Delayed Overshooting: Is It an 80s Puzzle?

Transcription

1 Delayed Overshooting: Is It an 8s Puzzle? Seong-Hoon Kim* Seongman Moon** Carlos Velasco*** *KERI **Chonbuk National University ***Universidad Carlos III de Madrid August 28, 26 (Asia Meeting, Kyoto)

2 Outline of the talk Delayed overshooting puzzle and Forward premim puzzle Our method (Volcker era (979-87) and post-volcker era (988-27)) Result : Exchange rates overshoot Result 2: A close link between conditional UIP and conditional exchange rate movements (Dornbusch Redux) Result 3: A close link between conditional UIP and unconditional UIP (Clue to the causes of the forward premium puzzle) Result 4: A close link between conditional and unconditional movements of exchange rates (Monetary Policy Effectiveness) Our explanation: Imperfect credibility of monetary policy

3 What made the Volcker era so extraordinary and overwhelming? US-AGG nominal exchange rates US long term interest rates and inflation rates Figure: Nominal exchange rates, long term interest rates, and inflation rates

4 Dornbusch Overshooting Hypothesis (976) Incorporate rational expectations and dynamics of exchange rates. One key assumption: UIP E t s t+ s t = i t i t Liquidity effect (money demand): m t p t = ηi t + φy t Prediction :exchange rate overshoots- in response to a domestic contractionary monetary shock, an increase in domestic interest rates relative to foreign interest rates leads to a large impact appreciation followed by a gradual depreciation of the domestic currency toward its long run value. Exchange rate volatility and persistence (key elements for DSGE models)

5 Delayed Overshooting Puzzle: an illustration

6 Delayed Overshooting Puzzle and Forward premium puzzle Delayed overshooting: Eichenbaum and Evans (995) using recursive identification methods in a unrestricted VAR framework In response to a contractionary US monetary shock, the US interest rate increases and the US dollar is gradually appreciated for a while (upto several years) and then depreciated later. Cause: E t s t+ s t = i t i t + d t where d t represents deviations from UIP. Implication: On average non-zero excess returns on foreign currency:(s t s t+ ) + i t i t >. A close link between failure from UIP (the forward premium puzzle) and the delayed overshooting puzzle: persistent appreciation of the US dollar Since then: Clarida and Gali (994), Grilli and Roubini (996), Kim (25), and Scholl and Uhlig (28). Cushman and Zha (997), Kim and Roubini (2), Bjornland (29), Faust and Rogers (23), and Forni and Gambetti (2)

7 Our Identification Follow the same empirical methods: model specification, identification of monetary shocks, and estimation method. Use the same data: also expands the data in terms of sample period and the number of currencies. Our innovation: take US monetary policy regimes to the front of empirical re-investigation of exchange rate behavior in response to US monetary policy shocks. Compare the behavior of exchange rates over three different sample periods: the entire sample period (976-27), Volcker era (979-87), and post-volcker era (988-27).

8 Our Identification Motivation : the Volcker monetary policy regime is...the most widely discussed and visible macroeconomic event of the last 5 years of US history (Goodfriend and King, 25, p. 98): Imperfect credibility of disinflationary monetary policy in early 98s: Historical background: (i) Inflation by then had been a generation-long phenomenon; (ii) a series of adverse real shocks in the 97s accelerated the pace of inflation; (iii) costly short-run trade off between inflation and unemployment. Evidence of imperfect credibility. Motivation 2: Statistical evidence Bernanke and Mihov (998) present evidence for structural breaks in 98 and in 988 or 989 based on Andrews (993). Judd and Rudebusch (998) concerns possible changes in the US Fed s reaction function proxied by the Taylor rule.

9 Empirical Method Consider a reduced form VAR: Y t = B Y t + B 2 Y t + + B p Y t p + u t, E[uu ] = Σ u t = Av t, E[vv ] = I () Specification (Eichenbaum and Evans (995)): VAR Model includes seven variables : Y = (y, p, y, i, nbrx, i, s + p p ) where y and y are industrial production, and nbrx is the ratio of US nonborrowed to total reserves. (2) Estimation (Scholl and Uhlig (28)): Bayesian VAR with lag length 6. (3) Identification of US monetary shocks using sign restriction on impulse responses (Scholl and Uhlig (28)): p k, nbrx k, and i k, for k =,,..., where k denotes months from impact.

10 Data Use the same data: also expands the data in terms of sample period and the number of currencies Our data set covers the 4 trading partners of the US: Austria (AT), Belgium (BE), Canada (CA), Denmark (DK), France (FR), Germany (GE), Italy (IT), Japan (JP), Netherlands (NL), Norway (NO), Spain (ES), Sweden (SE), Switzerland (CH), and United Kingdom (UK). Monthly data from 976: to 27:7. For the member countries of the European Monetary Union (EMU), the sample period ends in 998. But the US-GE exchange rate is replaced with the US-Euro after 998 based on a fixed rate of the GE-Euro. Construct two aggregate exchange rates against the US dollar. One aggregate (AGG98) is made up of all 4 bilateral US dollar exchange rates prior to 999. The other aggregate (AGG) extends the series until 27 while treating the Deutschmark as a representative currency of the former EMU currencies.

11 Four new empirical findings New result : Exchange rates overshoot (Dornbusch Redux) New result 2: A close link between conditional UIP and conditional exchange rate movements New result 4-: Forecast error variance decomposition: Monetary Policy Effectiveness New result 3: A close link between conditional UIP and unconditional UIP (Clue to the causes of the forward premium puzzle). New result 4-2: A close link between conditional and unconditional movements of exchange rates (Monetary Policy Effectiveness)

12 Results : Dornbusch Redux In the entire sample period, the exchange rates gradually fall in response to the contractionary US monetary shock and then return to their long run value (delayed overshooting). In the Volcker era, the maximal response of US real exchange rates is delayed over 8 to 36 months similar to the entire sample (delayed overshooting). In the post-volcker era, the maximal appreciation of US real exchange rates occurs within 2 months followed by gradual depreciation (immediate overshooting). Consistent with the behaviors of real exchange rates for the individual country pairs. Real and nominal exchange rates behave closely each other at both aggregate and individual levels.

13 Results : Dornbusch Redux Entire period Volcker era Post-Volcker era AGG AGG Figure: Impulse responses of real exchange rates to US contractionary monetary shocks

14 Delayed Overshooting Puzzle : an illustration

15 Results : Robustness () sensitivity analyses on the benchmark model in various directions (2) an -variable model with sign restrictions (3) a 7-variable model with recursive identification (4) a structural factor model with recursive identification

16 Results 2: Explanation (Conditional UIP and conditional exchange rate movements) Test for conditional UIP Investigate if there is also a significant difference between the two subsample periods in terms of deviation from UIP. UIP significantly fails in the Volcker era when the overshooting hypothesis also fails; UIP tends to hold in the post-volcker era when the immediate overshooting takes place too. The connection between the prediction of the hypothesis and its key assumptions.

17 Results 2: Explanation (Conditional UIP) Consider the following trading strategy in response to the contractionary US monetary shock: An investor borrows in foreign currency for k months, exchanges it for the US dollar, holds US treasury bills for k months, and exchanges the interest rate returns in the US dollar back for the foreign currency k months later. The return from this strategy between j to j + k is defined by ρ j+k = (s j s j+k ) + i j k i j k, () The cumulated returns from period to period T can be defined by ρ T = T /k l= ρ lk+k. Conditional UIP in our VAR analysis states ρ lk+k = for all l and thus ρ T =, under which gain from the interest rate difference and loss from currency exchange should offset each other.

18 Results 2: Explanation (Conditional UIP) Entire period Volcker era Post-Volcker era AGG Figure: Conditional UIP: cumulated three-month excess returns in response to the monetary shock after adjusting interest rate differences

19 Results 3: Link between Conditional and Unconditional UIP Test for unconditional UIP. Investigate if unconditional UIP tends to hold for the times of immediate overshooting and violates for the times of delayed overshooting. Our tests of unconditional UIP provide clues to sources for violations of UIP, which have long been sought for in the literature of the forward premium anomaly.

20 Results 3: Unconditional UIP Tests if foreign excess returns (s t s t+3 + i t 3 it 3 ) are predictable. Test this prediction using the variance ratio test for the serial dependence of excess returns over long horizon. Define the population variance ratio VR(q) by VR(q) = Var( q i= ξ t+i) qvar(ξ t ) q = + 2 i= ( i ) γ(i), q where ξ t denotes three month foreign excess returns; q an accumulation value ranging from 2 to 2 quarters; γ(i)= autocorr(ξ t+i, ξ t ). Sample split method by Moon and Velasco (23) since data frequency (monthly) is finer than the forecasting interval (3-month) in the same spirit of Hansen and Hodrick (98) and Newey and West (987).

21 Results 3: Unconditional UIP Entire period Volcker era Post-Volcker era AGG Figure: Unconditional UIP: t-values of estimated variance ratios of three-month excess returns

22 Results 4: Monetary Policy Effectiveness (Conditional and Unconditional movements) Link between conditional and unconditional exchange rate movements If unconditional exchange rate movements inherit the conditional behavior, there will be strong positive serial dependence of changes in exchange rates during the Volcker era that reflects the delayed overshooting behavior. Test this prediction using the variance ratio test for the serial dependence of exchange returns over long horizon. Define the population variance ratio VR(q) by VR(q) = Var( q i= ξ t+i) qvar(ξ t ) q = + 2 i= ( i ) γ(i), q where ξ t+i = s t+i s t+i denotes the change in exchange rates between period t + i and t + i; q an accumulation value ranging from 2 to 6 months; γ(i)= autocorr(ξ t+i, ξ t )

23 Delayed Overshooting Puzzle : an illustration

24 Results 4-2: Monetary Policy Effectiveness (Conditional and Unconditional movements) Entire period Volcker era Post-Volcker AGG Figure: t-values of estimated variance ratios of changes in real exchange rates

25 Summary New result : Exchange rates overshoot (Dornbusch Redux) New result 2: A close link between conditional UIP and conditional exchange rate movements (Assumption and Prediction) New result 3: A close link between conditional UIP and unconditional UIP (Clue to the causes of the forward premium puzzle). New result 4: A close link between conditional and unconditional movements of exchange rates (Monetary policy effectiveness). UIP and exchange rate movements are related in data as cohesively as in theory.

26 Our story: What made the Volcker era so extraordinary and overwhelming? Our answer: imperfect credibility of Volcker disinflation policy A common feature underlying delayed overshooting and a violation of UIP is persistent appreciation Theoretical model: Gourinchas and Tornell (24) s expectational error model can explain for both delayed overshooting and violations of UIP. Their key assumption: agents underestimate the effective horizon of a monetary shock. Evidence on imperfect credibility of Volcker disinflation policy: The Volcker Fed s u-turn in response to the first appearance of a recession in 98s The long-term Treasury yields remained above % until October 985. The difference between the long-term interest rate and inflation during the period is highest among the entire postwar period. The FOMC transcripts

27 What made the Volcker era so extraordinary and overwhelming? US-AGG nominal exchange rates US long term interest rates and inflation rates Figure: Nominal exchange rates, long term interest rates, and inflation rates

28 Our story: What made the Volcker era so extraordinary and overwhelming? [W]e ve had a heck of a difference in the interest rate structure since 79, in those terms, which one can interpret as the American people giving up on the idea that we were ever going to return to price stability, Volcker remarked (FOMC transcript, August 98; p.39). Our story about imperfect credibility of monetary policy in the Volcker era and credibility of monetary policy in the post-volcker era can explain why we only observe both delayed overshooting and violations of UIP in the Volcker era and why we do not in the other sample periods.

29 Conclusion and Related Studies Monetary Policy Regime matters. Delayed Overshooting Puzzle, Forward Premium Puzzle, and Purchasing Power Parity Puzzle. International transmissions of U.S. monetary shocks [Mundell-Fleming-Dornbusch model and Obstfeld and Rogoff (995)]: expenditure switching effects and demand effects.

30 International transmission of US monetary shocks Entire period Volcker era Post-Volcker era y Figure: Impulse responses of foreign output to the US contractionary monetary shock

31 A 7-variable model with recursive identification (Eichenbaum and Evans) Entire period Volcker era Post-Volcker era f nber Figure: Impulse responses of foreign output to the US contractionary monetary shock

32 Financial Crisis Period We do not include the global financial crisis in our sample. Reason (Structural): The Fed has conducted an extraordinary quantitative easing policy since the financial crisis. That is, a possible change in the US monetary policy regime. Reason 2 (Technical): Nonborrowed reserves turned to negative during 28:-28:. This raises a technical difficulty in empirical examination.

33 Pre-Volcker era( ) The pre-volcker era mixes up the observations from the transitional early years ( ) with thosee from a more complete floating regime (976.2 onward) Hansen and Hodrick (983) suggest to choose a starting date of February, 976. During the period of ( ), agents may have been expecting a return to a fixed rate regime and not have known or fully understood the intervention strategies of the central banks in the beginning of the era.

34 Pre-Volcker era( ) The IMF (the Committee of 2) put forward an Outline of Reform in June 974, essentially suggesting that the restructure [of the international monetary system] be based on stable but adjustable par values with limited floating. However, this solution failed to achieve a concensus among G6 countries The Rambouillet Agreement led the floating exchange rate system to be formally ratified in January 976 following the approval from the Interim Committee of the IMF. A signal for a complete transition from the Bretton Woods system to a modern flexible exchange rate regime from the viewpoint of market participants

35 Results of Pre-Volcker era AGG AGG98 AGGX Figure: Impulse responses of foreign output to the US contractionary monetary shock

36 Figure: Impulse responses of real exchange rates to US contractionary monetary shocks Bilateral RER (Pre-Volcker era) CA CH DK GE JP NO SE UK

37 Results 4-: Monetary Policy Effectiveness (Forecast error variance decomposition) The share of the forecast error variance of the aggregate exchange rate attributable to the US monetary shock follows the same pattern as its impulse response, varies significantly across sample periods, and appears to be far smaller than that at the individual country level in each of the three subsample periods. The maximal accountability of the US monetary policy shock occurs over 2 to 36 months at the median estimate in the Volcker era (2.4%), around to 4 months in the pre-volcker and post-volcker eras (9.9 and 22.4%), and throughout the 5-year horizon in the entire sample period (5.3%).

38 Results 4-: Monetary Policy Effectiveness (Forecast error variance decomposition) Entire period Volcker era Post-Volcker AGG AGGX Figure: Forecast error variance decomposition of real exchange rates

39 Results 4-: Monetary Policy Effectiveness (Forecast error variance decomposition) Aggregation effect: Because of the extensive dispersion in the timing of their maximal accountability, individual exchange rate fluctuations attributable to the common US monetary shock cancel one another out in the aggregate. Pooling effect: Because of the extensive dispersion in the timing of the maximal accountability across the subsample periods, the exchange rate fluctuations cancel one another out in the pooled sample. Can explain previous studies: For example, using monthly data from 974: to 99:5 during which the delayed overshooting appears to be most prominent, Eichenbaum and Evans (995) report that US monetary policy shocks account for approximately 42, 26, and 23% of the exchange rate fluctuations for the US-GE, US-UK, and US-JP country pairs, respectively, at lags 3-36 (see also Clarida and Gali (994)). using monthly data from 975:7 to 22:7, a coverage period that nearly overlaps with our entire sample period, Scholl and Uhlig (28) report a share of approximately % for the same countries at the same lags.

40 Results 4-: Forecast error variance decomposition (Volcker era) CA CH DK GE JP NO SE UK Figure: Forecast error variance decomposition of real exchange rates

41 Empirical Method: imposing sign restrictions based on a Bayesian approach Consider a VAR: Y t = B Y t + B 2 Y t + + B p Y t p + u t, E[uu ] = Σ u t = Av t, E[vv ] = I Step : Form a prior for the reduced-form VAR. Using the data, form the posterior. Step 2: Take a draw (B, Σ) from the posterior. Calculate the Cholesky decomposition Σ = ÃÃ Step 3: Take a draw α from the unit sphere (dimensions=number of variables). Calculate a = Ãαwhere a is a column vector of A and Σ = AA. Step 4: Calculate the impulse responses to a. If they satisfy the sign restrictions, keep the joint draw (B, Σ, a), otherwise discard. Given the draws kept, calculate statistics of interest.