Microeconomic Sources of Real Exchange Rae Variabiliy By Mario J. Crucini and Chris Telmer Discussed by Moren O. Ravn
THE PAPER Crucini and Telmer find ha (a) The cross-secional variance of LOP level violaions is large relaive o he ime-series variaion (b) The variaion of changes in LOP violaions is dominaed by idiosyncraic shocks raher han common locaion-pair specific variaion (c) Nominal exchange raes correlae closely wih he counry-specific variaion They conclude ha nominal exchange raes play only a minor role for relaive prices here s a lo more acion wihin he disribuion of relaive prices han in is mean.
THE DATA Daa on Price levels Of individual goods (up o 300) In muliple counries / ciies (123 locaions) Annual, ime-series dimension (1990-2005) This is a very ineresing daase. Allows one o invesigae he deerminans of absolue price level differences and convergence o he absolue version of he LOP.
THE PAPER Crucini has earlier worked on his daa in: Crucini, Telmer and Zachariadis, AER 2005: Roughly as many underpriced as overpriced goods beween any wo EU counries LOP violaions greaer for non-raded goods han for raded goods and for goods wih large inpus of nonraded inpus Crucini and Shinani, 2006: Half-life of LOP violaions around 1.2 year for raded goods and 1.9 years for non-raded goods
The Resuls Crucini and Telmer define LOP violaion as: q i, jk, = log P ij, P S ik, jk, The logarihm of he price of good i in locaion j a dae convered ino currency k divided by he price of good i in locaion k a dae And hey decompose he oal variance as: ( q i) = var E [ q i, jk] i, jk, jk ( ) ( ( ) + E var q i, jk var, i, jk, T i + jk i, jk F i
The Variance Decomposiion
The Variance of Changes in LOP Violaion Crucini and Telmer hen look a he variance of changes in LOP violaions: Δq = f + ε f i, jk, jk, i, jk,, ε jk, i, jk, The compue he common componen as: f ( q jk ) jk, = Ei Δ i, jk,, And hey examine he common componen s conribuion o he overall variance of he changes in LOP violaions
The Common Componen The common componen is small bu srongly correlaed wih NER
Commens 1. Trade vs. Macro. Crucini and Telmer find ha here is los of cross-secional variaion in deviaions from he absolue version of he LOP. - They claim ha his implies ha rade heory is more imporan han macro for undersanding relaive prices - I s emping o agree bu does his conclusion really follow from he resuls? - Could heir resuls simply be due o sampling uncerainy?
frequency Sampling uncerainy will make he economerician conclude in favor of rade over macro bu macro is he only hing ha maers US disribuion US year 1 disribuion year 0 UK disribuion UK disribuion year year 0 1 Dallas, shop A New York shop A New York, shop B London shop B London shop A New York, New shop York, B shop A London shop London A shop B price
Commens 2. The imporance of exchange raes / sicky prices: Is he following rue? Fac: Los of idiosyncraic changes in LOP violaions (a) Implies ha If you hink ha annual real exchange rae variabiliy is indicaive of nominal exchange raes moving around a disribuion of microeconomic sicky prices, you are wrong. (b) Which implies ha Nominal exchange raes maer lile? I m no sure abou neiher (a) nor (b).
Counerexample Take a exbook model of sicky prices (Calvo saggering, for example). his model wih display price dispersion in equilibrium suppose ha he only reason by prices differ across locaions is ha prices are sicky would his model necessarily say ha he common componen (he exchange rae) should dominae he cross-secional changes in relaive prices?
The exchange rae shock has shifed he disribuion bu here are sill large idiosyncraic movemens due o pas shocks. 0 q i
Commens So, how could we check he exen o which exchange raes maer? I hink he righ quesion is: If he nominal exchange rae changes by x%, he real exchange rae changes by y% over a horizon of z periods in oher words a sor of ERPT regression. The resuls of Crucini and Telmer do no answer his quesion he fac ha here are many oher variaions in relaive prices doesn mean ha nominal exchange raes are no imporan. In fac, he covariance beween changes in LOP violaions and NERs migh be reasonably high perhaps he evidence does no so srongly go agains sicky prices?
Things I would have liked o see The daa are very ineresing price level daa for a large no. of counries and a longish sample. I would have liked o know: which economic forces affec he fixed effec (he long run difference in prices)? which economic forces affec he adjusmen in prices? in response o shocks, do prices end o rever owards he LOP? do price differenials end o ge smaller over ime?
Oher commens 1) Wha happens in large devaluaions? Does he whole disribuion of prices change? 2) The common componen he auhors simply average prices cross-secionally o ge he common componen is his appropriae? 3) Can he daa be made available, please?
Summary The work of Crucini on he opic of he LOP is exremely imporan I look forward o more! - I have ried o be he devil s advocae, bu a he end of he say, I hink ha heir resuls are prey convincing.