Why a 10-Day Horizon and 99% Confidence Level Dictate Trillions in Bank Capital Reserves

On September 18, 2023, the Federal Register published a 1,089-page joint proposal by U.S. banking agencies to overhaul how large institutions calculate their regulatory capital. At the core of this framework—and the global Basel standards it implements—is a mathematical boundary: the requirement to measure market risk over a 10-day liquidity horizon at a 99 percent confidence level. This specific statistical pairing dictates exactly how many billions of dollars a bank must lock away in reserve rather than deploy into the market.[2]

The mechanism driving this calculation is Value at Risk (VaR), a metric that became the industry standard for quantifying financial risk. A 99 percent VaR over a 10-day horizon means a bank calculates the maximum amount it could lose over two trading weeks, such that it is 99 percent certain its actual losses will not exceed that number.[4]

As the quantitative analytics firm Quantt defines the metric, "Value at Risk summarizes the worst expected loss over a target horizon within a given confidence interval." If a trading desk has a 10-day, 99 percent VaR of $50 million, the model predicts that only one time in 100 will the desk lose more than $50 million over a two-week period.[4]

The 10-day horizon is not an arbitrary time limit. It represents the estimated duration a bank would need to liquidate a standard trading book during a period of severe market stress without triggering a fire sale that crashes prices further.

According to a 2014 analysis by the IEB on measuring market risk under the Basel Accords, this holding period assumes that market liquidity dries up during a crisis. Regulators force institutions to hold capital against a 10-day window because exiting complex positions instantaneously is impossible when buyers disappear.

However, the 99 percent VaR model contains a structural blind spot that became glaringly apparent during historical market crashes. The model identifies the threshold of the worst 1 percent of outcomes, but it provides zero information about what happens beyond that threshold.[3][4]

If a catastrophic event occurs—the 1-in-100 scenario—VaR does not calculate whether the loss will be $51 million or $5 billion. To address this unmeasured "tail risk," the Basel Committee introduced a new metric called Expected Shortfall (ES) as part of the Fundamental Review of the Trading Book (FRTB).[1][3]

The Bank Policy Institute noted in a May 23, 2023, analysis that Expected Shortfall fundamentally changes the mathematical question. Instead of asking for the minimum loss in the worst 1 percent of cases, ES asks for the average loss across all of those worst-case scenarios.[3]

Under the FRTB framework detailed by the Bank for International Settlements (BIS) in standard MAR33, the confidence level for Expected Shortfall is set at 97.5 percent, rather than the historical 99 percent used for VaR.[1]

While lowering the confidence level from 99 percent to 97.5 percent sounds like a relaxation of capital standards, the mathematical reality is the opposite. Averaging the worst 2.5 percent of outcomes captures extreme tail events that a simple 99 percent VaR threshold cuts off, generally resulting in a higher capital requirement.[1][3]

Sullivan & Cromwell’s March 1, 2026, memorandum on bank regulatory capital highlights that these internal models directly determine the denominator of a bank's capital ratios. The higher the calculated risk, the more tier-one capital the bank must hold against its trading assets.

The Federal Register rule proposed applying these stringent market risk capital requirements to banking organizations with significant trading activity, fundamentally altering the cost of maintaining large trading desks and holding complex inventory.[2]

The 10-day horizon itself is also being refined and expanded. Under the FRTB, the BIS requires banks to scale liquidity horizons based on the specific asset class, acknowledging that not all securities can be sold in two weeks.[1]

The revised framework applies a baseline 10-day horizon for highly liquid large-cap equities and sovereign bonds, but extends the horizon up to 120 days for complex credit derivatives and structured products.[1]

This granular approach prevents banks from assuming they can offload illiquid assets in the same two-week window as government bonds, forcing them to hold substantially more capital against hard-to-sell positions.[1]

The transition from VaR to Expected Shortfall requires massive computational power. Banks must now simulate thousands of historical market shocks, recalculating the value of their entire trading portfolio under each scenario to find the average of the worst 2.5 percent.[3][4]

The Bank Policy Institute argues that the design of the FRTB Expected Shortfall calculation, while theoretically superior at capturing tail risk, introduces significant operational complexity and potential volatility in daily capital requirements.[3]

The final implementation of these rules dictates the capacity of global markets. When the capital required to hold a specific asset rises due to a longer liquidity horizon or a severe Expected Shortfall projection, banks naturally reduce their inventory of that asset, shifting risk to non-bank financial institutions.[2]