Rand Dollar Digest
Week ending 17 July 2026 · Spot 16.41 · Fair value 16.66
Where the rand stands
The rand is trading at 16.41 to the dollar in the week just ended. Over June it averaged 16.40, and our fair-value model — a long-run relationship between the exchange rate and relative prices, money, activity and interest rates — puts equilibrium for that month at 16.66.
That leaves the rand -1.6% away from fair value: essentially at equilibrium.
| Against fair value | Rand per dollar |
|---|---|
| +10% | 18.41 |
| +5% | 17.51 |
| +2% | 17.00 |
| fair value | 16.66 |
| the rand now (-1.6%) | 16.40 |
| -2% | 16.33 |
| -5% | 15.85 |
| -10% | 15.07 |
A gap this small is well inside the model’s normal scatter (one standard deviation is about 13%), so it is not a trading signal — it says the rand is broadly where the fundamentals put it.
The forecast
Weekly
| Horizon | Target | Forecast | 90% range |
|---|---|---|---|
| 1 week (nowcast) | 20 Jul 2026 | 16.42 | 15.97 – 16.90 |
| 2 weeks | 27 Jul 2026 | 16.42 | 15.77 – 17.20 |
| 3 weeks | 03 Aug 2026 | 16.42 | 15.63 – 17.33 |
| 4 weeks | 10 Aug 2026 | 16.43 | 15.51 – 17.50 |
Monthly
| Horizon | Target | Forecast | 90% range |
|---|---|---|---|
| 1 month (nowcast) | 01 Jul 2026 | 16.44 | 15.65 – 17.35 |
| 3 months | 01 Sept 2026 | 16.55 | 14.86 – 18.36 |
| 6 months | 01 Dec 2026 | 16.69 | 14.38 – 19.68 |
| 12 months | 01 Jun 2027 | 16.98 | 14.27 – 19.75 |
A horizon marked nowcast is a period already under way — part of it is observed, so it is an estimate of a partly-known average rather than a forecast of an unknown one.
The ranges are empirical: they come from how far this combination’s forecasts have actually missed in the past, at that horizon, not from a model’s own assumption about its errors.
The picture
What is moving it
Fair value is not a mood; it is an arithmetic consequence of four things. Because the model is linear, we can split the past year’s move in fair value into exactly what each driver contributed.
| Driver | Moved by | Pushed fair value |
|---|---|---|
| SA vs US inflation | +1.5% | +2.9% |
| SA vs US money supply | +4.1% | -2.2% |
| SA vs US output | -4.7% | +2.0% |
| Interest rate gap | +0.45 pp | +0.9% |
| Net | +3.6% |
The spread across the suite is itself informative — when the models disagree, the outlook is genuinely uncertain rather than merely uncertain-sounding.
| Horizon | Lowest | Highest | Spread |
|---|---|---|---|
| 1 month | 16.40 | 16.66 | 1.6% |
| 3 months | 16.40 | 16.66 | 1.6% |
| 6 months | 16.40 | 16.86 | 2.8% |
| 12 months | 16.40 | 17.33 | 5.7% |
How the models have been doing
We publish the scoreboard whether or not it flatters us. The benchmark is the random walk — the assumption that tomorrow’s rate is today’s. Beating it consistently is famously hard, and mostly we do not.
| Model | 1 mth | 3 mth | 6 mth | 12 mth |
|---|---|---|---|---|
| Random walk (benchmark) | 1.00 | 1.00 | 1.00 | 1.00 |
| ARIMA | 0.99 | 1.00 | 0.99 | 1.02 |
| Exponential smoothing | 0.99 | 1.05 | 1.10 | 1.12 |
| Local trend | 1.01 | 1.02 | 1.01 | 1.06 |
| Fundamentals (error-correction) | 1.00 | 1.03 | 1.03 | 1.01 |
| Fundamentals (system) | 2.57 | 1.58 | 1.37 | 1.21 |
| Fair value | 4.13 | 2.29 | 1.76 | 1.37 |
Two honest points about that table. Most of the suite sits at or just above 1.00, which is what the exchange-rate literature has found since Meese and Rogoff in 1983. And our combined forecast is not reliably better than the single best model — its value is that you do not have to know in advance which model that will be.
Where the suite does earn its keep is the fair-value gauge above: not as a prediction of next week, but as a statement about whether the currency is expensive or cheap against its fundamentals.
Method and data
| Series | Latest month |
|---|---|
| SA consumer prices | Jun 2026 |
| SA money supply (M3) | May 2026 |
| SA manufacturing output | May 2026 |
| SA mining output | Apr 2026 |
| US consumer prices | Jun 2026 |
| US money supply (M2) | May 2026 |
| US industrial production | Jun 2026 |
Exchange rate data from SARB via EconData; US series from FRED. The target is the average rate over the period, not a closing snapshot. Models are estimated on logs and forecasts reported in levels. Fair value is a long-run cointegrating relationship estimated from 2002. Forecast ranges are empirical quantiles of past out-of-sample errors.
Rand Dollar Digest · produced 24 July 2026 · vintage: week ending 17 Jul 2026, fundamentals to Jun 2026