Photo: Reuters/Thomas White
Photo: Reuters/Thomas White

The dollar’s surge during the past two weeks appears to reflect an overall tightening in US financial conditions. There are many reasons for other currencies to be weak: uncertainties about the future Italian government, the policy meltdown in Turkey, the stalled NAFTA negotiations with Mexico, sanctions against Russia, and so forth. But the main reason for the dollar’s strength appears to be a shift to risk aversion in the United States.

The Goldman Sachs Financial Conditions Index is a black-box product derived from econometric analysis and is therefore opaque, but it has the advantage of including a very large variety of sensitive credit variables as well as liquid rates measures. A tightening of financial conditions starting at the end of January (reflecting the pullback in US equity markets among other things) anticipated the dollar’s rise.

This goes well beyond expectations of Fed tightening as reflected in the yield curve. As we see in the chart below, the flattening of the US swaps curve tracked an improvement in US financial conditions until early this year. Starting in March the two measures diverge.

Disentangling the components of the GS index is difficult, but some of the most important components (referring to the 2010 Jan Hatzius-Frederic Mishkin discussion of the index) are items such as consumer credit and ABS issuance. Both have been weak recently.

I have the impression that a general drift towards risk aversion rather than any single phenomenon is at work. Whatever the cause, the GS Financial Conditions Index has considerable predictive value for the euro exchange rate, along with the respective US and European swap rate slopes.

Between late 2016 and the beginning of May 2018, the respective 2-year/5-year swaps slope in euro and USD predicted the EUR exchange rate vs. USD with great accuracy. The dollar strengthened much further than the yield curve relationship predicted during May. Including the financial conditions index eliminates most of the model error.