probability on the last few weeks. VaR reacts quickly when volatility rises or falls. The cost is a smaller effective sample, which can add noise to the estimate.
• A higher λ behaves more like plain historical simulation, with a stable estimate that adjusts slowly.

A useful way to think about it is the effective window. With λ = 0.99, weights halve roughly every 69 observations, so the last three months dominate while a year of history still contributes. With λ = 0.94, the half-life drops to about 11 days, which is very responsive and statistically thin.

There is evidence the effort pays off. In the BRW paper's S&P 500 example, the hybrid method's absolute error in estimating 1% VaR was 30% to 43% lower than the exponentially weighted variance approach and 14% to 28% lower than ordinary historical simulation. That is one dataset and one period, so treat it as a reason to test weighting on your own currency pairs, not as a guarantee.

To settle on λ, backtest the whole setup. The Bank of England's working paper on historical-simulation models says backtesting should cover the complete model and its recalibration strategy. It describes the Kupiec, Christoffersen and mixed-Kupiec tests and recommends testing several portfolios, confidence levels and stressed historical periods. For FX books, our own guidance on 95% backtesting rules for FX VaR modeling walks through the exception counting side.

Scenario weighting vs. filtered simulation, stress tests and stressed VaR

These terms get mixed up in search results, so here is how they separate.

Filtered historical simulation (FHS) is a different fix for the same problem. The Bank of England describes it as devolatilising each historical return with its contemporaneous volatility estimate, then revolatilising the residuals with current volatility. Weighting changes how often a scenario is counted, while FHS changes the scenario itself. Some teams combine them.

Stress testing is complementary, not a substitute. MSCI's paper on stress testing explains that stress tests tie losses to an identifiable event, while VaR summarizes a percentile of a distribution, so a weighted VaR should sit next to scenario analysis. Methods for building those scenarios, such as pessimised scenarios, use the loss contributions of extreme historical scenarios to find which risk factors drive severe losses, and they call for expert review rather than purely mechanical shocks. For the treasury version of this, see our piece on FX stress scenarios that reveal funding gaps.

Stressed VaR fixes the observation window on a stressed period instead of reweighting. The two ideas can coexist, which brings us to the UK rules.

What UK rules say about weighting schemes

Most corporates do not hold regulatory capital against a VaR model, so these rules are not your obligation. They are still the clearest public statement of what a supervisor considers a prudent weighting scheme, which makes them a good benchmark for internal policy.

In the FCA Handbook version dated December 2021, the VaR-model rules in BIPRU 7.10 say that when a firm applies a weighting scheme, the effective observation period must be at least one year and the weighted average lag at least six months. The firm must also adjust the scheme if it produces an imprudent VaR. The same section requires a firm that uses a VaR model for regulatory capital to hold VaR-model permission and to meet standards on the model, risk management, stress testing, backtesting, capital calculation and reporting.

The FCA's technical standard on general assessment also lists what supervisors look at, including the VaR methodology, time-series length, stressed-period methodology, nonlinearities, correlation and basis risk, proxies and event risk. For stressed VaR, the PRA's supervisory statement SS13/13 says it should be based on a continuous 12-month period of significant financial stress relevant to the portfolio, reviewed at least annually.

Rules in this area are revised over time, so check the current FCA and PRA texts before you build anything that must satisfy a regulator.

A practical reading for a corporate team: do not let λ get so small that the effective window collapses below a year, and write down how often you review the choice.

Applying it to a corporate FX book

A company with EUR, USD and PLN or SEK payables and receivables has a portfolio that is mostly linear (forwards, natural exposures) with some options. That makes full revaluation under weighted scenarios cheap and transparent. A sensible setup:

1. Define exposures by currency, with settlement dates and hedge positions.
2. Build daily scenarios from at least two to three years of rate history, so a high λ still has depth.
3. Pick a λ in the 0.97 to 0.99 range as a starting point and compare against equal weights.
4. Report both numbers. The gap between weighted and unweighted VaR is a signal: when weighted VaR is well above unweighted, recent volatility is above the long-run level.
5. Backtest quarterly and log exceptions.
6. Pair with stress scenarios such as a 10% or 20% move in your main pair, and check cash and covenant headroom.

If you want more depth on each piece, our guides cover daily historical VaR for FX books with full revaluation and EWMA, the Value at Risk calculation workflow and VaR for corporates. Once you have a number you trust, the next question is what to hedge, which we cover under hedging based on value at risk.

Our perspective

We built Corphedge for companies that need to see their currency risk the way a bank desk does, without running a bank's infrastructure. Scenario weighting is a good example of why. The idea is simple. The hard part is doing it every day, on every position, with consistent data and a record of what you changed and why.

Our view is that a VaR number should never sit alone. It belongs next to live exposure, hedge ratios and stress scenarios, so the CFO and the treasurer read the same picture. That is how we designed the platform: real-time position tracking, VaR-based strategy simulation, portfolio reporting and automated notifications in one place, with expert advice for the judgment calls a model cannot make.

We also think weighting choices deserve the same discipline as hedging decisions: pick a parameter, document it, test it, and revisit it on a schedule. Teams that do this get a risk figure they can defend to a board, an auditor or a lender.

Bartas

See your weighted VaR on live positions

The Corphedge platform brings together currency position tracking, hedging strategy simulation, Value at Risk analysis, live market data and portfolio reporting for internationally active companies. Our operational expert advice helps with the choices around it, and the Corpay integration connects the analysis to execution. We are also expanding to Poland and Sweden, so teams with PLN or SEK exposure can expect more local support.

See how the Corphedge platform works for your exposures. If you want to build the methodology skills in your own team first, our FX hedging course is a one-off purchase at €220.

FAQ

Is scenario weighting the same as exponentially weighted VaR?

Not exactly. Exponential weighting is the most common way to do it, but scenario weighting is the broader idea of assigning unequal probabilities to historical scenarios. In the BRW method the weights decline by a factor of λ for each step back in time. Other schemes choose weights to hit a target stress with minimal distortion.

Which decay factor should I use?

There is no universal value. Higher values such as 0.98 or 0.99 keep a long effective history and give a stable estimate. Lower values react faster but add noise. Choose by backtesting against several portfolios and confidence levels, and keep the effective window consistent with your risk policy.

Does weighting make VaR higher or lower?

It depends on recent conditions. If recent days were calmer than the older history, weighting can lower the estimate. If recent days were turbulent, it can raise it. The size of the gap between weighted and unweighted VaR is itself a useful signal.

Do UK corporates have to follow the FCA weighting rules?

The FCA VaR-model rules apply to firms that use a VaR model for regulatory capital and hold the relevant permission. A corporate treasury typically does not, but the rules, such as a minimum one-year effective observation period, are a reasonable benchmark for internal policy.

Should I still run stress tests if I weight my scenarios?

Yes. Weighted VaR gives a percentile of a distribution, while stress tests show the loss in a specific event. They answer different questions and work best together.

Sources

• The Hidden Dangers of Historical Simulation, Federal Reserve
• Efficient Sensitivity Analysis via Scenario Weighting
• The Best of Both Worlds: a hybrid approach to calculating Value at Risk
• Computing VaR using exponentially weighted historical simulation, GlobalCapital
• Historical and Weighted Historical Simulation
• Bank of England Working Paper No. 525
• Stress Testing in the Investment Process, MSCI
• Building pessimised scenarios, Risk Magazine
• FCA Handbook, BIPRU 7.10 Use of a Value at Risk Model
• FCA Handbook, Article 9 General assessment standards
• PRA SS13/13 Market risk, Bank of England

Recommended

• 95% Backtesting Rules: FX VaR Modeling for Risk Analysts
• Streamline your Value at Risk workflow for FX risk
• Daily Historical VaR for FX Books: Full Revaluation, EWMA, Backtests
• VaR hedging: effective currency risk mitigation guide