Chapter 1 of 5 · 10 min

The square root of time

Independent days add in variance, so volatility scales with the square root of time. Daily to annual, annual to a ten-day horizon, a standard deviation to a value at risk: one rule, run in every direction.

By the end of this chapter you can
  • Annualize a daily volatility with √252, and go back
  • Scale volatility to any horizon and turn it into a dollar move
  • Compute a one-day 95% value at risk
  • Solve for the days a target move needs, and say when √t fails
1

The intuition

If each day's return is independent of the last, the variances of the days add up. A year of 252 trading days has 252 times the variance of one day, and √252 ≈ 15.9 times the volatility. A stock with a 1.26% daily standard deviation is a 20% annual stock. Traders use 16: divide annual vol by 16 for a typical daily move.

The same root runs everywhere. Twenty percent annual over ten trading days is 20% × √(10 ÷ 252) = 4%, so a $50M position has a one-sigma ten-day move of about $2M. A one-day 95% value at risk is 1.645 daily standard deviations: about $1M on that position. And the rule is a default, not a law: trending returns make long-horizon risk larger than √t says, mean-reverting ones make it smaller, and smoothed prices on illiquid assets make daily vol look far too low.

The key idea

σ_annual = σ_daily × √252. σ over h days = σ_annual × √(h ÷ 252). One-sigma dollar move = position × σ_h. One-day 95% VaR = 1.645 × σ_daily × position. Days for a one-sigma move to reach x: h = 252 × (x ÷ σ_annual)².

2

Why it works

  • The conventions here: independent daily returns with constant volatility, 252 trading days a year. VaR is one-sided at 95%, normal returns, zero mean.
  • Variances add, volatilities do not. Var(sum of h independent days) = h × daily variance, so σ_h = σ_daily √h. Multiplying a daily vol by 252 instead of √252 is the intern's mistake, and it is off by a factor of 16.
  • Every direction is the same formula. Daily from annual divides by √252; a horizon multiplies by √(h ÷ 252); a target move squares the ratio to get days back.
  • VaR is a quantile, not a worst case. 1.645 is the one-sided 95% point of a normal. Losses exceed VaR about one day in twenty; VaR says nothing about how large those losses are, and real returns have fatter tails. Expected shortfall answers that.
  • When √t understates risk. Positively autocorrelated returns, momentum or trending markets, cluster the bad days. Illiquid assets with smoothed marks are the classic case.
  • When it overstates. Mean-reverting returns partly cancel over long horizons. Neither correction is large for liquid equities over days to weeks, which is why the rule is the default.
Daily standard deviation 1.26%; a $50M position; a 10-day horizon
Annual vol: 1.26% × √252 = 1.26% × 15.8720.0%
Rough check: 1.26% × 1620.2%
10-day vol: 20% × √(10 ÷ 252)3.98%
One-sigma 10-day move: $50M × 3.98%$1.99M
One-day 95% VaR: 1.645 × 1.26% × $50M$1.04M
Days for a one-sigma move to reach 10%: 252 × (10% ÷ 20%)²63 trading days

Quadruple the horizon to 40 days and the vol doubles to 7.97%: four times the time, twice the volatility.

3

The formulas

σ_annual = σ_daily × √252; σ_daily = σ_annual ÷ √252

Independent days add in variance.

σ_h = σ_annual × √(h ÷ 252)

Any horizon in trading days.

One-day 95% VaR = 1.645 × σ_daily × position

The one-sided 95% point of a normal.

h = 252 × (x ÷ σ_annual)²

Days for a one-sigma move to reach x: square the ratio.

4

Worked example

Multiply the daily figure by the square root of 252. The follow-up is the mental shortcut: times 16.

Drawing the numbers…
5

See it move

Same asset. Change its annual volatility, the horizon, the size of the position and the target move.

Drawing the numbers…
Try this
  • Raise the annual volatility. Every volatility and every dollar figure rises, and the days a target move needs fall as the square.
  • Lengthen the horizon. The horizon volatility and the dollar move rise with the root of the days; the one-day VaR does not move.
  • Increase the position. Both dollar figures scale with it; no volatility changes.
  • Raise the target move. The days needed rise as its square.
6

Run it backwards

The annual volatility is quoted. What is a single day's standard deviation?

Drawing the numbers…

Divide by √252, about 15.87. Or, in your head, divide by 16.

The harder inverse is in the check: the trading days over which a one-sigma move reaches a stated size, which squares the ratio of the two volatilities.

7

Traps

Multiplying daily vol by 252.
Variances add, not volatilities. Multiply by √252.
Using 365 days for returns.
Returns are measured on trading days: 252. The 365 belongs to option expiry.
Reading VaR as the worst case.
It is a 5% quantile under a normal. The tail beyond it is what hurts, and it is fatter than normal.
Trusting √t for smoothed or trending series.
Positive autocorrelation makes long-horizon risk larger than √t says. Illiquid assets are the classic case.
Doubling the horizon and doubling the vol.
Twice the time is √2 times the vol. Four times the time doubles it.
8

Say it in the interview

The interviewer asks

Daily vol is 1.26%. What is the annual vol, and the ten-day risk on $50M?

Say yours out loud first, then compare.
9

Check yourself

5 fresh questions, with new numbers. Answer each one correctly to finish the chapter. Get one wrong and you will see the full working, then you can try it again with new numbers.

Answers within 1% are marked right. Type the number; $, %, x and M are fine. First tries count toward Learned: the topic is Learned once every chapter is done and 75% of first tries were right.

0 of 5
Drawing your questions…
Remember
  • σ scales with √time: × √252 to annualize, × √(h ÷ 252) for a horizon.
  • Divide annual vol by 16 for a typical daily move.
  • One-day 95% VaR = 1.645 × daily σ × position, a quantile with a fat tail beyond it.
  • √t is a default: trending or smoothed returns break it.