Tuesday , July 23 2019
Home / Econbrowser - James Hamilton / The New Fama Puzzle Persists

The New Fama Puzzle Persists

Summary:
In previous posts [1] [2], I described how in data up to the beginning of 2016, the Fama puzzle was overturned for major currencies. One question was whether the change would persist post-crisis and post-Zero Lower Bound (ZLB) exit. The short answer is, so far, yes. This is shown using data from a forthcoming paper w/Jeffrey Frankel (earlier version here), where we demonstrate that the reversal of the Fama regression coefficient holds, including when using forward rate data instead of interest rate data.  First, consider the Fama regression: s+1 – s = α’ + β'(i-i*) + error ε(s+1) is subjective market expectations of the future spot exchange rate (proxied using Currency Forecasters Digest/FT FX forecasts/FXForecasts survey data). Fama regression coefficients are the  β’ is the OLS

Topics:
Menzie Chinn considers the following as important: ,

This could be interesting, too:

Menzie Chinn writes Of Big Macs, PPP, the Penn Effect, and Currency Misalignment

Menzie Chinn writes To Worry or Not To Worry: Adjusted and Unadjusted Spreads

Menzie Chinn writes Guest Contribution: “A Blockchain solution for the technology war between China and the US “

Menzie Chinn writes Why Drop Rates?

In previous posts [1] [2], I described how in data up to the beginning of 2016, the Fama puzzle was overturned for major currencies. One question was whether the change would persist post-crisis and post-Zero Lower Bound (ZLB) exit. The short answer is, so far, yes.

This is shown using data from a forthcoming paper w/Jeffrey Frankel (earlier version here), where we demonstrate that the reversal of the Fama regression coefficient holds, including when using forward rate data instead of interest rate data.  First, consider the Fama regression:

s+1 – s = α’ + β'(i-i*) + error

ε(s+1) is subjective market expectations of the future spot exchange rate (proxied using Currency Forecasters Digest/FT FX forecasts/FXForecasts survey data).

Fama regression coefficients are the  β’ is the OLS regression coefficient. The estimated β’ for 1986-2008 and 2008M09 -2018M01 are shown in Figures 1 and 2, respectively.

The New Fama Puzzle Persists
Figure 1: Regression coefficient of 3 month annualized ex post depreciation on interest differential, for AUS, CAN, DNK, JPN, NZL, NOR, SWE, CHE, GBR, EUR, HKG, SGP, ZAF, all pre-crisis.

The New Fama Puzzle Persists
Figure 2: Regression coefficient of 3 month annualized ex post depreciation on interest differential, for AUS, CAN, DNK, JPN, NZL, NOR, SWE, CHE, GBR, EUR, HKG, SGP, ZAF, all post-crisis.

These results are robust to starting the post-crisis sample in 2010.

The key finding here, as in Bussiere, Chinn, Ferrara and Heipertz (2018), is that before the crisis, the correlation between ex post depreciation and interest differentials for advanced economies is negative. After the crisis, and during the zero lower bound (ZLB) era, it is positive. There are two non-reversal currencies (which were not included in the BCCH paper): Singapore dollar and South African rand. Note that these are emerging market currencies, and we do not expect to see the same factors and behaviors (neither at or near ZLB).

Currently, I’m at the NBER Summer Institute. Two excellent papers presented this morning in the International Finance & Macro meetings link up with these results, directly (Kalemli-Ozcan and Varela) and indirectly (Lilley, Maggiore, Neiman, and Schreger). For the latter, the question asked in the session is whether the results will persist far into the future (when many interest rates are above the ZLB) is also germane to our results.

(Of course, given how things are going, maybe there’ll be a lot of currencies returning to ZLB…)

Menzie Chinn
He is Professor of Public Affairs and Economics at the University of Wisconsin, Madison

Leave a Reply

Your email address will not be published. Required fields are marked *