Evaluating High-Frequency Microstructure without Tick Data: Range-Based Proxies, Interrupted Time Series and Structural-Break Evidence from the NIFTY 100

Abstract

Tick-level limit-order-book data are costly to obtain and process, which restricts microstructure research on emerging markets. This paper asks how much of the market-quality effect of algorithmic trading can be recovered from daily Open–High–Low–Close–Volume (OHLCV) data, and what an interrupted time series (ITS) design on such data can and cannot identify. Using NIFTY 100 data for December 2005–December 2025 and the January 2010 launch of co-location at the National Stock Exchange of India (NSE) as a candidate regime change, we construct spread estimators (Abdi–Ranaldo, Corwin–Schultz), the Amihud price-impact ratio and range-based variance estimators, and we treat the break date as a hypothesis, testing it with autocorrelation-robust estimation, crisis-window exclusions, sup-Wald and Bai–Perron break-date tests, placebo and permutation exercises and a formal test of a second (2015) regime. Range-based variance is lower after 2010 (Garman–Klass level change −2.97 × 10⁻⁴, Holm-adjusted p = 0.07), but the decline shrinks by about three-fifths when 2008–09 is excluded, disappears when a crisis indicator is included, is not unusual relative to estimates at other dates, and is not preferentially dated at January 2010: where the data prefer a break, it falls in mid-2009. Spread proxies and price impact show no detectable change, and index-level spread estimators are truncated or negative in most months and so carry little spread information. A 2015 regime is detected, if at all, only in the spread proxies, and in the direction of rising, not falling, spreads. A conventional daily specification with heteroskedasticity-robust errors would have given overwhelming evidence of a 2010 effect, which illustrates the weight such specifications cannot bear. The contribution is a transparent, replicable protocol, with open code, that states what daily OHLCV evidence can and cannot say about algorithmic trading.