Realized Volatility Under Microstructure Noise

What this microstructure-noise volatility coding problem tests

This is a hard Python coding problem that appears in quantitative finance interviews, particularly at high-frequency trading firms. It asks you to implement a real statistical estimator—the Two-Scale Realized Variance (TSRV)—used to filter out bid-ask bounce and other microstructure effects from raw tick data. The problem tests both your ability to translate a formal mathematical specification into clean, correct code and your understanding of why the formula works.

Strong solutions carefully manage index arithmetic across multiple sampling scales, compute log-returns accurately, and handle edge cases (empty or too-short price sequences). Interviewers often follow up by asking you to reason about why the bias-correction term reduces noise, how to extend the approach to higher-frequency data, or how performance scales with the number of ticks.

• Logarithmic returns and realized variance
• Index slicing and subgrid construction
• Noise bias estimation and subtraction
• Numerical stability and edge-case handling