One axis, and the days it is the wrong axis

One axis, and what is actually moving along it

Risk-on and risk-off is the claim that on most days capital is moving in one direction along a single line — out of assets that pay you for bearing an uncertain outcome, into assets that pay you for avoiding one, or the other way round. One number describes the whole tape.

Put like that it sounds too crude to survive contact with a real session, and it explains a surprising amount anyway. On a great many days, equity indices in four time zones, the price of copper, the spread a weak borrower pays over a strong one, the price of an option and the direction of foreign money into Indian shares all move as though somebody had turned a single dial. No news connects them. Nothing happened to copper because of what happened to an Indian bank.

The reason one dial can do that is the subject of the next section, and it is worth getting right, because the mechanism is also the thing that tells you when the frame has stopped applying. A frame you have only learned as a pattern cannot tell you it is not fitting. A frame you have learned as a mechanism can.

This article is a chain link off the macro signals pillar, which sorts a cross-asset page into five channels. The pillar's second question — is capital moving toward risk or away from it — is this article's whole subject. The companion piece, what a risk-off day looks like across assets, takes the cross-section of a single day; this one takes the idea underneath it.

A change in the price of safety reprices everything relative to it

Every financial asset is the same object in different clothing: a claim on cash that has not arrived yet. What it is worth today is that future cash, discounted for two separate things.

The first is time — a rupee next year is worth less than a rupee now, and the baseline rate is what a lender charges when repayment is not in question. The second is uncertainty — the extra compensation demanded for bearing the chance that the cash is smaller than expected, or later, or absent. That second piece has a name worth having: the risk premium.

Here is the whole idea. A risk-on or risk-off move is a change in that second piece, and the second piece is common to every risky claim at once. Nobody has revised their forecast of a cement company's cash flows. What changed is the compensation being demanded for holding uncertain cash flows in general — so every uncertain claim reprices in the same direction on the same afternoon, and the correlation across hundreds of unrelated assets is not a coincidence, it is one number appearing in hundreds of valuations.

The arithmetic also explains why the move is not evenly distributed. Take an illustrative claim on ₹100 a year, forever. Discount it at 8% and it is worth ₹1,250; at 9%, ₹1,111 — an 11% fall, with the ₹100 untouched. Now take a claim on a single ₹100 payment one year out: at 8% it is ₹92.6, at 9% it is ₹91.7, a fall of under 1%. The same one-point change in the price of safety costs the long, uncertain claim roughly twelve times what it costs the short, near-certain one.

That is illustrative arithmetic, not a market figure, and it is the entire reason the frame ranks assets rather than merely sorting them. Claims whose value sits far out in time and depends on things going well — small companies, speculative growth, weak borrowers, emerging markets — carry more of that sensitivity than short, senior, near-certain ones. When the price of safety moves, they move more. Nothing about their own prospects changed.

One trade-off is worth naming before the frame gets used, because it is the thing people forget. Safety is bought, not given. The asset that reprices least is also the one that pays least, and the investor who holds it through a calm stretch has paid for insurance that was not needed. There is no side of this axis that is free.

The two ends, and what you are paid at each

The two poles are usually described by listing assets, which is the least useful way to learn them, because the list changes and the reason does not. The reason is the middle column.

Two entries in that last row deserve a flag, because they are where the neat picture starts to fray.

A currency is not a risky asset, yet it sits on the axis. The dollar is scored as a safety destination because most cross-border borrowing is written in it, so reducing a leveraged position means buying it back — the mechanism, and what it does to an emerging market, is in the dollar index and emerging markets. The yen sits at the same pole for a different reason: it is what the position was financed with, which is why a fast yen move is its own event rather than a point on a scale. That is the subject of the yen carry trade explained.

Gold is at the safety end without being a promise. It pays nothing, so what it costs to hold is the income given up elsewhere — which means it is priced off the same rate that sits at the centre of this article, worked through properly in gold and real rates. Keep that oddity in view. It returns in the fifth section as one of the two places the single word “safety” is doing two different jobs.

How the axis reaches an Indian share price

None of the above is about India yet, and the transmission is worth stating in one paragraph rather than assumed, because the axis reaches an Indian share price through channels that each have their own article.

A rise in the global risk premium reaches Mumbai four ways. It lowers what a foreign allocator will pay for the same Indian earnings, because the discount applied to them just rose — the arithmetic of that is in US yields and Indian share prices. It moves money out of the category rather than out of a company, which is what a foreign institutional cash figure records after the close — and the reason that line answers to a dollar rather than to anything Indian is worked through in why FII flows follow the dollar. It moves the rupee, which changes the dollar value of an unchanged rupee return — USD/INR and your portfolio. And it reprices crude, which for an importer lands in the accounts rather than only in sentiment: crude oil and the Indian economy.

One structural fact shapes how all four arrive. The NSE and BSE cash sessions run 09:15–15:30 IST, so a US session lands overnight. An Indian investor at the open is not watching a global event happen; they are watching the local market price one that finished hours ago. A gap at the open is the same shock arriving on local time, not a second shock — the handoff is worked through in how global indices lead the Indian open.

Five days the frame does not fit

Now the part that makes the frame worth learning rather than merely worth having. Each row below is a tape that a one-axis reading labels confidently and wrongly. The third column is what distinguishes the two stories — a check to run, not a conclusion.

Three of those need the mechanism spelled out. The last two are treated in full elsewhere: the funding case, and why a failing hedge is itself information, is in what a risk-off day looks like, and the rate shock is in US yields and Indian share prices and yield curve inversion.

The growth downgrade is the one that fools the frame most cleanly, because it produces the identical picture. Go back to the two pieces of the discount. A flight to safety is the premium rising: nothing about expected cash changed, and people paid up for claims that do not depend on it. A growth downgrade is the forecast falling: expected earnings are lower, and because a weaker economy implies a lower expected path of policy rates, government yields fall alongside. Equities down, yields down, both times.

The mechanical difference is that in the second story nobody repriced risk. And a risk premium is exactly what an insurance price is made of. So the distinguishing question is whether the things that are pure premium moved: what a weak borrower pays over a strong one (credit spreads), and what a seller charges for carrying somebody else's downside (the VIX and India VIX, and for the bond market the MOVE index). A cheaper growth forecast gives no reason for the price of insurance to change. A rise in the compensation demanded for uncertainty gives every reason. This is a diagnostic to test on the screen in front of you, not a rule — the two stories can arrive on the same day, in which case the honest answer is that the tape is not decomposable and the label is not worth applying.

The dollar and gold rising together is the second case, and it is where the single word “safety” quietly covers two different things. Ordinarily the two are pushed apart by mechanism rather than by mood: gold is quoted in dollars, so a stronger dollar makes the same ounce dearer in every other currency, and gold's carrying cost is the real income given up to hold it, which is a dollar-rate story. When both are bid at once, something is overriding that link. The reading worth holding is that there are two kinds of safety being demanded and the axis cannot distinguish them: safety as the promise of the strongest issuer, which is a bet that the system pays, and safety as no issuer's promise at all, which is a bet that you do not need it to. This is a judgement about what the word conceals, offered as such. We are not claiming to know which one is operating on any given day, and neither should a page that renders both as one colour.

The third case is the least dramatic, which is exactly why it gets a global label it has not earned. An Indian market that falls while the dollar, credit and the overseas tape sit still is not a global event, and the axis will still produce a reading, because a composite built from global inputs always produces a reading. Earnings, an index rebalance, a tax or regulatory change, a large domestic seller and the balance between foreign and domestic flow are all outside its coordinate system. Our own page carries domestic institutional cash beside the foreign line for precisely this reason — the foreign figure alone does not describe who was on the other side of the trade.

Our composite is a fit meter for this frame, not a damage meter

The Macro page compresses the tiles into one number between −100 and +100. How that number is built is the subject of how a composite macro score works, and this section takes only the part that bears on the frame: what a small reading means.

Two facts from that construction are all this argument needs. Each input's daily move is clamped to the range −1 to +1 before it is weighted, so no single channel can shout. And the model carries twenty weighted tiles summing to 1.33, plus foreign institutional cash folded into the same average at 0.12 — twenty-one contributions and a divisor of 1.45 on a day when every one of them reports. At −20 or below we label the day risk-off, at +20 or above risk-on, and a separate stress state takes precedence over both. Those weights, signs and cut-offs are FNOTrader's design judgement, not measured constants — not estimated from a regression, not a property of the world, and published precisely so that a reader who disagrees can see what they are disagreeing with.

Now work the ceiling out, because it makes the design unmistakable. The dollar index is the heaviest input at 0.20. Move it further than the model is capable of registering, so its contribution is pinned at the clamp, with every other input reporting exactly flat, and the score is 100 × 0.20 ÷ 1.45 = 13.8 points. The single largest input in the model, moving further than the model can even see, cannot on its own produce a risk-off label, because the boundary sits at 20. Foreign flow at full clamp reaches 8.3 points and falls shorter still. Crossing the boundary needs company: the dollar and the US 10-year together, both pinned and agreeing, come to 100 × 0.35 ÷ 1.45 = 24.1 points.

Which is the argument of this whole article expressed as arithmetic. A composite built this way cannot measure how violent a day was. It measures how many channels agree — which is to say, it measures how well the one-axis frame fits the day.

So a reading near zero has two entirely different meanings and the number cannot tell you which. It can mean a quiet tape. It can equally mean channels moving hard in contradictory directions and cancelling — a day the frame does not fit, which is the most interesting kind and the one a single number erases. That is why the gauge is published beside a list naming the day's largest contributors in words rather than on its own, and why the page carries a separate stress state that bypasses the average altogether when one channel moves violently and alone.

The colours encode impact, and three groups of tiles do not follow the rule

One convention before anyone reads a macro page against this frame, because it makes the page confusing on precisely the days it should be clearest.

A tile's colour shows the modelled effect on Indian equities, not the direction of the number. Green is supportive, red is a headwind, and the legend on the page says so in those words. A falling dollar renders green. A rising USD/JPY also renders green — the model's sign on that pair is positive, because a weaker yen is what keeps a carry position financed. Two currency tiles, moving opposite ways, can carry the same colour.

Which is fine until you put a genuine flight to safety through it. Falling government yields are scored as helpful, because a lower discount rate raises the present value of distant earnings. Falling crude is scored as helpful, because India buys most of its oil abroad. So in the middle of a global risk event, the yield tile and the crude tile render green while the market falls. Both scorings are correct as general statements. What a tile cannot encode is the reason the number moved — and on that kind of day the reason is the same aversion doing the damage everywhere else on the page.

Now the part that is a genuine wrinkle rather than a deliberate convention, and it is worth knowing because it produces a contradiction you can see on one screen. The colour rule has several branches, and only some tiles reach the one that consults the model's sign.

• The 2s10s curve tile is never coloured at all. It is contextual, and colouring it would assert a direction the model does not take.
• Yield and spread tiles are coloured by the direction of the level: a falling yield or a tightening spread renders green. That branch runs before the score is consulted, so the 5-year and 30-year yield tiles carry impact-style colour despite having no weight in the score at all.
• Weighted price tiles — the dollar, crude, gold, copper, the vol indices, the scored equity boards — use the model's signed contribution, which is the convention as advertised.
• Unweighted price tiles fall through to raw direction: rising renders green. That is WTI, silver, natural gas, Dow futures, Shanghai, KOSPI and the FTSE.

The last branch produces the tell. Brent carries a weight and a negative sign, so a rising Brent renders red. WTI carries no weight, so a rising WTI renders green — same commodity, same session, opposite colours, a few centimetres apart. Nothing is broken: one tile is expressing a modelled effect on Indian equities and the other is expressing which way a number went, and the page does not distinguish them visually. Once you have seen it, the discipline follows on its own — read the colour of a tile you know the model scores, and read the number on the rest.

Worth being exact about one more thing, because it is easy to assume otherwise and the assumption changes how you read the page. No weight is printed anywhere on it. Each tile's weight travels in the page's data payload and is never rendered, and the list under the gauge is not a weight list either — it ranks contributors by weight multiplied by the size of the move, so a lightly weighted input pinned at the clamp can appear above the dollar on a day the dollar barely moved. That list describes the day's arithmetic, not the standing importance of an input. Every weight quoted in this article comes from the source, not from the screen.

The frame's real weakness is that it always has an answer

Everything so far has been about cases where the axis mislabels. The deeper problem is structural, and it survives every improvement you could make to the labelling.

Risk-on and risk-off is a description that is very easily mistaken for a cause. “Indian equities fell because it was a risk-off day” contains no information: the fall is part of what makes the day risk-off. The sentence restates its own evidence and sounds like an explanation, which is the most comfortable kind of wrong. A frame with a name for every possible tape can never be contradicted by one, and a claim that cannot be contradicted is not telling you anything.

So where does the frame earn its keep? Only when it commits to something you have not looked at yet. Read the equity tape and infer that credit should have widened and insurance repriced — then go and look. If they did, the one-axis reading survived a test it could have failed, which is what makes it worth something. If they did not, you have learned the more valuable thing: this day has a different shape, and the label would have hidden it.

That is the honest form of the frame. It is a consistency check across a tape you have not finished reading, and it is worth nothing as an account of the part you already read. Used the first way it is one of the most efficient ideas in markets. Used the second way it is a vocabulary for sounding informed about a day nobody has explained, and it costs you the moment when the pattern broke and you were told a story instead.

Which leads to the boundary this article will not cross. A cross-asset read 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 an entry or an exit. What a drawdown does to a long-horizon plan is a different question with a different answer, worked through in asset allocation and in what happens when you stop a SIP in a drawdown.

Four ways this goes wrong in practice

1. Treating the label as the mechanism. “Risk-off” names a pattern across holdings; it is not an agent and it does not act on anything. If you cannot say which channel carried the move into the price you are looking at, you have a word rather than an explanation.
2. Reading a small composite as a calm day. Every input is clamped before it is weighted, so a near-zero score is produced both by a quiet tape and by channels tearing in opposite directions. The number cannot distinguish them. The named contributors beside it can.
3. Reading colour as direction. Green means supportive for Indian equities, not up — except on the unweighted price tiles, where it does mean up, which is how WTI and Brent end up disagreeing about the same barrel on the same screen.
4. Promoting one day to a regime. Our labels attach to a score crossing ±20 on the day, and those cut-offs are a design choice. Nothing in the construction makes a single crossing a state of the world, and nothing in this article says what follows a crossing.

There is a fifth that is quieter and costs more. The frame is memorable, which means you will start seeing it in tapes that only half fit — and the half that does not fit is precisely the part carrying information you do not already have. The discipline is to look for the limb that contradicts the label before accepting it, which is the opposite of how a satisfying explanation is normally assembled.

Where this sits in the app

Testing a one-axis read means having the channels that could contradict it on the same screen, 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 in five sections — currencies; rates, bonds and credit; commodities; volatility; global equities — each carrying a one-line reason and a longer transmission note into Indian equities behind a Why it matters toggle, the composite score with the day's largest contributors named in words, a separate stress state on the cut-offs stated above, foreign and domestic institutional cash, central-bank policy rates with their stance and their three-month change, a correlation panel over 30, 60 and 90-session windows, and a per-tile history chart at one month, six months, one year and five years so a move can be read against its own range.

On the correlation panel, one caution that belongs with this article. It is a Pearson correlation of daily returns between Nifty and a fixed list of ten drivers, computed over the dates each pair shares. A correlation is a property of a window, not of a day and not of a mechanism — a coefficient that flips sign between the 30-session and the 90-session window has told you about the window. It also cannot establish direction: two assets move together on these days because both respond to the same decision by the same holders. Use it to check whether the recent record is consistent with a mechanism you can already state, and not to discover one.

The weights, signs, regime boundaries and stress triggers are the ones in this article, and they are ours.

Common questions

Continue reading

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