5. Data and Results5.2 Pooled Impulse Response Function Results5.3 Pass-through Ratios5.4 Variance DecompositionsCountry%change in Industrial Production%change in Real Exchange Rate%change in CPI%change in Nominal Exchange Rates%change in Import PricesTable 3. Variance Decomposition of %change in Nominal ExchaCountryA New Look at Pass-through Jay Shambaugh∗Dartmouth College Abstract: This paper examines the relationship between exchange rates and prices. Rather than assume exchange rate changes are exogenous shocks that affect prices, I use a long-run restrictions VAR to identify shocks and explore the way domestic prices, import prices and exchange rates react to a variety of shocks. Consumer price pass-through is nearly complete in response to some shocks, but low in response to others. Alternatively, import prices and exchange rates typically respond in the same direction, and pass-through seems quick. This supports the idea that import prices are set in the producer’s currency and that lower CPI pass-through is a result of changes in quantities or margins further down the supply chain. JEL Classifications: F3, F4, E31 Keywords: Exchange Rates, Pass-through, Import Prices, Inflation, Vector Autoregression ∗ Jay C. Shambaugh, HB 6106, Economics Department, Dartmouth College, Hanover NH 03766. phone: 603-646-9345. Fax: 603-646-2122 email: [email protected] I: Introduction The way the exchange rate interacts with the real economy depends critically on exchange rate pass-through, the degree to which exchange rate changes are passed through to price level changes. Our typical view of pass-through (notional pass-through) is one where the exchange rate changes for an exogenous reason and we see how prices change over time. Given this view, pass-through plays a major role in determining the extent of exchange rate volatility, the response of the terms of trade and trade balance to exchange rate changes, and the optimal monetary policy and exchange rate regime. When measured, though, pass-through seems to vary both across countries and time, and understanding the reasons for the different reactions of prices to exchange rate changes requires an examination of the root causes of exchange rate and price changes. The wide variation in previously found pass-through coefficients may in fact be nothing more than measuring the fact that different shocks were hitting the exchange rate at different times. I seek to remedy some of the problems of interpreting pass-through as measured by asserting that macro-level pass-through as typically measured (measured pass-through) can be misleading because it assumes that the change in the exchange rate is the shock itself. In this paper, rather than focus on the correlation of exchange rates and prices, I will identify fundamental shocks to the economy and test their effects on exchange rates, consumer prices, and import prices. Depending on the way in which exporters and others in the supply chain set their prices, we could expect the change in the exchange rate to be fully reflected in domestic import prices (full pass-through) or could expect foreign exporters to compress their profit margin to avoid drastic changes in the prices their customers face. The pass-through to the overall price level will depend on the way in which domestic and import prices interact and the reaction of the overall economy to the increase in import prices. It is likely that the way price setters react will depend on why the exchange rate is changing. For example, during a hyperinflation, money grows, prices rise, and the exchange rate depreciates, generating a picture of full pass-through. At the same time, changes in foreign money supply levels may change the exchange rate and foreign price levels at the same time generating little or no change in import or domestic prices. In neither case is the law of one price nor purchasing power parity violated, yet these examples give very different pass-through coefficients when pass-through is simply measured as the change in domestic prices with respect to changes in the exchange rate. Alternatively, one might see changes in the exchange rate generated by real factors: changes in the world economy that require a change in the home currency to maintain internal and external balance. This may lead to exchange rate changes without any changes in domestic prices, while import prices may change. In this case, it seems that pass-through into the overall price level truly was incomplete. Previous macro-level studies have measured the effects of changes in the exchange rate on domestic prices. Many, though, failed to adjust for the fact that foreign costs and the exchange rate may be changing at the same time. To combat these issues, various researchers have tried to control for changes in costs as well as 1other pressures on domestic prices. Micro-level studies of a particular product or industry have had more success in creating sufficient controls, but then limit the ability for international comparisons due to lack of data availability. Even when econometrically valid pass-through coefficients are found, studies of different countries over different time periods often generate widely varying pass-through coefficients leaving us with no clear picture of the importance of exchange rate changes for domestic economic conditions.1 The difficult methodological question is the appropriate way to identify the shocks an economy faces. While many of the examples above were monetary changes, we cannot simply control for changes to the money supply because money and monetary policy respond endogenously to many other variables. A large body of literature has developed surrounding the issue of the effects of monetary shocks or monetary policy shocks (Christiano, Eichenbaum and Evans (1999)). Unfortunately, as the literature shows, the assumptions regarding the short run behavior of money, prices, and other variables, which are needed to statistically identify monetary shocks, have a substantial impact on the results. Faust and Rogers (2003) provide a strong rejection of recursive ordering procedures that assume some variables can or cannot respond to other variables in the first period of a shock. They show that if one tests a wide variety of reasonable restrictions on the relationships between the variables, the responses to shocks can vary a great deal. The