时间序列与投资组合理论（英文课件，OFFICE版）--LECTURE19.docVIP

I. Introduction Cointegration is a concept that explains why time series that appear, individually, to behave as random walks, or in a nonstationary fashion, may not drift apart from one another. That is, in combination they may exhibit behavior that appears stationary. There are many applications in economics. A. Economic Applications 1. Purchasing Power Parity The notion of purchasing power parity is that the relative value of two currencies, i.e. their nominal exchange rate, S(t), should reflect their puchasing power over a bundle of commodities in the two contries. The nominal exchange rate or dollar price of the pound is: S(t) = $/￡. The dollar price index for a bundle of goods is PUS(t), and the Pound price index for a bundle of goods is P￡(t). If the bundles are comparable, prices are measured accurately, and we abstract from transportation costs, then the two currencies will have purchasing power parity if: PUS(t) = S(t) P￡(t). Taking logarithms, ln PUS(t) = ln S(t) + ln P￡(t). This relationship should hold in the long run. Note that individually, the logarithms of the price indices in the two countries, and the logarithm of the exchange rate may behave approximately as random walks. However, any deviation, e(t), from the long run equilibrium condition should be stationary: e(t) = ln PUS(t) - ln S(t) - ln P￡(t) = ln[PUS(t)/P￡(t)] - ln S(t). So, the difference between the logarithm of the relative price indices and the logarithm of the exchange rate should be stationary. 2. The Term Structure of Interest Rates A short term interest rate, rS(t), and a long term interest rate, rL(t), may each individually behave in an apparent evolutionary fashion. In a term structure of interest rates, the long rate may be some multiple, ?, of the short rate: rL(t) = ? rS(t) + e(t) where e(t) represents stationary deviations from this long run term structure rela