Sources of Real Exchange-Rate Fluctuations: How Important Are Nominal Shocks?

The paper positions itself directly between two camps: Mussa (1986), who “argued that sluggish price adjustment must play a central role in explaining the short run movements in real and nominal exchange rates” based on the fact that real and nominal exchange-rate volatility have closely tracked each other since Bretton Woods, and Stockman (1987), who argued the “equilibrium approach” implies this co-movement could instead “reflect…the influence of real shocks with large permanent components” (Introduction). The authors note that findings of a unit root in real exchange rates and a weak link between real interest differentials and real exchange-rate changes (Huizinga 1987; Campbell and Clarida 1987; Meese and Rogoff 1988) had “tended to cast doubt on the relevance of the basic ‘sticky price’ open macro model,” with Meese and Rogoff concluding bluntly that their evidence provides “no support whatsoever” for models emphasizing sticky prices and monetary disturbances.

Revisiting the Campbell-Clarida (1987)/Meese-Rogoff (1988) approach – generalized “to allow for a much wider class of stochastic processes for the real interest differential” using the Beveridge-Nelson-style forward solution q_t^rr = q_t − q̄_t = −E_t Σ r_{t+j} (Eqs. 2-6) – the paper confirms the earlier null result: “the transitory component in dollar real exchange rates implied by the time series on real interest differentials accounts for only a trivial fraction of the variance in the log change in real exchange rates” (Section on Real Interest Rates). This sets up the paper’s key methodological move: if the interest-parity-implied transitory component is trivial but a multivariate Beveridge-Nelson decomposition (adding lagged inflation differentials and relative output growth to the information set) finds a much larger transitory component – which the authors do find, 70% of the variance of the dollar-DM change and 59% of the dollar-yen change – then the earlier null result reflects an information-set problem, not the absence of a quantitatively important transitory (potentially nominal) component.

The model is “a stochastic version of the two country, rational expectations open macro model developed by Obstfeld (1985)” that “exhibits the standard Mundell-Fleming-Dornbusch results in the ‘short run’…but it also embodies the ’longer run’ properties that characterize macroeconomic equilibrium…once prices adjust fully.” In the flexible-price equilibrium the system is triangular (Eq. 17): only supply shocks z influence the long-run level of relative output; both supply and demand shocks (d) influence the long-run real exchange rate; and all three shocks – including money shocks v – influence the long-run relative price level. Under sluggish price adjustment (indexed by a parameter θ), the actual real exchange rate deviates from its flexible-price path by a term proportional to the gap between actual and market-clearing prices (Eq. 19), and the model exhibits “the renowned Dornbusch (1976) overshooting result”: the nominal exchange rate overshoots its long-run response to a monetary shock whenever a stated parameter condition holds, while real supply/demand shocks instead produce undershooting of the real exchange rate relative to its flexible-price path.

Using the Blanchard and Quah (1989) approach, the paper exploits exactly three long-run restrictions implied by the model’s triangularity: neither money nor demand shocks have a long-run effect on relative output (C₁₂(1) = C₁₃(1) = 0, Eq. 26), and money shocks have no long-run effect on the real exchange rate (C₂₃(1) = 0, Eq. 27). These three restrictions are exactly enough to identify the contemporaneous structural matrix C₀ from the estimated reduced-form VAR’s variance-covariance matrix (Eqs. 21-25), without requiring any assumption about the dynamic path or the contemporaneous exclusion of shocks from particular equations – the paper’s stated “important advantage” of this identification scheme over more restrictive alternatives (the system is estimated as a trivariate VAR in output, the real exchange rate, and inflation for each of the US-Germany, US-Japan, US-Britain, and US-Canada pairs, 1973-1992).

The headline result is a sharp split: “more than 41 percent of the unconditional variance of the change in the dollar-DM real exchange rate, and more than 35 percent of the variance of the change in the dollar-yen real exchange rate are attributed to nominal shocks” (Tables 3-4), while for Britain “only 2.8 percent” and for Canada “only 2.5 percent” of the unconditional variance of the exchange-rate change is attributed to nominal shocks. At a 4-quarter forecast horizon, more than 50% of the German and nearly 35% of the Japanese real-exchange-rate forecast variance is due to nominal shocks, versus 1.3% for Britain and under 1% for Canada. For all four countries, “of the remaining variance…that is not attributed to monetary shocks, virtually all is attributed to demand shocks, and virtually none is attributed to supply shocks” – for Germany specifically, of the 49.6% of 4-quarter variance not due to money, 42.7 points come from demand shocks and only 6.9 from supply shocks.

The paper decomposes the actual path of each real exchange rate into the cumulative contributions of each shock type since 1974. It finds “virtually all of the dollar’s real depreciation against the DM in the late 1970s is attributed to nominal shocks, while most of the dollar’s real appreciation against the DM in the first half of the 1980s is attributed to demand shocks” – with a sizable portion of the early-1980s appreciation against both the DM and yen also attributed to nominal shocks (consistent with the well-documented US monetary disinflation of that period), and the post-Plaza-Accord dollar depreciation against the DM split roughly evenly between nominal and demand shocks (while for the yen it is attributed almost entirely to demand shocks). For Britain and Canada, by contrast, “the time path of the real exchange rate is essentially the time path attributed to demand shocks.”

Checking impulse responses against the Mundell-Fleming-Dornbusch predictions, the paper finds nominal shocks for both Germany and Japan match closely: “an initial 3.8 percent real depreciation of the dollar, a nearly 0.5 percent rise in US output relative to German output, and a 0.3 percent rise in US inflation” for Germany, with the nominal exchange rate overshooting its long-run response and depreciating further for several quarters before reversing – a delayed-overshooting pattern the paper notes “has been recently uncovered using a completely different empirical strategy by Eichenbaum and Evans (1993).” Demand shocks also match prediction closely (real appreciation, higher output, higher inflation, in both the short and long run). Supply shocks are the clear exception: “Although US prices fall initially in response to the supply shock, the dollar appreciates in real terms, in contradiction to the open macro model presented above” for the US-Germany and US-Britain pairs, while for Japan and Canada the supply-shock response does match the model’s prediction of real depreciation – the paper reports this discrepancy plainly rather than suppressing it.

The authors benchmark against Eichenbaum and Evans (1992), who isolate US monetary-policy innovations directly (via the federal funds rate or non-borrowed reserves) rather than via long-run restrictions, and find their own 12-quarter-ahead estimates (28% for the DM, 15% for the yen) fall “near the respective midpoints” of Eichenbaum and Evans’s range (13-42% and 13-23%). Against Lastrapes (1992), who uses a Blanchard-Quah bivariate VAR that can identify only a single “real” shock rather than separate supply and demand shocks, Clarida and Galí find substantially larger nominal-shock contributions (45% vs. Lastrapes’s 27% for the 4-quarter DM variance; 34% vs. 6% for the yen) and suggest this gap may reflect Lastrapes’s own acknowledged vulnerability: “if the world is subject to more than a single real shock, his identification strategy is potentially compromised.”

The authors conclude that the dispute is not an either/or: “these findings…suggest to us that sluggish price adjustment must be incorporated into any effort to explain short run fluctuations in real exchange rates (Mussa (1986)),” but “according to our estimates, real shocks to supply and demand account for more than 50 percent of the variance in forecasting real exchange rates – except at short horizons for the dollar-DM real exchange rate,” so that “Campbell and Clarida (1987) and Stockman (1987)…‘real’ explanations for real exchange rate volatility…[are not] misplaced.” The paper’s bottom line is therefore that nominal shocks matter a great deal for some currency pairs (Germany, Japan) at short-to-medium horizons without being the dominant source of variance anywhere, while demand shocks are the single largest source of real-exchange-rate variance across all four countries studied, and supply shocks are nearly irrelevant throughout.