The risk-off signature, and its missing limbs

What the whole pattern looks like

A complete risk-off day has a signature. The dollar strengthens against most currencies but weakens against the yen, US Treasury yields fall, gold rises, crude falls, volatility jumps, credit spreads widen, equity indices fall in every time zone, and foreign investors sell Indian shares.

Nine moves, one decision. Somebody with money at risk decided to hold less of it, and every line above is that decision seen from a different desk.

Recognising the pattern when it is whole takes no skill at all — the newspapers will have named it by the evening. The pattern is almost never whole, and the missing limb is where the information is. Equities down with the dollar flat, credit unmoved and the global tape fine is not a global risk event. It is a domestic or sectoral story wearing macro clothes, and reading it as the former is how a stock-specific problem gets treated as a reason to change an allocation.

This article is a chain link off the macro signals pillar, which sets out the five channels one at a time. This one takes the cross-section instead: what it means when every channel says the same thing on the same day, and what it means when they do not.

Why these particular assets move together

Because they are not nine decisions but one. An investor reducing risk is trying to hold fewer claims whose value depends on things going well, and more claims that pay regardless. Each limb of the signature is one step in carrying that out.

Equities fall everywhere at once, which is the part that looks strange and is not. A shareholder is the residual claimant — paid after everyone else, out of whatever is left. Someone cutting equity exposure in general is not forming a fresh view on each market first; they are reducing a category. So Tokyo, Frankfurt, New York and Mumbai fall together without any of them causing the others, which is exactly why "global markets fell in sympathy" explains nothing. There was no sympathy. There was one seller with holdings in all four.

Government bond yields fall because bonds are the destination. The buyer wants a promise that does not depend on growth, and US Treasuries are the deepest market for that kind of promise. Buying pushes prices up, and a bond's yield is its price read backwards — the arithmetic is in yield to maturity explained. Note carefully what this is not: a falling yield here is not the market taking a view on inflation or policy. It is the mechanical result of demand for the instrument. The same tile carries two entirely different meanings and the tile cannot tell you which one you are looking at — the fuller treatment of both is in US yields and Indian share prices.

The dollar strengthens for two separate reasons. Buying Treasuries takes dollars, so the flight into them is dollar demand. And most cross-border borrowing is written in dollars, so anyone unwinding a leveraged position has to buy back the currency they borrowed. Both push the same way. What that does to an emerging market is the subject of the dollar index and emerging markets.

The yen goes the other way, and it is the exception that proves the mechanism. The yen is a funding currency: positions get financed by borrowing it cheaply. Unwinding those positions means buying yen back, so USD/JPY falls while the dollar rises against nearly everything else. Both facts are the same risk reduction. Why that unwind is fast, and why it sells assets with no connection to Japan, is in the yen carry trade explained.

Gold rises for two reasons that happen to arrive together. It is a claim on nobody, which is the point of holding it when the question is whether promises get kept. And it pays no income, so the cost of holding it is the yield given up elsewhere — a cost that just fell, because yields fell. Demand up and carrying cost down, at the same moment.

Crude falls because it is consumed, not held. Its price is a running bet on industrial demand, so a day that reprices growth reprices the barrel. For India that is a tailwind arriving inside a headwind, which is the odd asymmetry worked through in crude oil and the Indian economy.

Credit spreads widen and volatility jumps because both are prices of insurance. A spread is what a lender charges for the chance of not being repaid; an option premium is what a seller charges for carrying somebody else's downside. Doubt raises both, and the VIX is a price of protection rather than a measurement of mood. Why credit registers this before equity usually does — and where that ordering fails — is in credit spreads and where trouble shows up first.

Hold on to the distinction the rest of this article rests on. Each limb has a mechanism, which is durable and re-derivable. The limbs appearing on the same day is an observation about days. The mechanisms are what let you diagnose an incomplete signature; the observation on its own only lets you recognise a complete one.

The signature, limb by limb

The middle column is the one to learn. It is what the move is pricing, which is not always what it appears to be pricing.

Nine rows, and the complete set is rarer than the phrase "risk-off" suggests. We have not put a count on how often all nine land together, and neither should anyone quoting one at you — the honest statement is that the partial case is the ordinary one, and the partial case is what the rest of this article is about.

Why half the tiles turn green on a bad day

Because a tile's colour shows the modelled effect on Indian equities, not the direction of the number. A falling dollar shows green because a softer dollar is mechanically a tailwind here. There is no rule that up is red, and that convention is what makes a macro page confusing on precisely the days it should be clearest.

The two currency tiles make that unmissable, because on a risk-off day they move in opposite directions and render the same colour. The dollar index rises, and a stronger dollar is scored as a headwind, so that tile is red. USD/JPY falls — and the model's sign on USD/JPY is positive, because a weaker yen is what keeps the carry trade financed and the money it funds in the market. A stronger yen is therefore also scored as a headwind, and that tile is red too. Two currency tiles, moving opposite ways, both red. A rising number is green on one tile and red on another, which is the whole convention in one screenshot.

Now apply that to the signature above, and something awkward happens. On a textbook flight-to-safety day, US Treasury yields fall — and a falling yield is scored as helpful, because a lower discount rate raises the present value of distant earnings. Brent falls — and a cheaper barrel is scored as helpful, because India buys most of its crude abroad. So in the middle of a genuine global risk event, the yield tile and the crude tile both render green while the market falls.

That is not a defect, and it is worth being precise about why. Both scorings are correct as general statements: a lower discount rate does help valuations, and a cheaper barrel does help the import bill. What the tile cannot encode is the reason the number moved. A tile knows the usual consequence of a move. It does not know the cause of it — and on a risk-off day the cause is the same aversion that is doing the damage elsewhere on the page.

The practical consequence is the whole argument of this article in one line: on days like this, individual tiles mislead and the shape of the whole page does not. Read the pattern across the tiles, not the colour of any one of them.

The diagnostic: reading an incomplete signature

Here is where the pattern earns its keep. Each row below is a partial signature and the reading it most plausibly supports. These are diagnostic hypotheses to be checked, not conclusions — the third column would be better read as "the first thing to test".

Two of those deserve unpacking, because they are the ones most often read wrongly.

Equities down with yields up is not a risk-off day. It is close to the opposite in mechanism. In a flight to safety, bonds are where the money goes; in a rate shock, bonds are where the trouble starts, and equities fall because the rate their future earnings are discounted at just rose. Same red screen, different transmission, and the tell is the sign on the yield. An investor who files both under "risk-off" has thrown away the one piece of information that distinguishes them.

Gold falling alongside equities is the tell for a liquidity event. When the constraint is a margin call rather than a change of preference, the thing that gets sold is whatever has a bid — which is often the position that is still working. So the hedge gets liquidated to fund the loss, and for a stretch everything correlates to one. This is a judgement about the usual order of events rather than a measured regularity, and it is stated as such: the useful part is not the label but the check, which is to notice that a hedge failing to hedge is itself information about who is selling and why.

One caution before using any of this. A limb can look missing because the data has not published. The US 2-year and both credit-spread tiles come from FRED, whose daily series typically appear a session or two behind, and India's FII cash figure lands after the close. An unchanged number is sometimes a market that did not move and sometimes a feed that has not updated, and treating the second as the first will produce a confident diagnosis of an incomplete pattern that was never incomplete.

The Indian limbs arrive last, and two of them are narrower than they look

India sits at the end of the chain for a reason that is purely about clocks. The NSE and BSE cash sessions run 09:15–15:30 IST, which means a US session lands overnight. An Indian investor looking at the screen at the open is not watching a global event happen; they are watching the local market price one that finished hours ago.

That has a specific consequence worth naming. A gap at the open is not a second event. It is the same event arriving on local time, and counting it as fresh information — on top of the overnight move you already saw — is double-counting one shock.

Three Indian limbs sit on the page, and each has a limitation.

Foreign institutional cash is the most direct evidence available: someone actually sold. But the figure covers the cash segment only, and it publishes after the close. Foreign positioning also runs through index and stock futures, options and the primary market, none of which appears in the print. A large cash sale is proof of selling in the cash segment, not proof that total exposure fell.

The rupee tells you how much of the global dollar move has been transmitted here rather than adding a new channel. In our composite USD/INR carries a weight of 0.06 with a negative sign — a weaker rupee counts as a headwind — against the dollar index's 0.20. That ordering follows the mechanism: the index is the cause being watched and the pair is the local symptom.

Which sets up the one loop in the whole chain, and the reason those two Indian limbs are not independent. A foreign investor who sells Indian shares ends up holding rupees they did not want. To the extent the proceeds are repatriated rather than left onshore, that is rupee selling and dollar buying — so the FII line and the currency line are, in part, the same flow recorded twice, once by the exchange and once by the currency market. Counting them as two independent confirmations of a risk-off day inflates the evidence. It is one limb arriving through two doors.

India VIX against the US VIX is the pair worth reading together rather than separately. India VIX calm while the US VIX is climbing means the domestic options market is pricing insulation from something the global market is not. That is a fact about what is priced, and it has two readings — the insulation is real, or it is not yet priced. The page shows the gap. It does not resolve it, and nothing here says which side moves.

Why our composite argues with itself on a risk-off day

The Macro page compresses the tiles into one number between −100 and +100. The formula is a weighted mean: each input's move is converted to a contribution, clamped to the range −1 to +1 and signed by whether rising helps or hurts Indian equities, then score = 100 × Σ(wⁱ·cⁱ) / Σ(wⁱ). Below −20 we label the day risk-off; above +20, risk-on.

Those weights and cut-offs are FNOTrader's judgement, not measured constants. They were not estimated from a regression and they are not a property of the world. Someone with a different reading of what matters to Indian equities would choose differently and would not be wrong on the arithmetic. Treat them as a stated opinion with a formula attached.

The model carries twenty weighted tiles whose weights sum to 1.33, and foreign institutional cash is folded into the same average at 0.12 without appearing as a tile — twenty-one contributions and a divisor of 1.45 on a day when every one of them reports. A feed that fails drops its weight out of the top and the bottom of the fraction alike, which is why the formula divides by the weights actually present rather than assuming they sum to one. The five largest tiles are the dollar index at 0.20, the US 10-year at 0.15, Brent at 0.12, USD/JPY at 0.10 and the VIX at 0.10 — the largest components, not the whole set. The rest carry the remaining weight: the five global equity indices together come to 0.18, the three volatility inputs to 0.21, the two credit-spread inputs to 0.10, copper sits at 0.04 and gold at 0.03.

Gold's entry is worth pausing on, because it shows what a design choice looks like when you open one up. Its sign is negative — a rising gold price is scored as a headwind for Indian equities. That only makes sense if you read gold as a fear gauge rather than as a commodity price, which is what we have decided to do. It is defensible and it is contestable in the same breath, and nothing in the arithmetic settles it.

Now put the signature through it, and the non-obvious part appears.

On a textbook risk-off day, a bit over a fifth of the model's total weight pushes the score the wrong way — by construction. The US 10-year at 0.15, the US 2-year at 0.06 and Brent at 0.12 all carry a negative sign, meaning a fall in each is scored as helpful. On a flight-to-safety day all three fall. That is 0.33 of weight out of the 1.45 that divides the score when everything reports, a shade under 23%, contributing positively while the equity, credit, volatility, currency and flow inputs contribute negatively. The yield and crude pair alone, at 0.27, outweighs the dollar's 0.20.

This is not a flaw to be fixed, and understanding why is the useful part. The signs are right in general: cheaper money and a cheaper barrel genuinely help Indian equities. What a weighted average cannot represent is that on this particular day those two moves are symptoms of the aversion driving the rest of the page rather than independent good news. An average adds up directions. It cannot read a pattern.

Two design consequences follow, and both are deliberate. First, because every input is clamped at ±1, the score measures breadth of agreement across channels rather than severity within one — which makes it, almost by accident, a completeness meter for the signature described in this article. A wide, shallow risk-off day scores more negative than a violent single-channel one. Second, because that average is slow and partially self-cancelling, the page carries a stress state that bypasses it: a score at or below −35 triggers it, and so does a one-day VIX jump of 20% or more, or a one-day fall of 1.2% or more in USD/JPY. Those three cut-offs are ours too, on the same footing as the weights.

The USD/JPY trigger is the one that connects back to the signature. A yen leg moving that fast is a carry unwind, and an unwind is the case where a single limb moves violently while the rest of the page has barely moved — which an average over twenty-one inputs registers faintly, by construction rather than by accident. The override exists to say so out loud instead of letting one number smooth it away. The trade-off in the whole design is worth stating plainly: one number is easier to read and tells you less, and it can never tell you which channel is doing the work.

The heatmap will not tell you a day was risk-off

The correlation panel measures how tightly two series moved together day by day — the Pearson correlation on daily returns, over a window of 30, 60 or 90 sessions, using the dates the two series share. It is the natural place to look for confirmation that assets are moving as a bloc, and it is the wrong place.

A correlation is a property of a window, not of a day and not of a mechanism. A 30-session coefficient tells you what those 30 days produced; it cannot isolate the one session you are interested in, and a single violent week inside a 30-session sample can carry the whole number. When a coefficient flips sign between the 30-day and the 90-day window, the honest reading is that you have learned something about the window.

It also cannot establish direction. Two assets move together on a risk-off day because both respond to the same decision by the same holders — neither one is driving the other, and the sentence "the dollar is driving Nifty" is not something a correlation can support. In macro, two things moving together because both respond to a third is the ordinary case rather than the exception.

What the panel is genuinely good for is the reverse check. Take a channel whose mechanism you can already state, and ask whether the recent record is consistent with it. Consistency is mild support. Inconsistency is a prompt to ask what else has been dominating, which is a question and not an answer.

Five ways this read goes wrong

1. Concluding from one limb. "The dollar is up, so it's risk-off" is the commonest version. The dollar rises on relative growth too, and on rate differentials, and on a single large flow. One limb is a prompt to check the other eight.
2. Filing a rate shock as a flight to safety. Equities down with yields up is a different event with a different transmission, and the sign on the yield is the whole tell. Both produce a red screen; only one is people buying safety.
3. Reading colour as direction. Green means helpful to Indian equities. On a genuine risk-off day the yield and crude tiles are green while the market falls, because a tile encodes the usual consequence of a move and not its cause.
4. Treating one day as a regime. A day is a day. Our regime labels attach to a score crossing ±20 on the day, and those cut-offs are our choice; nothing in the construction makes a single crossing a state of the world.
5. Turning an environment read into an entry. This is the expensive one. A cross-asset pattern describes the conditions a position sits in — it has no view on your holding period, your allocation or the price you paid, and none of the mechanisms here run on a timescale that produces a trade. What a drawdown does to a long-horizon plan is a separate question, worked through in asset allocation and in what happens when you stop a SIP in a drawdown.

There is a sixth, quieter one. The complete signature is memorable precisely because it is rare, and a pattern you have learned to recognise is a pattern you will start seeing in four limbs out of nine. The discipline is to count the limbs before naming the day.

Where this sits in the app

Checking a signature means having the limbs on one screen at the same time, which is the practical reason a macro page exists at all.

FNOTrader's Options Analytics app carries the macro page described here: cross-asset tiles grouped by channel with the modelled effect on Indian equities attached to each, the composite score with its largest contributors named in words, a separate stress state on the cut-offs stated above, foreign and domestic institutional cash, the correlation heatmap across 30, 60 and 90-session windows, and a per-tile history chart from one month to five years so a move can be read against its own range.

The weights, signs, regime boundaries and stress triggers are the ones in this article, and they are ours. They are published rather than hidden so that a reader who disagrees can see exactly what they are disagreeing with.

Common questions

Continue reading

More in Macro & Markets · App: Options Analytics · Definitions: glossary · Free tools: calculators · All: every article