Quantitative Research

Understanding Option Volatility: Why HAR-RV Outperforms Standard HV30 for Volatility Risk Premium (VRP)

In quantitative options trading, evaluating an option's expected value requires looking beyond traditional technical analysis and focusing strictly on the Volatility Risk Premium (VRP). Options are decaying financial assets priced on expected future variance — and the model you use to measure realized variance changes the answer.

When selling premium, your core statistical edge relies on implied volatility (what the options market expects) overstating realized volatility (how much the underlying stock actually moves). If you sell options when the VRP is negative — meaning options are cheap relative to physical market movement — you face a structural disadvantage over time, regardless of directional accuracy.

To measure this effectively, traders cross-reference implied volatility against historical realized movement. However, the method used to calculate realized volatility makes a substantial difference. This article covers the metrics that define a volatility regime, why the standard 30-day historical volatility (HV30) estimator falls short, and how the HAR-RV model fixes it.

Key metrics for analyzing volatility regimes

Before comparing realized volatility models, it helps to review the primary metrics used to assess whether options are pricing in an attractive risk premium:

The limitations of standard HV30

Most retail platforms evaluate realized volatility using a simple 30-day historical volatility calculation (HV30). While easy to compute, HV30 has notable limitations as a forward-looking tool:

• Equal weighting — HV30 treats every day in the 30-day window identically. A market shock from 29 days ago carries the exact same mathematical weight as price action from yesterday.
• Lag in regime shifts — because it relies on an unweighted lookback, HV30 reacts slowly when market volatility suddenly spikes or calms down.
• The "drop-off effect" — when a single large volatile day reaches day 31, it abruptly falls out of the calculation, causing an artificial drop in the HV30 reading even if market conditions remain unchanged.

To model volatility more accurately, traders need an estimator that accounts for memory in financial markets and acknowledges that recent price shocks carry more signal than older ones.

The Heterogeneous Autoregressive (HAR-RV) model

To address the shortcomings of simple historical lookbacks, quantitative research frequently utilizes Fulvio Corsi's HAR-RV (Heterogeneous Autoregressive model of Realized Volatility).

The HAR-RV model builds on the Heterogeneous Market Hypothesis, which recognizes that market participants operate across different timeframes — ranging from intraday market makers to weekly swing traders and monthly institutional investors. Because these groups react to news at different speeds, volatility cascades across different time horizons.

The model structure

Instead of applying a single flat lookback window, HAR-RV forecasts realized volatility by decomposing historical price action into three distinct, overlapping components:

The standard HAR-RV regression model is formulated as:

where c is a constant, the β coefficients are the weights assigned to each timeframe, and ε_t+1 is the residual error.

Why HAR-RV improves volatility estimates

• Captures volatility clustering — recent shocks immediately impact the daily parameter (β_d), allowing the model to adapt quickly to regime changes.
• Smooth decay — as an extreme move ages, its influence transitions smoothly from the daily to the weekly (β_w) and monthly (β_m) components, avoiding the sudden cliff effects typical of HV30.
• Realistic expectations — by weighting time horizons hierarchically, HAR-RV provides a clearer baseline when evaluating whether current option IV is truly elevated or simply catching up to recent physical swings.

Summary: HV30 vs. HAR-RV at a glance

HV30 is a flat, unweighted average. A massive market shock on day 29 affects today's HV30 calculation exactly as much as a shock from yesterday. On day 31, that shock abruptly falls off the lookback window, causing a sudden, artificial drop in your volatility reading.

HAR-RV is a time-decayed cascade. It explicitly breaks the lookback into daily, weekly, and monthly components. While the exact weights (β) are dynamically calibrated via linear regression based on the specific stock's historical behavior, empirical data almost always assigns the heaviest weight to the daily component, followed by the weekly, and then the monthly.

By weighting recent volatility more heavily, HAR-RV instantly reacts to new market shocks, while the weekly and monthly components act as a stabilizing "memory" that allows the shock to smoothly decay over time rather than dropping off a cliff. That is exactly why institutional risk engines prefer it for pricing options — and why a VRP screen built on HAR-RV will disagree with an HV30 screen precisely when it matters most: at a regime turn.

If you're new to reading implied versus realized volatility, start with our companion guide on decoding volatility metrics and the options Greeks, which covers IV Rank, IV Percentile, skew and the Rule of 16. The FinoAgent quant engine applies this machinery to live options chains, ranking defined-risk strategies against options-implied probability distributions rather than raw yield.

Frequently asked questions

What is the volatility risk premium (VRP)?

The gap between implied volatility — what the options market expects the underlying to move — and the volatility it actually realizes. When implied exceeds realized, sellers are overpaid for risk. When it inverts, options are cheap relative to physical movement and sellers face a structural disadvantage regardless of directional accuracy.

What is the HAR-RV model?

The Heterogeneous Autoregressive model of Realized Volatility, introduced by Fulvio Corsi. It forecasts realized volatility from three overlapping components — daily (1 day), weekly (5 days) and monthly (22 days) — with regression weights calibrated from the asset's own history.

Why is HAR-RV better than HV30 for forecasting volatility?

HV30 weights every day equally and reacts slowly to regime shifts. HAR-RV puts the heaviest weight on the daily component, capturing volatility clustering and adapting immediately to shocks, while the weekly and monthly components let that shock decay smoothly instead of dropping off a cliff.

What does an IV/HV ratio above 1.0 mean?

Implied volatility is pricing a wider distribution of returns than recent realized price action reflects — a potential VRP edge for sellers. Below 1.0, realized movement is outstripping what options cost.

What is the drop-off effect in HV30?

When a single large volatile day reaches day 31 it falls out of the 30-day window, so the HV30 reading drops sharply even though nothing changed in the market — an artifact of the window, not a real shift in regime.

What are the daily, weekly and monthly components in HAR-RV?

Realized variance over the past 1, 5 and 22 trading days respectively. Each enters the regression with its own β coefficient, and empirically the daily component almost always carries the heaviest weight.