Katarina Juselius Department of Economics University of Copenhagen

1. Katarina Juselius Department of Economics University of Copenhagen

2. Economic theory: Shock structure, equilibrium relations, and expectations formation • Model-based rational expectations: a tool to ensure theoretical consistency • Principles for taking such models to the data • Empirical consistency entails carefully matching the basic assumptions underlying the theoretical model with the empirical regularities of the data as structured by a statistically adequate model

3. An empirically oriented methodology • There are many economic models but one economic reality. Hence one should be able to use the economic reality as structured by the available data to create confidence intervals within which potentially empirically relevant models should fall. • Economic data typically exhibit both pronounced persistence and structural breaks. These are informationally rich features of the data that can be exploited in particular when choosing between competing explanatory theories. Economists often try to rid the data of these features from the outset (differencing the data, Bayesian priors, calibrating parameters, ignoring breaks, etc.) and by doing so use empirical evidence to illustrate their beliefs rather than asking sharp and novel questions. • The Cointegrated VAR model can handle both unit root persistence and breaks. In its unrestricted form it is just a convenient reformulation of the covariances of the data.

4. Using persistence as a structuring device

7. A proposal for classification

8. Structuring persistence with the CVAR The I(1) model:

9. The I(2) model:

10. Bridging theory and evidence: A CVAR scenario How to treat unobservable expectations?

11. Principles for formulating a scenario • Translate the postulated behavioral equations of a theoretical model into a set of conditions on the persistency properties of the steady-state relations. For example, REH-based theoretical models mostly assume that both the purchasing power parity and the uncovered interest rate parity hold as stationary (or at most as a near I(1)) conditions, whereas IKE-based models assume that the real exchange rate and the interest rate differential are near I(2) processes and cointegrate to I(1), and by adding the inflation spread that they cointegrated to I(0). • Express the expectations variable(s) as a function of observed variables. For example, Uncovered Interest Rate Parity (UIP) assumes that relative interest rates are equal to the expected change in the nominal exchange rate. Thus, provided the parity holds, the observed interest rate spread is a measure of the expected change in nominal exchange rate and its persistency property can, therefore, under Assumptions A and B be studied empirically.

12. 3. For a given order of integration of the variable(s) determined outside the model, derive the order of integration of all remaining variables. 4. Translate the stochastically formulated theoretical model into a theory-consistent CVAR scenario by formulating the basic assumptions underlying of the theoretical model as a set of testable hypotheses on cointegration and common trends properties. 5. Estimate a well-specified VAR model and check the empirical adequacy of the theory-consistent CVAR scenario.

13. Defining shocks

14. Doornbush/Frankel type of overshooting models • Basic Equilibrium Assumptions: • PPP holds as a stationary condition • UIP holds as a stationary condition • Domestic Fisher parity holds as a stationary condition • International Fisher parity holds as a stationary condition

15. Anchoring expectations to observables

16. Persistency properties of exchange rates and prices

17. REH prices are I(2)

18. International Fisher parity under REH

19. Real exchange rate under REH

20. Time series implications of REH:

21. A CVAR scenario for Dornbush/Frankel model

22. Cointegration implications:

23. An IKE based model for nominal exchange rate Basic implications:

24. Empirical illustration

26. Difference between IKE and REH REH: Real exchange rate at most I(1) IKE: real exchange rate near I(2)

28. The Frydman-Goldberg monetary model • Equilibrium conditions: • Uncertainty Adjusted UIP (UA-UIP) • PPP with international Fisher Parity The uncertainty premium is measured by the gap effect (deviation from PPP)

29. Equilibrium in the goods market

30. Anchoring expectations to observables

31. Interest rates are near I(2) under IKE

32. Difference between REH and IKE Under IKE, the best predictor of nominal interest rate: Under REH, it is:

33. The UA-UIP condition Or: Replacing up with the gap effect:

34. The Fisher parity condition under IKE Inserting expression for interest rate: Integrating:

35. To summarize: IKE is consistent with

36. IKE based scenario Two polynomially cointegrating relations: One medium-run relation:

37. Rank determination

38. Tests of time-series properties I(2) versus I(1)

40. The pushing forces

41. The pulling forces

43. Summary and conclusion • Under IKE, speculative behavior in the currency market is likely to drive prices away from long-run PPP benchmark values for extended periods of time. Such persistent movements away from equilibrium PPP values are likely to have the property of a near I(2) process, i.e. real exchange rates are likely to be near I(2). REH-based models assume that movements away from long-run PPP values are stationary, or at most near I(1). • Such persistent swings in real exchange rates have to be offset by something else. The IKE theory tells us that it should be the real interest rate differential. Hence, real interest rate differentials should exhibit a similar persistence as real exchange rates, i.e. be near I(2). The REH-based theory assume that the real interest rate differential should be stationary or at most near I(1).2.

44. Conclusions cont. • According to IKE, real exchange rates and real interest rate differentials should cointegrate to a stationary relation, whereas according to REH, they should be individually stationary albeit allowed to exhibit some persistence. • According to IKE, prices need not be rigid to produce the long swings in real exchange rates, but the speed of adjustment has to differ between relative prices and nominal exchange rate, i.e. to explain the long swings a de-linking of prices and nominal exchange rate is needed. REH-based overshooting models with price rigidities do not assume de-linking whereas endogenous money versions do.

45. Discussion • What is initiating a long swings cycle? • Why is this currency speculation profitable? • What is the likely impact of speculation on the real economy? • What causes a reversal of the long swings?