A plain-English guide to the dollar smile.

DXY 99.03

A strategist will sometimes say the dollar is in a "left-side" regime, or that the smile has "flattened," and the reader is expected to know what they are talking about. This is the explainer. The dollar smile is a three-regime framework for how the US dollar behaves against the rest of G10. It was published by Stephen Jen at Morgan Stanley in the early 2000s, and it has held up better than most macro frameworks of the same vintage because the structure it identifies is real.

The three regimes

The framework takes one variable: the US economy's growth performance relative to the rest of the developed world. Call that g. Plot the dollar level (DXY is a fine proxy) on the y-axis and g on the x-axis. You get a smile.

1. Left side: global risk-off. When g is sharply negative, the world's growth is slowing, financial conditions are tightening, and capital flees into the safest paper. That paper is the US Treasury. Demand for dollars rises because foreign investors are net buyers of US safe assets. The dollar strengthens on what looks like a contradiction: US growth is bad, but the dollar is bid. The mechanism is safe-haven flow, not US rate-differential math.
2. Middle: US underperforms. When g is modestly negative or close to zero, growth is mediocre but not panic-inducing. Investors look outside the US for yield, into Europe, EM equities, Asian credit. Dollars get sold to buy foreign assets. The dollar weakens. This is the trough of the smile.
3. Right side: US exceptionalism. When g is strongly positive, the US is growing faster than the rest of G10. Real rates rise, equities outperform, foreign capital flows into US assets for yield, and the dollar strengthens. The mechanism is rate-differential and capital-account, not safe-haven.

The picture

The math, such as it is

The original Jen framework was qualitative; the version a desk uses today is a regression. A simple working model is:

ΔDXY = α + β₁·g + β₂·g² + β₃·VIX + ε

Where g is the US growth differential and the squared term gives you the smile shape: small g in either direction is dollar-positive on net because both ends of the parabola lift the y-axis. VIX picks up the risk-off intensity so you can separate the left-side flow from the middle. Fit on monthly data over the post-1995 sample, the coefficient on g² is consistently positive and significant at 5 percent; the coefficient on the linear term is small.

Three operational reads come out of the regression:

• The trough of the curve sits a touch to the right of zero, around US-vs-G10 of plus a quarter point. That is the modal regime in the post-1995 sample: the US grows a shade faster than the average G10 country, but not enough to pull capital in.
• The left tail kinks up faster than the right tail, because safe-haven flows are non-linear. When the VIX crosses about 30, the safe-haven multiplier on the left side compounds. That is why crisis episodes show much sharper dollar moves than the slow grind of US outperformance.
• The right tail is sensitive to the rate-differential channel. Each 25 basis points of real-rate divergence vs G10 buys roughly half a point of DXY on the historical fit.

Where DXY sits today

The latest read places the US growth differential modestly negative. The Q4 2025 advance pegged real GDP at 0.5 percent annualised, with the Atlanta Fed GDPNow model running Q1 2026 at 1.2 percent before the print rose to a 2.0 percent advance. Against a G10 average closer to 1.5 percent for the same period, the US sits in the trough-to-left-edge region of the smile. DXY at 99.03 is consistent with that read: weaker than the right-tail rate-differential pull would imply, stronger than a clean left-tail safe-haven spike would imply. The dollar is in the middle of the curve.

Stephen Jen himself flagged a structural risk to the left side earlier this month, writing that continued US fiscal slippage and a return to quantitative easing would erode the safe-haven premium the dollar collects in risk-off episodes. If that structural read is right, the left side of the smile gets shallower over time, and the dollar's downside in the next risk-off becomes worse than the historical sample suggests. We do not yet read that into the data; the fiscal premium is the risk to watch, not the base case.

What the smile does not tell you

1. Timing. The regime classification tells you what regime drives the dollar; it does not tell you when one regime hands off to another. The hand-off is usually visible only after the fact.
2. Magnitude. The fitted curve gives you a sign and an order of magnitude. It does not pin down the size of the move within the regime. Two left-side episodes can produce five-percent and fifteen-percent DXY moves; the difference is the size of the risk-off shock, not the framework.
3. Structural breaks. The smile assumes the US is still the world's reserve-asset anchor. That assumption is not metaphysically guaranteed. If the dollar's reserve status erodes, the left tail flattens and the framework needs adjusting. Jen's recent piece is exactly that adjustment in progress.
4. The 24-hour window. The smile is a multi-month framework. Daily dollar moves are dominated by flow, calendar releases, and positioning, not by where you sit on the curve.

Related reading

• Today's analysis: DXY finally moved, the live application of the framework
• The original dollar piece in this thread
• The correlation matrix, which reads cleanest through the regime lens
• The term premium piece, which connects to the right-side rate-differential channel