Wha is Driving Exchange Raes? New Evidence from a Panel of U.S. Dollar Bilaeral Exchange Raes Jean-Philippe Cayen Rene Lalonde Don Colei Philipp Maier Bank of Canada The views expressed are he auhors and need no reflec hose of he Bank of Canada.
Roadmap. Moivaion 2. Relaed Lieraure 3. Empirical mehodology Dynamic facor model Coinegraion ess Sae space model Hisorical decomposiion Robusness checks 4. Conclusion
08: US$/CDN$ since 970 (quarerly).0 976Q3.00 986Q 99Q3 2002Q 0.90 0.80 0.70 0.60.0.00 0.90 0.80 0.70 0.60 06: 70: 72: 74: 76: 78: 80: 82: 84: 86: 88: 90: 92: 94: 96: 98: 00: 02: 04:
Why did he Canadian dollar appreciae so rapidly? Real oil prices and non-energy commodiy prices increased by 50% and 70%, respecively, beween 2002 and 2008. Large appreciaions of commodiy exporers (Ausralia, Canada, New Zealand) agains he U.S. dollar likely linked o commodiy prices; If his is rue, why did commodiy imporers (euro area and U.K.) appreciae, oo, agains he U.S. dollar?
Changes in exchange raes 2000-2008 200 75 50 25 00 Jan- 02 Jul- 02 Jan- 03 Jul- 03 Jan- 04 Jul- 04 Jan- 05 Jul- 05 Jan- 06 Jul- 06 Jan- 07 Jul- 07 Jan- 08 Jul- 08 Jan- 09 CAD EURO YEN NZD AUD GBP
2 Relaed lieraure Models linking exchange raes o macro fundamenals Models linking exchange raes o commodiies Laen facor models (empirical finance) Overall limied success Empirical resuls somewha beer Good resuls, bu economic inerpreaion no always clear Meese and Rogoff 983, Obsfeld and Rogoff 2000, Cheung e al. 2005, Engel and Wes 2006 Amano and van Norden 995, Gruen and Korian 996, Djoudad e al. 200, Chen and Rogoff 2003, Issa e al. 2006 Diebold and Nerlove 989, Mahieu and Schoman 994, Dungey 999
3 Empirical mehodology Exrac common movemens in he panel of exchange raes using: Dynamic principal facor analysis Sae-space model Each has relaive advanages and disadvanages.
Models The principal facor model: Purely saisical echnique, relies on a minimum of resricions and assumpions The sae-space model Imposes more srucure on he daa Faciliaes saisical inference Allows for explici links beween common/specific componens and an economic variable Can be sensiive o he assumpions regarding he processes of he unobservable variables
Daa We use real U.S. bilaeral exchange raes for Ausralia, Canada, he euro area, Japan, U.K. and New Zealand. Sample: 98Q-2007Q4 (currencies floaed freely agains he U.S. dollar). Mos exchange raes I(); small firs-order auoregressive erm (0.3), so we use he daa in firs differences
3. Dynamic principal facor model We sar by esimaing a dynamic facor model. We normalize each exchange rae by is sandard deviaion o ensure equal conribuion o he oal variance of he model. Panel of p real U.S. dollar exchange raes (X=(x,... x p ) ) wih mean µ can be expressed as linear funcions of m (m<p) hypoheical common facors (f,...f m ), plus an error erm. q q p p f l f l f l f LF X...... 2 2 2 2
Resuls from he dynamic facor model We find wo facors (eigenvalue of he hird facor 0.3)
Loading facors dynamic facor model Facor Facor 2 AU 0.6 0.45 CA 0.43 0.37 EU 0.79-0.34 JA 0.52-0.25 NZ 0.74 0.2 UK 0.74-0.26 Resuls robus o esimaion of a saic or dynamic facor model
3.2 Wha drives he facors? Facor models useful o deec paerns in he daa, bu more evidence needed for economic inerpreaion. Sraegy:. Coinegraion ess Relaive fiscal posiion Shor-erm ineres raes Commodiy prices No evidence for produciviy differenials 2. Building a sae space model ha incorporaes he facors direcly.
Wha drives he facors? Insighs from he lieraure. Produciviy shocks (Balassa/Samuelson effec) 2. Fiscal policy: Rising governmen deb leads o fall in savings and lowers ne foreign asses-o-gdp raio; leading o a depreciaion o faciliae build-up of curren accoun surpluses (Blanchard, 985; Weil, 989). 3. Commodiy prices, which proxy exogenous erms-ofrade shocks (Chen and Rogoff, 2003).
Wha drives he firs facor ( U.S. facor )? We esimae dynamic facor models for governmen deb-o-gdp raios (relaive o US) and produciviy differenials. In boh cases, he firs facor moves in he same direcion for all counry pairs. Firs facor of our exchange rae panel is coinegraed wih he firs facor of governmen deb, bu no wih produciviy differenials. A deerioraion in he U.S. fiscal posiion - relaive o all oher counries - leads o a long-run, mulilaeral U.S. dollar depreciaion. This is in line wih overlapping generaions models (Ganelli, 2005; and Kumhoff and Laxon, 2007).
Firs facor: relaive U.S. fiscal posiion
Wha drives he second facor? Paern of he loading facor suggess ha commodiy prices migh play a role. Indeed, second facor coinegraed wih real energy and non-energy commodiy prices.
Second facor: Commodiy prices
Second facor: Commodiy prices
Coinegraion ess Firs facor Second facor Tes saisic (Saikonnen) -4. -4.80 Tes saisic (Engle-Granger) -4.07-4.22 Criical values % -4.44-4.44 5 % -3.83-3.38 0% -3.52-3.52
Loading facor 2 vs. ne commodiy impors, relaive o he US
3.3 Sae space model We build a sae-space model which incorporaes: A direc link beween he firs facor and he firs facor of he relaive deb-o-gdp raio A direc link beween he second facor and energy- and non-energy commodiy prices The facor in he shor-erm ineres rae differenial drives he shor-erm dynamics of he exchange rae
p oil : real WTI oil price, p NE : real IMF non-energy commodiy prices Deb : symmerical common componen of he deb/gdp raio, relaive o he U.S. deb/gdp rdiff : Symmerical common componen of he shor-run real ineres rae differenial wih he U.S. D93: dummy for energy liberalizaion in Canada Sae-space model i for all ) ( Deb rdiff 2 4 3 2 2 2 2 2 2 2 93 2 i i Oil NE i CAN i i i v L D p p C C C C C C v C D C X
Esimaion of he srucural sae space US Facor Commodiy facor
Esimaion of he srucural sae space: Coinegraion relaionships
U.S. facor: Facor model vs. sae-space model
Commodiy facor: Facor model vs. saespace model
3.4 Hisorical decomposiion: Euro area Frazscher (2007): Europe has borne he bulk of he adjusmen of he U.S. dollar effecive exchange rae over he pas 25 years.
Hisorical decomposiion: Canada
Hisorical decomposiion: Japan
Hisorical decomposiion: UK
Hisorical decomposiion: Ausralia
Hisorical decomposiion: New Zealand
3.5 Robusness checks Key resuls are robus o; Esimaion sample (pos-breon Woods or 980-2007) Mehodology o exrac he common componens (principal facor model or sae-space model) Changing he nummeraire currency (USD or euro) Modelling he link beween he facor and economic variables explicily, or esimaing hese links ex-pos (like in he facor model).
4. Conclusion We sudy a panel of six bilaeral U.S. dollar real exchange raes using a principal facor and a sae-space model. Key insighs:. U.S. shocks are imporan role in explaining exchange raes over he 2002-2007 period. 2. U.S. fiscal shocks seem o have played an imporan role in he recen (2002-2007) generalized depreciaion of he U.S. dollar. 3. Commodiy prices help explain he behaviour of commodiy currencies, including Canada