In yesterday’s FT, “All in this together” assessed the possibility of a roughly synchronized downturn in the world’s major economies, with the United States, ironically enough, suffering the smallest hit. This brings up all sorts of interesting questions regarding exchange rates, if one believes that Taylor rules define monetary policy making to some degree, and that interest differentials affect exchange rates.
First, consider the evidence. The Euro area and Japan experienced negative growth in 2008Q2. The US experienced rapid GDP growth, but the consensus is for far more tepid growth in 2008Q3 (John Kitchen‘s estimate based on data available on 9/5 indicates 0.3% SAAR growth, although the underlying growth rate is 1.7%).
To the extent that monetary authorities respond to economic slack with lower interest rates, then one should anticipate lower interest rates in the Euro area and Japan as slow growth drives the output gaps negative in these regions (for a discussion of the distinction between growth and output gaps, see this post). However, it’s not clear to me that contemporary (i.e., in the past ten years) Japanese monetary policy is adequately modeled in such a way, so let’s take that off the table (but we can add in the UK).
In this post, I discussed how well a Taylor rule based model explained movements in the USD/EUR exchange rate. To the extent that this type of model works well for USD/GBP as well, what we should expect to see is a strengthening of the dollar as Euro area and UK interest rates fall. Here’s the 4 quarter ahead forecast from a slightly revised version of the model:
st-st-4 = 0.060 – 3.119 ogt-4 – 6.703 pit-4 – 0.020 (qt-4-0.109) – 6.546 ((qt-4-0.109)3) + 1.25 it-5 + ut
where og is the relative output gap, and pi is relative inflation, and q is the real exchange rate gap. The adj.R2 = 0.43, SER = 0.076, sample 1994q1-08q2; bold face indicates significance at the 10% level [update: using HAC robust standard errors].
If the reaction functions stay unchanged, then the implied 4 quarter ahead forecast is as depicted in Figure 1, i.e., a dollar appreciation over the next year ($1.42 by end-June 2009, vs. $1.58 end-June 2008).
Figure 1: US-Euro Area output gap (blue, left axis), log dollar/euro rate (red, right axis), in-sample 4 quarter ahead prediction from Taylor rule fundamentals based on 2008Q2 data (red square, right axis). NBER defined recession dates shaded gray. Output gaps estimated as deviations from quadratics in time. Source: author’s calculations.
Here begins the speculative portion of this post (so no predictions here — just some musings.)
There is an interesting feature of the model, insofar as the target real exchange rate is included in the central bank reaction functions. As long as the target real exchange rate remains constant, the model should be stable; and indeed estimation of the model is predicated upon no-parameter-shifts. However, should the target shift (let’s say be depreciated from the US perspective), this will work to offset the Euro area and UK currency depreciation, and hence prevent expenditure switching toward those regions.
Now, would the target real exchange rate ever shift for the US? Followers of the Taylor rule literature know that shifts in targets are not typically contemplated. And indeed, it is debatable whether the Fed includes such a target in its Taylor rule.
That being said, we know that net exports accounted — in a mechanical sense — for almost all of the GDP growth in 2008Q2 . About half of that is from import compression, and the other half is from export growth. To the extent that this is the only sector in the US sustaining growth, one could see the temptation to allow for a looser monetary policy. And as growth slackens in the rest-of-the-world(see Figure 2), surely the temptation will only strengthen.
Figure 2: e-forecasting global activity index, and IMF industrial country index, Sept. 8, 2008.
Now consider what happens if the ECB or BoE also shift (depreciate) their target real exchange rates. That would mitigate the incipient dollar depreciation. In this case, interest rates are driven ever further downward — in the short run (with some lag, inflation will rise, stopping that effect).
This scenario is merely a modern version of the “competitive devaluation” phenomenon often used to interpet international monetary policies during the interwar period.
As I said before, no predictions here. Just something to think about. (As a parenthetical aside, if we end up extending EGTRRA and JGTRRA on top of adding in another $1.3 trillion in tax cuts, piled on top of whatever liability the Treasury will end up after the Fannie/Freddie bailout, then you can skip all the above text; you won’t need any monetary policy to drop the dollar’s value — all those structural budget deficits and associated debt will do the trick.)
[Update: 4:35pm, 9/11, posted in response to GNP's query]
Yes, there is serial correlation, as indicated by the standard Q-statistic. I should have mentioned that the standard errors used to judge statistical significance are HAC robust standard errors. Below is the requested graph.
Figure 3: Actual (red), fitted (green) and residual (blue) from equation.