Time horizon is a critical factor in VaR calculations because it determines the period over which potential losses are assessed. A short time horizon might show lower potential losses compared to a longer horizon, as it considers a smaller window of market fluctuations. This difference influences risk interpretation; a 1-day VaR will provide insight into immediate potential losses, while a longer-term VaR might better capture longer-term risks and systemic factors. The choice of time horizon should align with the risk management objectives and the nature of the portfolio .

The non-sub-additive property of VaR implies that the VaR of a combined portfolio can exceed the sum of the VaRs of individual portfolios. This characteristic can discourage diversification, as combining two "safe" portfolios into one might result in a higher risk measure than initially expected. This could lead to counterintuitive risk management decisions, where maintaining separate portfolios might appear safer than merging them. Additionally, it highlights the potential for underestimation of risk when applying VaR to complex financial instruments and multi-asset portfolios .

Historical VaR is effective in utilizing actual past market data to estimate potential future losses, making it relatively straightforward to implement. However, it assumes that past market scenarios are indicative of future conditions, which might not capture unprecedented market events. Monte Carlo VaR, on the other hand, uses simulations to create a wide range of potential future market scenarios, offering a more robust approach to capturing tail risks and complex instruments. It is computationally intensive and requires extensive modeling expertise, which can be a challenge to implement effectively .

During the 2008 financial crisis, correlations between assets that had previously been low or negative suddenly increased, rendering diversification strategies less effective. VaR models, which were based on historical correlations and volatilities, failed to predict the simultaneous downturns in different markets and sectors. This unexpected change in correlations led to substantial underestimation of risk and higher losses than anticipated. The crisis highlighted the inability of VaR to adapt quickly to rapid changes in market conditions, emphasizing its limitation as a risk management tool during periods of extreme stress .

Backtesting VaR models with historical data from periods of financial stress is important to validate the accuracy and reliability of the models under extreme conditions. These tests help identify potential weaknesses in the model's assumptions and provide insights into how well the VaR estimates reflect actual market behavior. By including stressed periods, such as the 2008 financial crisis, in the backtesting process, financial institutions can ensure that their VaR models account for the volatility spikes and correlation changes experienced during market downturns, thus making them more robust for future uncertainties .

The choice of VaR model parameters acts as a risk factor because it influences the accuracy and reliability of the VaR estimates. Different models and parameters can yield significantly different VaR values for the same portfolio, leading to potential misjudgments in risk management. Changing market conditions, like the sudden rise in correlations and volatility during a crisis, can render the traditional model parameters insufficient, leading to underestimated potential losses .

Liquidity concerns are incorporated into VaR calculations through Liquidity-Adjusted VaR (LVaR), which extends the traditional VaR model by accounting for the cost and potential difficulty of liquidating large positions quickly. It adjusts the VaR estimate to reflect the increased risk of not being able to sell positions at expected prices during market stress. LVaR considers factors like market depth, trading volume, and bid-ask spreads, providing a more realistic picture of potential losses in illiquid markets .

Expected Shortfall (ES), also known as Conditional Value at Risk (CVaR), addresses a critical limitation of VaR by considering not just the threshold of potential losses but the average of all losses beyond that threshold. While VaR provides the potential loss that will not be exceeded with a certain probability, it does not specify how much can be lost if this threshold is exceeded. ES offers a more comprehensive measure by focusing on the tail end of the loss distribution, making it a better indicator of extreme risks in a portfolio .

Diversification is crucial in portfolio VaR calculations because it reduces the overall risk exposure through imperfect correlations among assets. When the correlation between assets is less than 1, a diversified portfolio is typically safer than the sum of individual asset risks. However, during financial crises, asset correlations tend to rise towards 1, significantly reducing the benefits of diversification. This was evident during the financial crisis when diversified portfolios suddenly appeared riskier, and portfolio VaR calculations based on previous data became inaccurate, leading to underestimated risks .

The primary limitation of Value at Risk (VaR), as revealed by the trading incident on August 24, 2015, is that it does not measure extreme tail risk. VaR only covers normal market conditions and does not account for the size of losses beyond a certain percentile cutoff. In this incident, the trader's VaR was $2.3 million, but he ended up losing $120 million due to the market dropping 11% in 8 minutes, highlighting that VaR assumptions are invalid during crashes when correlation and volatility spike .