Here's a question that keeps quantitative traders up at night: is the factor strategy that crushed it pre-COVID still the same beast it was? Value, momentum, low-volatility — these premia looked stable for decades. Then March 2020 happened, and the rulebook got shredded. Testing whether that rupture was a permanent regime shift or a temporary shock is genuinely hard, and getting it wrong is expensive.
The direct answer is this: you need formal statistical tools to distinguish noise from structural change. Eyeballing a rolling return chart won't cut it. The two workhorses here are the Chow test — introduced by Gregory Chow in 1960 — and the CUSUM family of tests. Each attacks the problem differently, and used together they give you a much sharper picture of when, and whether, your factor's behaviour fundamentally changed.
Think of the Chow test like cutting a road trip log in half and asking: did this car perform the same way in both halves? You run separate regressions on sub-periods, then test whether the coefficients are statistically different. Applied to ASX value premia, for instance, you'd split your return series at a nominated date — say, April 2020 — and compare factor loadings across periods. A significant F-statistic says something broke.
CUSUM — cumulative sum of recursive residuals — is more elegant. Rather than forcing you to nominate a breakpoint, it plots the running deviation of your model's residuals over time. When those residuals drift outside the confidence bands, the data is telling you the model broke. For ASX momentum strategies, post-COVID CUSUM charts frequently show a dramatic excursion right through 2020–2021, a period when momentum got absolutely demolished by sector rotation. You can read the theoretical foundation of structural change testing through regression analysis fundamentals on Investopedia, explore the formal econometric history via Wikipedia's structural break article, and understand how the original methodology was formalised through the Chow test's Wikipedia entry.
Your practical takeaway: before trusting any pre-COVID factor backtest, run a Chow test at March 2020, then layer on a CUSUM plot across your full sample. If both flag a break, your strategy needs re-estimation on post-break data alone — or a regime-conditional model that handles both worlds.
A backtest that ignores structural breaks isn't a backtest — it's historical fiction dressed up in a spreadsheet.
This content is for educational purposes only and does not constitute financial product advice. Past performance is not indicative of future results. Profit Logic Ltd (ACN 688 669 936) accepts no responsibility for errors or omissions in this content or anywhere on this website. Always seek advice from a licensed financial adviser before making investment decisions.