What Is Value at Risk? Starting From the Very Basics
Imagine you’re about to drive from Mumbai to Pune. Before you leave, you check the weather app, and it tells you: “There is a 95% chance you will reach within 3.5 hours.” It doesn’t promise you’ll never hit traffic. It doesn’t tell you what happens on the unlucky 5% of days when there’s a landslide or an accident that shuts the highway for six hours. It simply gives you a threshold you can plan around, for the vast majority of ordinary days.
Value at Risk, almost always shortened to VaR, does the same thing for money instead of time. It answers one very specific, very practical question that every bank, fund manager, and trading desk asks every single day:
“On a normal bad day, how much money could I realistically lose?”
That’s it. That’s the entire idea at its core. Everything else in this article is just detail on how that one question gets answered precisely.
The Formal Definition of VaR
Once you understand the driving analogy, the textbook definition becomes easy to read instead of intimidating.
Value at Risk (VaR) is a statistical measure that estimates the maximum potential loss a portfolio could suffer over a specific time period, at a given confidence level, under normal market conditions.
Written as a formula, a VaR statement looks like this:
VaR (Confidence Level %, Time Horizon) = ₹ X
Example:
1-Day VaR (99%) = ₹50,00,000
Meaning: “There is a 99% chance that this portfolio
will NOT lose more than ₹50,00,000 in one trading day.”
Equivalently: “There is only a 1% chance that losses will EXCEED ₹50,00,000 in one trading day.”
Notice something important already: VaR is always a single number, in currency terms, describing a threshold. It never describes an average loss or a typical loss. It describes a boundary line that separates “normal bad days” from “unusually bad days
Breaking Down a VaR Statement: Three Moving Parts
Every VaR number is meaningless without three pieces of context sitting alongside it. A risk manager who says “our VaR is ₹50 lakh” without specifying the other two components hasn’t actually said anything useful yet.
1. The Loss Amount (the ₹ figure itself)
This is the threshold in currency terms — the answer to “how much.”
2. The Confidence Level
This tells you how strict the boundary is. Common levels used across the industry are 95%, 99%, and occasionally 99.9%. A 99% confidence level is stricter than 95% — it pushes the boundary further out, capturing rarer, larger losses within the “normal” zone.
3. The Time Horizon
This tells you over what period the loss could occur — one day, ten days, or one month. A trading desk that turns over its portfolio daily typically uses a 1-day VaR. A bank calculating regulatory capital often uses a 10-day VaR, because regulators assume it could take that long to safely exit a large position in a stressed market.
Worked Example
Scenario: A bank’s trading desk holds a bond portfolio worth ₹10 crore. The risk team calculates:
1-Day VaR (99%) = ₹35,00,000
How to read this correctly: Under normal market conditions, there is a 99% probability that the desk will not lose more than ₹35 lakh in a single trading day.
Put differently, on roughly 1 out of every 100 trading days (about 2-3 days a year), the desk could reasonably be expected to lose more than ₹35 lakh —
VaR simply does not tell you how much more. This last point is the single most misunderstood — and most important — idea in this entire article. Keep it in mind; we return to it in the disadvantages section below.
Why Is VaR Used in Risk Management?
VaR became the dominant risk metric across global banking for four practical reasons:
- A common language across the business. A single number lets a CEO, a board member, and a trader all discuss risk using the same unit — currency — instead of technical statistics that only quants understand.
- Capital allocation. Banks use VaR to decide how much capital to set aside against potential losses. Regulatory frameworks (including the Basel capital adequacy rules) have historically used VaR-based calculations as an input to market risk capital requirements.
- Setting and monitoring risk limits. Trading desks are assigned VaR limits — a ceiling they cannot exceed without approval. This keeps individual traders from taking on hidden, oversized risk.
- Comparability across very different assets. A bond desk, an equity desk, and a currency desk hold completely different instruments. VaR converts all of them into one comparable currency-denominated number, making it possible to compare — and aggregate — risk across an entire institution.
Key Properties and Characteristics of VaR
- It is a quantile-based measure. VaR is essentially picking a specific point on a loss probability distribution — the point beyond which only the chosen “tail probability” (1% or 5%) of outcomes fall.
- It depends entirely on three chosen parameters. Change the confidence level, the time horizon, or the underlying statistical assumption about how returns behave, and the VaR number changes — even though the actual portfolio hasn’t changed at all. VaR is a modelling choice, not a fixed physical property of a portfolio.
- It scales with time — under assumptions. A commonly used shortcut, the “square-root-of-time rule,” estimates a 10-day VaR by multiplying the 1-day VaR by the square root of 10. This is a simplification that assumes returns are independent and identically distributed day to day — an assumption that often breaks down during real market stress.
- It is portfolio-level and aggregable. VaR can be calculated for a single position, a desk, or an entire firm, and (with appropriate correlation adjustments) can be rolled up from smaller units to larger ones.
- It is silent beyond the threshold. This is discussed in detail below, but it’s worth stating as a defining property: VaR tells you the boundary, not what lies beyond it.
Advantages of VaR
- Easy to explain to non-technical stakeholders — a single rupee figure
- Enables consistent risk limits and internal controls across an organisation
- Allows comparison of risk across very different asset classes and desks
- Forms a recognised basis for regulatory capital calculations
- Can be calculated using several different methods depending on data availability (covered below)
Disadvantages and Drawbacks of VaR
This is where a genuinely useful understanding of VaR begins — knowing exactly where it breaks down.
1. It says nothing about the size of losses beyond the threshold
This is VaR’s most quoted weakness. Two portfolios can have an identical 99% VaR of ₹35 lakh, yet one might lose ₹36 lakh on a bad day while the other loses ₹3.5 crore. VaR treats both scenarios identically because it only measures the boundary, not what’s beyond it. This is precisely why a related measure, Expected Shortfall (also called Conditional VaR or CVaR), was developed — it answers “given that we crossed the VaR threshold, what is the average loss we should expect?” We’ll cover Expected Shortfall in a dedicated article later in this series.
2. VaR is not a “coherent” risk measure — it can violate diversification logic
In risk theory, a sound risk measure should never say that combining two portfolios increases risk beyond the sum of their individual risks (this property is called sub-additivity). VaR can violate this rule in certain cases, particularly with instruments that have unusual, non-linear payoff patterns (like options). This means VaR can occasionally penalise diversification instead of rewarding it — a serious theoretical flaw for a risk measure.
3. It relies heavily on assumptions and historical data
Every VaR calculation method makes assumptions — about how returns are distributed, or about how representative recent history is of future risk. When markets behave in ways that haven’t been seen in the historical window used, VaR can dramatically understate real risk. This was a widely discussed criticism following the 2008 financial crisis, when many banks’ VaR models had not been calibrated to the kind of extreme, correlated losses that actually occurred.
4. Different calculation methods can give meaningfully different answers
As we’ll see below, there isn’t one single way to calculate VaR. The same portfolio, on the same day, can show different VaR figures depending on which of the three standard methods is used — which can create a false sense of precision if the underlying methodology isn’t well understood.
5. It can create incentives to “hide” risk in the tail
Because VaR only measures a threshold, there is a structural temptation (sometimes unintentional) to build portfolios that look safe by this measure but carry a small probability of very severe losses — a pattern sometimes described as “picking up pennies in front of a steamroller.”
6. It assumes “normal” market conditions
The definition itself includes this caveat, but it’s easy to forget in practice: VaR describes risk in ordinary market conditions. It is not designed to capture — and typically does not capture well — the risk of genuine crisis periods, which is why banks pair VaR with separate stress testing and scenario analysis frameworks.
The Three Methods of Calculating VaR
Everything above describes what VaR measures and why it’s used — but not how the number is actually calculated. In practice, there are three widely used methods, and each makes different trade-offs between simplicity, accuracy, and the assumptions it relies on:
1. The Historical Simulation Method
Calculates VaR directly from actual past returns of the portfolio, with no assumption about the shape of the return distribution.
2. The Variance-Covariance (Parametric) Method
Assumes portfolio returns follow a known statistical distribution (typically normal/Gaussian) and calculates VaR using the portfolio’s volatility and correlations directly through a formula.
3. The Monte Carlo Simulation Method
Generates thousands of random, computer-simulated future scenarios based on statistical models of how the underlying assets behave, then derives VaR from the simulated outcomes.
Each method has distinct strengths, weaknesses, data requirements, and suitability depending on the portfolio being measured. We will dedicate one full article to each of these three methods next in this series, working through the mechanics, formulas, and worked examples for all three.
Frequently Asked Questions
Is VaR the maximum possible loss on a portfolio?
No — this is the most common misunderstanding. VaR is a threshold that is expected to be exceeded a small, defined percentage of the time (1% at a 99% confidence level). It is not a worst-case number.
What confidence level is most commonly used?
95% and 99% are the most widely used in practice. Regulatory capital calculations have historically leaned toward higher confidence levels such as 99%.
Is VaR still relevant given its drawbacks?
Yes — despite its limitations, VaR remains one of the most widely used risk metrics globally because of its simplicity and comparability. Most institutions today use it alongside complementary measures like Expected Shortfall and stress testing rather than relying on it alone.
What’s the difference between VaR and Expected Shortfall?
VaR tells you the threshold loss at a given confidence level. Expected Shortfall (Conditional VaR) tells you the average loss given that the VaR threshold has already been breached — addressing VaR’s blindness to tail severity.
https://decode-finance.com/category/market-risk-credit-risk-and-operational-risks