Exchange rate forecasting

A currency forecast is needed whenever a return is earned in one currency and consumed in another, which is most institutional portfolios. Four inputs, each useful at a different horizon. Relative purchasing power parity says the exchange rate adjusts for the inflation differential:

E(%Δ spot) = (1 + π_domestic) / (1 + π_foreign) − 1

With Indian expected inflation at 4.50% and US expected inflation at 2.30%, that is 2.150538% of rupee depreciation a year, projecting an illustrative spot of 88.00 to 97.88 after five years and 108.87 after ten. Parity is close to useless over one year and is the best single anchor available over ten, exactly the horizon an assumption table cares about.

What other inputs drive exchange rates?

Relative economic strength works the other way and at shorter horizons. A currency with a higher real interest rate attracts capital and tends to appreciate, at least until the trade unwinds. India's ex-ante real ten-year rate of 2.00% against the US 1.90% gives a 0.10 point differential, which pushes the rupee opposite to the parity forecast and is small enough to leave the ten-year view intact.

Capital flows drive the level and the volatility rather than the trend. An economy financing itself with foreign direct investment is in a different position from one financing itself with portfolio money that can leave in a week, and the distinction is visible in the balance of payments rather than in the exchange rate. Savings-investment imbalances are the slowest input and the one that sets the sign of the pressure: a current account deficit is an economy investing more than it saves, and a deficit that must be financed every year is standing depreciation pressure regardless of what parity says.

How do you check and apply a parity forecast?

A covered-parity cross-check costs one line. Using the two ten-year yields as a proxy, the no-arbitrage forward premium on the dollar is (1.065/1.042) − 1 = 2.207294%, six basis points from the parity forecast, and those six basis points are the real-rate differential showing up where it should. Deflating both yields multiplicatively rather than by subtraction turns the 0.10 point differential quoted above into 0.057 points, which is what the forward is pricing. The forward prices the interest differential while parity prices the inflation differential, and the two agree only when real rates are equal.

The practical rule for a ten-year table is to use relative purchasing power parity as the central case, check that it does not contradict the real-rate differential by more than a point, and record the current account as the reason you would revise. The macro and corporate-issuer branches own the parity conditions and the balance-of-payments arithmetic in full.

One consistency check catches a surprising number of errors. A foreign return built in dollars and converted to rupees must be converted using the same two inflation numbers that sit inside the domestic bond and equity assumptions. Gold makes the point cleanly: a zero expected real return in dollars plus 2.30% US inflation gives 2.30% in dollars, and converting at the 2.150538% parity rate gives exactly 4.50% in rupees, precisely the Indian inflation assumption. If your table does not reproduce that identity, the inflation assumptions inside it are inconsistent.