Exchange Rate Bands of Inaction and Play Hysteresis in German Exports Sectoral Evidence for Some OECD Destinations

Size: px

Start display at page:

Download "Exchange Rate Bands of Inaction and Play Hysteresis in German Exports Sectoral Evidence for Some OECD Destinations"

Willis Richardson
8 years ago
Views:

1 Exchange Rate Bands of Inaction and Play Hysteresis in German Exports Sectoral Evidence for Some OECD Destinations by Ansgar Belke, Matthias Goecke and Martin Guenther Vorlesung Empirie der internationalen Geld- und Finanzmärkte, SS

by Ansgar Belke, Matthias Goecke and Martin Guenther

2 Content 1. Introduction 2. Episodes of $/ EXR pain thresholds and stylized facts 3. Hysteresis in exports: the band of inaction 4. Modelling play-hysteresis: empirical results 5. Conclusion 2

3 1. Introduction European politicians and business persons express concerns nearly every time when the euro appreciates. Idea that a stronger reaction of exports beyond some boundaries of the exchange rates occurs is widely accepted. But at what specific level does the external value of the euro hit its pain threshold? Or is the interrelation of exchange rate and exports linear and neither pain threshold nor band of inaction exists? We intend to answer the question concerning the existence of hysteretic effects in foreign trade and assess their empirical relevance for German exports. 3

But at what specific level does the external value of the euro hit its pain threshold?

4 1. Introduction Hysteretic impact of real exchange rates on German exports from 1995Q1 to 2010Q3. Destination- and sector-specific pain thresholds? Non-linear model: suddenly strong spurts of exports occur when changes of the exchange rate go beyond a zone of inaction the play area analogous to a mechanical play. We develop an algorithm describing play-hysteresis and implement it into a novel regression framework. Looking at some of the main export destinations of Germany outside the euro area and some of the most important tradable sectors we find significant hysteretic effects on German exports. 4

mechanical play. We develop an algorithm describing play-hysteresis and implement it into a novel regression framework.

5 1. Introduction Keywords: exchange rate movements; play hysteresis; modelling techniques; switching regression; export demand Seminal literature: Amable, Henry, Lordon and Topol (1991), Cross (1994) and Krasnosel'skii and Pokrovskii (1989). Aggregation of non-linear path-dependent microeconomic activity to a sectoral or macroeconomic analysis not straightforward. Reason: path-dependent dynamic pattern may differ between micro perspective of a firm and aggregated macro perspective of an entire sector/economy consisting of heterogeneous firms 5

Aggregation of non-linear path-dependent microeconomic activity to a sectoral or macroeconomic analysis not straightforward.

6 1. Introduction Compared to existing studies of hysteresis in foreign trade, our approach is closer to the original concept of a macroeconomic hysteresis loop. Why? (i) It is not based on the discontinuous non-ideal relay interpretation as in the microeconomic firm level case (ii) the path-dependent structural breaks in the macroeconomic relations are not added to the system as an exogenous information. On the contrary, in our approach the structural shifts are explicitly determined by the history of the exchange rate and the exports are simultaneously estimated together with the (pathdependent) relation of exports to the exchange rate. 6

the macroeconomic relations are not added to the system as an exogenous information.

7 2. Episodes of $/ exchange rate and stylized facts 24,000 20,000 16, ,000 8,000 4, US-$/ -Exchange Rate German Exports to USA in Mil. Source: Own presentation based on Eurostat and Bundesbank data. 7

8 90 92 94 96 98 00 02 04 06 08 10 US-$/ -Exchange Rate German

8 German exports tend to move much slower than the $/ exchange rate. If exchange rate changes are one-directional and steady also exports react much stronger. Weak reaction to small and medium-sized exchange rate movements is caused by hedging of exchange rate uncertainty, relatively low price elasticity of German exports, sunk costs related to foreign trade activities, and relief on the export side by cheaper imports as well as the specific pricesetting of German exporting firms. 8

Weak reaction to small and medium-sized exchange rate movements is caused by hedging of exchange rate uncertainty,

9 3. Hysteresis in exports: the band of inaction Discontinuous MICRO hysteresis loop: export activity of a single firm h state of activity of an exporting firm active exit band of inaction inactive x out (exit rate) x c variable unit costs entry x in (entry rate) x t exchange rate sunk exit costs + option value of delaying exit sunk entry costs + option value of delaying entry 9

inaction inactive x out (exit rate) x c variable unit costs entry x in (entry rate) x t

10 3. Hysteresis in exports: the band of inaction On a microeconomic level hysteresis occurs via a band of inaction, i.e. differences between both trigger/thresholds. Band-of-inaction will be the wider, the lower the demand elasticity is, the higher the absolute values of the sunk entry and exit costs are and the higher the uncertainty about the future situation of the exporter is. Belke and Goecke (2005) focus on the shape and the location of a macroeconomic hysteresis loop, i.e. on the problem of aggregation. Aggregation is not trivial if heterogeneity regarding the value of sunk exit/entry costs and/or the level of uncertainty about future market situation and/or the elasticity of demand is taken into account, i.e. if the entry and exit trigger exchange rates are different for different exporting firms. 10

of the exporter is. Belke and Goecke (2005) focus on the shape and the location of a macroeconomic hysteresis loop, i.e. on the problem of aggregation.

11 3. Hysteresis in exports: the band of inaction In this (realistic) case of heterogeneity the transition from the micro to the macro level leads to a change of the hysteresis characteristics: the aggregate hysteresis loop shows no discontinuities (as known from ferromagnetics). However, a pattern not very different to a band of inaction remains. 11

12 3. Hysteresis in exports: the band of inaction Belke and Goecke (2005) show that even the macro behaviour can be characterized by areas of weak reactions which are corresponding to mechanical play called play -areas. For play hysteresis, see Krasnosel'skii and Pokrovskii (1989). Persistent aggregate (export) effects do not result from small changes in the forcing (exchange rate) variables, as far as the changes occur inside a play area. However, if changes go beyond the play area, sudden strong reactions (and persistence effects) of the output variable (i.e. exports) occur. 12

Persistent aggregate (export) effects do not result from small changes in the forcing (exchange rate) variables, as far as the changes occur

13 3. Hysteresis in exports: the band of inaction The specific realization of the exchange rate which materializes instantly after the complete passing of the play area can be denoted as a pain threshold, since, having passed this realisation of the exchange rate, the reaction of exports to changes in the exchange rate becomes much stronger. 13

14 3. Hysteresis in exports: the band of inaction Play-hysteresis is in two aspects different to the micro-loop. First, the play-loop shows no discontinuities. Second, analogous to the play in mechanics (e.g. when steering a car) the play area is shifted with the history of the forcing variable (exchange rate): Every change in the movement of the forcing variable starts with traversing a play area. Only after the play is passed, a spurt reaction will result, if the forcing variable continues move in the same direction. In the following, a straightforward empirical framework to test for a play-type impact of the exchange rate on exports is presented. We adapt an algorithm developed in Belke and Goecke (2001) to describe play-hysteresis and implement it into a regression framework. 14

..... Every change in the movement of the forcing variable starts with traversing a play area.

15 y Linear play-hysteresis and spurt areas aggregate exports spurt line spurt line (downwards) (upwards) D C a B E F G x 5 x 4 x 3 x 2 A x 0 x 1 exchange rate x play p 15

16 The hypotheses to be tested refer to the generalised equation: yt = C + xt + st( ) + zt with: < + pt = with: 0 xt = hysteretic regressor x st = spurt variable based on hysteretic regressor X zt = non-hysteretic regressors. Dependent variable yt is determined by past spurts and current reaction. Spurt variable st summarizes all preceding and present spurt movements leading to a shift of the current relation between x and y ( filtered x ). If we want to check whether play is relevant, we have to test the hypothesis (H1) 0 against the alternative = 0. 16

Dependent variable yt is determined by past spurts and current reaction.

17 4. Empirical results Estimation output from a standard regression without play / spurt (restriction = 0) SICT 78 (Road vehicles including air-cushion vehicles) 17

18 4. Empirical results Estimation output resulting from a one-dimensional grid search with constant play over ( p = = 0.23) 18

19 4. Empirical results Real exchange rate and the resulting spurt variable ( =0.23) 19

20 Export Destination Regression results with constant play the main pattern Denmark Japan Norway Switzerland United States SICT Group 0 to without 78 8 *** *** *** *** *** β> 0 *** play= 1.55 play= 1.5 play= 1.2 play= 1.14 play= 1.14 t= t= t= t= t= t= *** β> 0 ** *** *** *** *** play= 13.5 play= 44 play= 22 play=44 play= 50 t= t= t= t= t= t= *** *** *** *** *** play= 3.2 play= 3.2 play= 2.7 play= 2,95 play=3.25 play=2.9 t= t= t= t= t= t= -2,9606 *** * *** * ** play= 0.34 play= 0.41 play= 0.27 play= 0.09 play= play= 0.43 t= t= t= t= t= t= β> 0 ** β> 0 * β> 0 *** *** q.d. ** play= 0.23 play= 0.3 t= t= t= t= t= t= Notes: β>0 - coefficient of the SPURT variable is positive which is not in line with hysteresis theory; q.d. - SPURT variable turned out to be a dummy for some structural breaks;*** 1 %, ** 5 %, * 10% probability level referring to the usual student t-statistic of the SPURT variable; t refers to the empirical realisation of the t-statistic of the SPURT-variable. 20

2673 t= -5.7406 t= -2.9419 t= -9.2607 t= -9.4409 *** *** *** *** *** play= 3.2 play= 3.2 play= 2.7 play= 2,95 play=3.25 play=2.9 t= -3.4465 t= -10.5618 t= -3.8175 t= -3.6859 t= -1.

21 5. Conclusion Our main aim has been to identify a band of inaction for German exports. Analyzing some of the largest export partners outside the Euro-zone to which 14.8% of the total German exports were directed, we find hysteretic effects in more than 8% of total German exports. Not every increase or decrease of the exchange rate will, automatically, lead to positive or negative reactions of the volume of exports. However, a large appreciation of the euro means passing the play-area (i.e. a kind of 'pain-threshold') and results in a strong reaction of exports. Et vice versa! Policy conclusions (FOREX market interventions and other monetary policy reactions to strong euro) 21

RUHR. Exchange Rate Bands of Inaction and Play-Hysteresis in German Exports ECONOMIC PAPERS. Sectoral Evidence for some OECD Destinations #327

RUHR. Exchange Rate Bands of Inaction and Play-Hysteresis in German Exports ECONOMIC PAPERS. Sectoral Evidence for some OECD Destinations #327 RUHR ECONOMIC PAPERS Ansgar Belke Matthias Göcke Martin Guenther Exchange Rate Bands of Inaction and Play-Hysteresis in German Exports Sectoral Evidence for some OECD Destinations #327 Imprint Ruhr Economic