Time Series Analysis, Cointegration, and Applications
An important objective of empirical research in macroeconomics is to test hypotheses and estimate relationships, derived from economic theory, among aggregate variables. The statistical theory that was applied well into the 1980s in building and testing large simultaneous-equation models was based on the assumption that the variables in these models are stationary. The problem was that statistical inference associated with stationary processes is no longer valid if the time series are indeed realizations of nonstationary processes.
Granger,C. (2003) Time Series Analysis, Cointegration, and Applications, Les Prix Nobel.