🎯 Indian sugar mills — sugar-to-cane ratio cycle model

The ratio

ln(WPI Sugar) − ln(WPI Sugarcane). Cane is bought at a price the government fixes each season, and Indian sugar futures have been suspended since December 2021, so neither leg is a number the market watches on a screen.

The signal

A 48-month rolling z-score of that ratio. The raw ratio trends hard — cane has risen 144% since 2012 against sugar’s 38% — and a trending series against a persistent one manufactures correlations out of nothing, so only the de-trended z-score is used.

The basket

Equal-weight monthly log returns across whichever mills have a price that month, so a new listing changes breadth without putting a step in the index. A minimum of 6 mills is required.

What the band shows

This model’s band is a forward 12-month return relative to Nifty 500, not a price target — the tracked instrument is a basket, so a rupee price would mean nothing.

What was ruled out

Before publishing, the signal was put through the tests that killed two earlier candidates. It clears a phase-randomised null (p = 0.008 against a null 95th percentile of 0.587), holds in 4 of 6 non-overlapping-window offsets, keeps the same sign in all four sub-periods, does NOT predict the market itself (so it is not beta in disguise), and survives being measured against a beta and size matched peer basket rather than an index.

Tested against random signals

The whole model was re-run with the sugar-to-cane ratio replaced by unrelated commodity prices, holding the mill basket, the horizon and the benchmark fixed. On the zone spread only 1.5% of them matched it. On return per rupee invested the bullish-zone rule beat 97% of them if you are told its direction in advance, but only 87% if the comparison signals are allowed to pick their own direction too — and this model's direction was read off the data, so the second number is the honest one. That is why the bullish side stays marked unproven.

How to read a zone

Each reading describes the 12 months after it, so the way to act on it is to hold that call for twelve months rather than trade the zone as it flips. The label lasts a median of three months; the claim it makes lasts twelve.

Nothing here has hindsight

Every historical call uses an expanding window: the quartile cuts at month t use only months up to t, and the outcome pool only months up to t−12, whose result was actually known by then. Bands pool on the label a month was given at the time, never by re-sorting history under today's cuts. The record above is what this model would have published then.

Where it breaks

• The bullish side is not established. The bullish zone leans on 2015 and 2020; drop either year and its median falls to roughly zero. Only the bearish zone survives dropping any single year.
• Avoiding the bearish zone is worth much less than it looks. Measured per rupee actually invested it beat the index by under 3 percentage points a year, and one bad year removes even that. Its value is in what it avoids, not what it earns.
• The band is badly calibrated on the bullish side too — only about 28% of bullish-zone outcomes land inside it against 60% by design, and the misses run above the band. In the bearish zone it covers about 94%.
• It may not really be a margin spread. The cane cost leg is statistically insignificant with the wrong sign, so this behaves largely as sugar-price mean reversion rather than the input/output economics the name suggests. How much so is unsettled: on one sample the sugar leg alone scored better than the ratio, and on a later one the ratio clearly beat both its legs.
• Roughly 14 independent 12-month windows sit behind it. That is thin, and every figure carries wide error bars.
• The same signal shows up weakly in refinery stocks, which we cannot explain from sugar economics and which may mean it is partly a broader commodity-cycle effect.
• WPI is revised after first publication and lands about two weeks after month end, so the newest month can move slightly.
• Sugar mill earnings are heavily policy-driven — export quotas, the sugar floor price and ethanol diversion can all overwhelm the ratio.