Why foreign flows follow the dollar

The flow number is a decision taken in dollars

A foreign fund's return on Indian shares is earned in rupees and spent in dollars. The dollar's move is therefore part of its return before any Indian company enters the calculation — which is the channel commentary has in mind when it pairs the foreign flow print with the dollar, and the pairing is easier to over-read than almost anything else on a macro page.

Start with who is behind the number. The daily figure reported as foreign institutional investment is the net cash-market purchase of registered foreign portfolio investors: pension and sovereign funds, index-tracking vehicles, regional active managers, hedge funds. They differ in almost every way except one. None of them reports to its own investors in rupees.

That single fact does more work than it looks like it should. A fund reporting in dollars does not hold Indian equities. It holds a dollar position whose value happens to be produced by Indian equities and an exchange rate, and it is judged on the pair. The rupee is not a friction sitting between the fund and the return. It is half the return.

This article traces what follows from that, and only that. What the print counts, what it structurally cannot see, and how the foreign and domestic figures relate to each other belong to the flows article; the dollar's four channels into Indian share prices belong to the dollar index article; the Macro page as a whole belongs to reading the macro signals together. None of the three is repeated here.

Two legs, approved by two different people

The realised arithmetic is short: a dollar-based holder earns the index move multiplied by the rupee's move against the dollar, and a domestic holder earns only the first. That multiplication, worked through with illustrative numbers, is in the dollar index article and there is no point restating it.

The part worth adding is that the two legs are not merely multiplied together at the end. They are approved separately at the beginning, by different teams, in a fixed order.

At a large allocator the currency and country exposure is typically set at one level and the security selection at another. A macro or asset-allocation committee decides how much of the book sits in emerging-market equity and, within that, how much sits in India. A regional equity team then decides what to own inside whatever sleeve it has been given. The first decision is denominated in dollars and framed against dollar alternatives. The second is a view about companies.

So the ordering is: currency and country first, companies second. A change in the first resizes the second without anyone revisiting it. That is a statement about how the institution is wired, not about how often the wiring fires.

The funding leg — the hurdle rises for every allocation at once

The currency leg is only half of what a dollar move does to a foreign allocator. The other half is the cost of the money.

A stronger dollar ordinarily arrives alongside tighter dollar funding conditions, because the same circumstances tend to produce both: a higher return available on dollar cash, scarcer dollar liquidity in the banking system, or a general preference for the currency that everything is priced in. That is the conventional reading and the mechanism behind it is straightforward, but it is a reading rather than a law — the dollar can rise on a domestic growth surprise with funding conditions perfectly easy, which is the “two dollars” problem the dollar index article sets out.

Where funding does tighten, three things happen to an allocator at once, and none of them is about India:

• The alternative improves. The return available for taking very little risk in dollars is the baseline every risk allocation is measured against. When the baseline rises, the extra return demanded from emerging-market equity rises with it — the same discounting arithmetic worked through in US yields and Indian valuations.
• Leverage costs more. Any holder financing a position in dollars pays more to keep it, which is the mechanism behind the carry trade in its yen-funded form and behaves the same way when the funding currency is the dollar.
• Redemptions arrive at the wrong end. A fund meeting withdrawals sells what it can sell cleanly. Large liquid emerging markets are not sold because they are disliked; they are sold because they can be.

The important word is “every”. The hurdle does not rise for India specifically. It rises for the whole emerging-market allocation simultaneously, which is why the flow number can turn negative across several countries in the same week with nothing in common between their domestic circumstances. India is not being compared with its own past on those days. It is being compared with dollar cash.

The sale that has no opinion in it

Here is the consequence that most commentary on the flow print misses entirely.

When a top-down allocation is cut, the execution is usually proportional. A sleeve reduced by a tenth is sold down roughly across what it holds. Nobody in that chain formed a view on any Indian company; the instruction was to hold less India, and the instruction came from a committee that was looking at the dollar, dollar rates and the relative case for emerging markets as a block.

So a large foreign sale is compatible with foreign investors thinking exactly what they thought last month about Indian earnings. The sale carries no information about companies, because no company was assessed in producing it. Reading “FIIs are bearish on India” off a negative print imports an opinion into a decision that may not contain one.

That is the specific mistake to watch for, and it has a tell. Ask what changed. If the answer is a dollar move, a rates move or a global risk event, the flow is the allocation layer speaking. If the answer is an Indian earnings season, a policy change or a domestic shock, the flow may genuinely carry a view about India. The same number means different things depending on which layer produced it, and the number itself cannot tell you which.

The honest limit on all of this: a print is a net figure. Buyers and sellers with entirely different reasons are aggregated into one rupee number, so “the flow was allocation-driven” is an inference about the balance of activity and never a reading off the print.

Two decision sheets for the same shares

The clearest way to see why the two investors behave differently on identical news is to write out what each of them is actually deciding.

Read the last row again, because the rest of the article turns on it. For the foreign holder, the equity transaction and the currency transaction are the same decision executed in two markets, not two events that happen to coincide. The exception proves the point rather than softening it: a holder who leaves the proceeds in rupees, or who sold the currency forward months earlier, has moved the currency leg to a different date — it has not stopped existing.

Why the co-movement cannot settle the direction

Everything above describes a channel running from the dollar to foreign flows. Commentary routinely reports the dollar index and net foreign flow leaning the same way, and treats the pairing as settled. The co-movement does not settle it, and the reason is worth being precise about.

Three different arrangements produce the same picture:

• Dollar to flows. The channel this article has described: dollar strength raises the hurdle and lowers the rupee leg, and the allocation shrinks.
• Flows to dollar. A foreign fund leaving India sells rupees for dollars. That is dollar demand created by the exit itself. The flow is not responding to the currency move; it is part of it.
• A third thing to both. A global risk event moves the dollar and moves every country allocation, with no causal link between the two at all — the configuration described in what a risk-off day looks like.

All three fit the data equally well, because the data is two series that lean together. Correlation between the dollar and foreign flow is consistent with the funding-and-hurdle channel and is not on its own evidence for it. A mechanism earns belief when it implies something checkable beyond the co-movement it was invented to explain — whether the effect shows up across other emerging markets at the same time, whether it appears in the hedged and unhedged parts of the flow differently, whether it survives episodes when the dollar rose for a benign reason.

The practical consequence for reading the page: the flow panel and the rupee are not two independent confirmations. A single large foreign sale moves both, so seeing them agree can feel like two witnesses when it is one event recorded twice. Tiles agreeing is informative in proportion to how separate the mechanisms behind them are. Here they are the same mechanism.

What a currency hedge removes, and what it leaves

The obvious objection is that a foreign fund can simply hedge the rupee and make the first leg go away. Many do, and it works — for that leg. The cost of doing it is the gap between Indian and US interest rates over the period, which is set by arbitrage rather than by anyone's view, and which the dollar index article works through.

A hedge removes the currency leg. It does not remove the funding leg. A fully hedged foreign fund still measures Indian equity against dollar cash, still pays more to finance a leveraged position when dollar funding tightens, and still sits inside an allocation decision taken in dollars one level above it. Two legs, one hedge, and the leg that survives is the one that operates on the allocation rather than on the translation.

This has a consequence for the print. Hedged and unhedged foreign money respond to the same dollar move by different amounts, and the published net figure aggregates them without distinguishing them. The flow number is a sum over holders with materially different exposures to the thing being used to explain it.

The trade-off cuts the other way too, and it is worth stating because the domestic investor's position is usually presented as a straightforward advantage. Having no currency leg is not free. It means the assets, the income and the liabilities all sit in one currency, so a rupee that weakens raises the cost of imported goods, foreign education and travel with no offsetting gain anywhere in the portfolio — the exposure set out in USD/INR and your portfolio. The foreign holder carries a currency risk and gets currency diversification with it. The domestic holder carries neither. Those are different positions, not a better and a worse one.

What our Macro page actually does with the flow number

The flow figure appears twice on FNOTrader's Macro page — once as a panel of its own, and once inside the composite score. The two are worth separating, because the panel is a report and the score is an opinion.

The panel shows the most recent published session's net foreign and net domestic cash purchases in ₹ crore, net meaning buy minus sell. The composite folds in the foreign number only. The domestic figure is displayed and never scored — it is there because a market where foreign selling is being absorbed by domestic buying is a different market from one where both are selling, which is a reading the score has no way to express.

Four properties of the scored input decide how much weight it deserves in your reading. All four were read from the running code rather than described from memory. The first two are judgement calls we made; the second two are consequences of how the score is assembled:

• Weight 0.12. Applied to the foreign net figure. The five heaviest 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 US volatility index at 0.10; the other fifteen carry less. So the flow input sits level with Brent crude and below the dollar.
• Saturation at ±₹5,000 crore. A larger net figure counts exactly the same as one at the ceiling, so an extraordinary session cannot dominate a cross-asset average. What that ceiling costs you in exchange is worked through in the flows article.
• It is not a tile. The twenty tile-table inputs carry weights summing to 1.33; the flow input is folded into the score at runtime and added to both the top and the bottom of the average. Twenty-one inputs, and a divisor of 1.45 when everything reports.
• The divisor moves. A failed feed drops out of the numerator and the denominator together, so 1.45 is the all-present maximum rather than a constant, and every remaining input's share of the score rises when something is missing.

Now the arithmetic that reframes the whole thing, and it is short enough to redo. The score runs from −100 to +100 and is 100 × Σ(wₓ · cₓ) ÷ Σ(wₓ), with every contribution clamped to the range −1 to +1. So the most the flow input can push the score away from wherever the rest of the board has put it, on a day when everything reports, is 100 × 0.12 ÷ 1.45 — about 8.3 points on a 200-point scale. The number Indian market commentary discusses more than any other is worth roughly a twelfth of our composite. The dollar, at 0.20, tops out close to 14 points on the same arithmetic. Both shares rise when feeds fail, because the divisor shrinks.

Those weights and that ceiling are FNOTrader's modelling judgement, not measurements. Nobody has established that foreign flow is worth 0.12 of anything. A different considered view would set it differently and would not be wrong. The same applies to the regime cut-offs at plus and minus 20 and to the stress override, whose mechanics belong to the pillar article.

Two more things about that panel, both easy to miss. The flow input is not live. Every price tile beside it updates through the session; the flow figure is the most recent published session found in a twelve-day lookback, so it can be a session or more behind the prices it is being read against. And on every tile that carries a weight, colour marks the scored effect on Indian equities rather than the direction the number moved — which is why a rising USD/JPY shows green: a weak yen is the condition under which the yen-funded carry trade stays intact.

The relationship this article describes is not on the heatmap

A reader who has followed the argument will reasonably go looking for it in the correlation panel. It is not there, and the reason is a fact about the panel's construction rather than about the world.

The panel computes Pearson correlation on daily returns of Nifty against a fixed list of ten drivers — the dollar index, the US 10-year, Brent, USD/JPY, gold, the US volatility index, copper, the S&P 500, the Nasdaq and USD/INR — over a window of 30, 60 or 90 days, pairwise on the dates the two series share. Nifty is the only left-hand side, and net foreign flow is not one of the ten drivers.

So the dollar-to-flows relationship cannot be read off the grid at all. What the grid can show is the dollar's relationship with the index, which is the far end of the chain and includes whatever else was happening. Using it as though it measured the flow channel is reading one link of a chain and reporting it as the chain.

Two limits apply to whatever you do read there. A correlation flipping sign between the 30-day and the 90-day window is describing the window, not a change in how the world is wired — a fortnight of large days can set a short coefficient by itself. And Pearson on daily returns is silent on which series moved first and on whether a third thing moved both, which is precisely the ambiguity the previous section could not resolve.

The usable discipline is the same one the pillar states: name the channel first, then check whether the correlation is consistent with it. Never in reverse.

Three ways this gets misread

Each of these is a specific error with a specific correction, and all three are common enough to be worth naming.

Reading one day's print as a verdict. A single session's net is one number aggregating buyers and sellers with unrelated reasons, published for the cash segment alone. Foreign positioning also runs through index and stock futures, options and primary market participation, none of which is in the figure — the coverage limits are set out in the flows article. A large cash sale is evidence about the cash segment and not proof that total exposure fell.

Treating flow and currency as two confirmations. They are one event observed twice whenever the flow is large, because a repatriated foreign exit sells shares and rupees in the same decision. Agreement between two tiles is worth what the independence of their mechanisms is worth.

Inferring a company view from an allocation decision. A proportional sleeve reduction is executed without anyone assessing an Indian company. When the trigger was a dollar or global-rates event, the flow is the allocation layer speaking and carries no earnings opinion — and the crude channel, which runs through the current account and inflation, is a genuinely India-specific driver that gets confused with it constantly.

What all three share is a missing question: at which layer was this decision taken? The number is identical either way. The interpretation is not.

Where to look at this

The reading is the part that has to be yours. Keeping the series side by side, on one scale, rather than reconstructing them from separate screens, is what the Macro page in FNOTrader's Options Analytics app does.

It shows the foreign and domestic cash-market net for the latest published session in ₹ crore alongside the cross-asset tiles, each weighted tile coloured by its scored effect on Indian equities rather than by the direction it moved; the composite with its regime label; the rationale lines naming the largest contributors to the current score; and a Pearson correlation grid on daily returns of Nifty against ten drivers with a selectable 30, 60 or 90-day window, so a relationship can be checked on more than one sample before it is believed.

One thing the page does not put on screen is the weight table itself — it names the heaviest inputs in words and leaves the numbers to the payload. That is why the four figures above are set out in this article: the weights and the saturation ceiling are our judgement, and a judgement you cannot see is one you cannot argue with. The correlations are arithmetic on price series and carry the limits described above whoever computes them.

Common questions

Continue reading

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