A stock gaps down 3% at the open, rallies 4% from the low, and closes flat. Standard close-to-close volatility for that day? Nearly zero. The actual intraday range told a completely different story. That gap between what closing prices show and what actually happened during the session is exactly what Parkinson Volatility was designed to capture.
Michael Parkinson introduced his range-based volatility estimator in 1980, and the core insight still holds: the distance between a session’s high and low contains more information about true price variability than the closing price alone. If you have been relying only on historical volatility from close-to-close returns, you are working with a fraction of the data your charts already contain.
What Parkinson Volatility Measures
Parkinson Volatility is a range-based estimator. Instead of measuring how far the close moved from yesterday’s close, it measures how far apart the high and low were during each session. The idea is straightforward: the high-low range reflects the full extent of price discovery within a trading day. Close-to-close returns throw most of that information away.
Think of it this way. Two stocks both close at $100 today and closed at $100 yesterday. One traded between $99.50 and $100.50. The other traded between $95 and $105. Close-to-close volatility treats these identically. Parkinson does not. The second stock was ten times more volatile during the session, and Parkinson’s formula captures that difference.
I track Parkinson Volatility alongside standard deviation on my daily scans specifically because it flags when intraday movement is expanding while closing prices look calm. That divergence between range expansion and flat closes is often an early sign that a breakout or breakdown is building pressure.
The Parkinson Volatility Formula
The formula uses the natural logarithm of the high-low ratio for each session, then scales the result:
\sigma_P = \sqrt{\frac{1}{4n \ln 2} \sum_{i=1}^{n} \left( \ln \frac{H_i}{L_i} \right)^2}
Where H_i is the session high, L_i is the session low, and n is the number of sessions in your lookback window.
The \frac{1}{4 \ln 2} scaling factor is not arbitrary. Parkinson derived it from the properties of geometric Brownian motion. Under the assumption that price follows a continuous random walk within the session, this factor makes the estimator unbiased. The constant 4 \ln 2 \approx 2.773 comes from the expected value of the squared range of a Brownian bridge.
A common mistake is omitting the \ln 2 term and just dividing by 4n. That gives you a number, but it is not the Parkinson estimator, and it will consistently overstate volatility by about 44%.
To annualize, multiply the daily result by \sqrt{252}, same as any daily volatility measure.
Parkinson Volatility in Practice With Real Data
Here is how the calculation works on real SPY data from the week of April 14-17, 2026:
April 14: High $694.58, Low $687.66. \ln(694.58 / 687.66) = 0.01001
April 15: High $700.28, Low $694.20. \ln(700.28 / 694.20) = 0.00876
April 16: High $702.78, Low $698.53. \ln(702.78 / 698.53) = 0.00608
April 17: High $712.39, Low $705.76. \ln(712.39 / 705.76) = 0.00940
Sum of squared log ratios: 0.01001^2 + 0.00876^2 + 0.00608^2 + 0.00940^2 = 0.000301
Plugging into the formula: \sigma_P = \sqrt{\frac{0.000301}{4 \times 4 \times 0.6931}} = \sqrt{0.0000271} = 0.0052 daily, or about 8.3% annualized.
Close-to-close standard deviation over the same four days gives roughly 7.7% annualized. The difference is small here because SPY’s intraday ranges were fairly symmetric that week. The gap gets much wider when sessions have large ranges but small net moves.
Compare AAPL over the same period: highs and lows of $261.93/$257.19, $266.56/$257.81, $267.16/$261.27, and $272.30/$266.72. The Parkinson estimate comes to about 23.2% annualized. AAPL’s intraday ranges were wider relative to its price, especially on April 15, when the stock covered nearly $9 from low to high. That kind of session range is invisible in close-to-close volatility if the stock closes near its open.
Why Range Beats Close-to-Close
Parkinson’s estimator is statistically more efficient than close-to-close volatility. “Efficient” here has a specific meaning: for a given number of observations, it produces a tighter estimate of the true underlying volatility. Parkinson showed his estimator is roughly 5 times more efficient than the standard close-to-close method.
This matters most when you have limited data. If you are estimating volatility over 10 trading days, the close-to-close method gives you 10 data points: ten closing prices. But each of those sessions also produced a high and a low. Parkinson’s formula uses those highs and lows, extracting information that was sitting in your data all along.
The practical payoff is faster signal detection. When intraday ranges start expanding, Parkinson picks it up sooner than a close-to-close measure. I have seen this repeatedly in pre-earnings setups: the stock’s closing price drifts sideways, but the daily range widens for three or four sessions. Parkinson flags that expansion. Close-to-close volatility does not react until the close actually moves.
If you use Bollinger Band Width for squeeze detection, running a parallel Parkinson calculation over the same lookback gives you a second opinion. When Band Width is flat but Parkinson is rising, the squeeze may be closer to firing than Band Width alone suggests.
Where Parkinson Volatility Fails
The biggest limitation is gaps. Parkinson’s formula only sees what happens between the session’s high and low. It completely ignores the gap between yesterday’s close and today’s open. A stock that gaps down 5% at the open and then trades in a tight 1% range all day looks calm through Parkinson’s lens. Close-to-close volatility captures that gap. Parkinson does not.
This is not a minor edge case. Stocks gap every day. Futures gap on Sunday opens and around economic releases. Any market with regular overnight sessions or pre-market trading will have significant price movement that falls outside the regular session’s high-low range. Parkinson systematically understates volatility for gap-heavy instruments.
The assumption underneath the formula also matters. Parkinson assumed prices follow geometric Brownian motion with zero drift. Real prices trend. When a stock is in a strong directional move, the high-low range is asymmetric. The estimator still works, but its efficiency advantage shrinks. For strongly trending instruments, the theoretical 5x efficiency advantage drops considerably.
Another failure mode: thin or illiquid markets. If a stock’s high of the day was a single errant print 2% above the rest of the session’s range, Parkinson treats that as real volatility. Close-to-close volatility is more resilient to those prints because it only uses the closing auction. I filter Parkinson signals on anything with average daily volume below 500,000 shares for exactly this reason.
Parkinson Volatility vs. ATR
The Average True Range and Parkinson Volatility both use the session range, but they answer different questions. ATR measures average range in dollar (or point) terms. Parkinson measures range-based volatility as an annualized percentage. ATR is a practical tool for setting stop distances. Parkinson is a statistical estimator you compare against other volatility measures.
The key structural difference: ATR’s “true range” includes overnight gaps by taking the maximum of (high minus low), (high minus previous close), and (previous close minus low). Parkinson ignores gaps entirely. So ATR and Parkinson can diverge significantly after a gap-heavy week. If you are using ATR bands for swing stops, layering Parkinson on top tells you whether the intraday range alone is expanding or whether ATR is being driven primarily by overnight gaps.
That distinction matters for stop placement. Intraday range expansion (Parkinson rising) suggests the market is actively volatile during your trading hours. Gap-driven ATR expansion suggests overnight risk you cannot manage with intraday stops.
Common Mistakes Traders Make
The first mistake is treating Parkinson as a drop-in replacement for historical volatility. It is not. It measures something different. If your strategy depends on overnight gap risk, Parkinson will understate your exposure. Use it as a complement, not a substitute.
The second mistake is using too short a lookback window. With only 5 sessions, the estimate is noisy enough to mislead. I use 20-day Parkinson for swing trading context and 60-day for broader regime classification. Below 10 sessions, the efficiency advantage over close-to-close largely disappears because you just do not have enough observations.
The third mistake is comparing Parkinson values across different instruments without normalizing. A $500 stock will naturally have wider dollar ranges than a $20 stock. The formula handles this because it uses log ratios, not dollar ranges. But traders who hand-calculate or build quick spreadsheets sometimes accidentally use dollar ranges instead of log(H/L). That gives you something, but it is not Parkinson.
The fourth, and maybe most costly, mistake is ignoring what Parkinson tells you about your volatility regime and using the same position sizing regardless. When Parkinson rises from 10% to 25% annualized over two weeks, the intraday environment has fundamentally changed. Your position sizes and stop widths need to adjust. I have watched traders hold the same share count through a doubling of Parkinson volatility, and the equity curve damage is predictable.
Pairing Parkinson With Your Volatility Toolkit
Parkinson Volatility becomes most useful when you compare it to close-to-close historical volatility. The ratio between the two is informative. When Parkinson is significantly higher than close-to-close, it means intraday ranges are wide but net price movement is small. That is compression. When Parkinson and close-to-close converge, the market is moving directionally with less intraday reversal.
For TrendsAndBreakouts readers who already use the Relative Volatility Index, Parkinson adds a different dimension. RVI tells you whether volatility is expanding or contracting relative to its own history. Parkinson tells you whether the range is capturing more or less of the true price movement than closing prices suggest. They are complementary signals, not substitutes.
The practical workflow: run a 20-day Parkinson estimate alongside your standard 20-day close-to-close volatility. When Parkinson is more than 1.5 times the close-to-close estimate, something interesting is happening inside the session that closing prices are hiding. That is often a pre-breakout setup, a distribution phase, or an accumulation zone where large players are active intraday but managing the close.
When Parkinson Is the Right Tool
Use Parkinson Volatility when you trade instruments with minimal overnight gaps: forex pairs during liquid sessions, index futures during regular hours, or intraday timeframes where every tick falls between the session’s high and low. In those cases, Parkinson gives you a tighter, faster-reacting volatility estimate than close-to-close.
Be cautious with individual stocks that gap frequently, with futures around economic releases, and with anything that has a meaningful overnight session. In those cases, Parkinson will understate risk. Layer it with close-to-close or use a more complete estimator like Yang-Zhang that accounts for gaps.
The estimator is over 45 years old. It is simple, fast to compute, and extracts real information from data you already have. No exotic inputs, no optimization parameters, no curve fitting. Just highs, lows, and a bit of log math.
Educational content only. Not investment advice. Trading involves risk. You are responsible for your decisions.