The 90% Carry Trade: What Global Volatility Does to Returns

The carry trade is one of the oldest and most persistent sources of return in currency markets: borrow in low-yield currencies, invest in high-yield currencies, collect the differential. Done at scale, across a diversified basket of G10 pairs, it has generated positive returns over most multi-year rolling windows since at least the 1980s.

It also crashes. Spectacularly. 2008, August 2015, March 2020 — carry trades tend to unwind all at once, in a rush, with correlations that approach 1.0 as everyone exits through the same door at the same time.

The question is not whether carry works. The empirical record is clear: it does, over the right horizon. The question is what drives the crashes — and whether the crash signal is observable in advance.

The 90% Finding

Lukas Menkhoff, Lucio Sarno, Maik Schmeling, and Andreas Schrimpf (2012) sorted global currency pairs into portfolios by interest rate differential — the standard carry construction — and tracked returns over multiple decades. They then asked: what single factor best explains the variation in those returns across time?

The answer: global FX volatility. Measured as the average realized volatility across all currency pairs, this factor “captures more than 90% of the cross-sectional excess returns in five carry trade portfolios” — the published abstract’s own words. The quantity is the gap between the highest- and lowest-yielding baskets, which the paper’s conclusion states as more than 90% of the spread; it is not a share of any one currency’s return. What it is not is a single-factor claim: the cross-sectional R² above 90% belongs to a two-factor model, the dollar factor alongside volatility innovations, so crediting that figure to volatility on its own overstates it. When global FX volatility rises sharply, carry returns collapse. When it falls, carry performs well. That is a cross-sectional pricing result — it says volatility risk explains why some carry portfolios earn more than others — rather than a claim that volatility accounts for nine tenths of any single strategy's return through time.

This is not just a statistical curiosity. It is a structural finding about what carry trade returns are actually compensating investors for: volatility risk. Carry investors are, in effect, writing volatility. They collect the premium when conditions are calm and absorb the losses when volatility spikes.

The Treasury Market Link

Global FX volatility does not move in isolation. It is systematically correlated with broader cross-asset stress indicators, including treasury market volatility. When treasury market conditions become disorderly — when bid-ask spreads widen, when auction demand drops, when rate moves become large and unpredictable — FX volatility tends to rise in sympathy. This is why bond market conditions function as an upstream warning signal for carry crashes.

The first article in the Macro-to-FX Transmission Series covers this relationship in full. Bond volatility leads equity volatility in crises, which is why the treasury market is where this chain starts rather than ends.

Using This in Practice

What the result describes is a volatility-contingent return: across portfolios, the ones that pay most are the ones most exposed to unexpected volatility, and that exposure is the compensation rather than a free lunch. Carry is closer to a regime exposure than a directional view. What position size follows from that is not something this paper answers — it is a cross-sectional pricing result, not a timing rule, and nothing here recommends one.

The FX Monitor publishes a risk-regime state — the volatility reading, cross-asset signals, the risk-appetite picture — alongside the rate differentials. Menkhoff et al. establish that volatility risk is what carry portfolios are compensated for; they do not test the VIX as a timing signal, and no such claim is made here. The differential describes a structural force. The volatility state describes the environment it would have to express itself in. Both are published as conditions, and what to do about them is the reader's call.

The full carry trade framework, including the G10 sorting methodology and how to read historical carry return data, is in Article 9 of the transmission series.