In 2003, the Royal Swedish Academy of Sciences awarded the Nobel Prize in Economics to Robert Engle and Clive Granger for their methods of analyzing economic time series with a common trend, known as co-integration. Their paper laid the foundation for this field and changed the approach of applied macroeconomists in analyzing data. The concept of co-integration is a natural extension of the idea of economic equilibrium, considering the non-stationarity of most macroeconomic variables. While stationary time series values remain close to their mean, non-stationary variables have an infinite expected return to the mean and can drift far from it. The non-stationarity of most macroeconomic indicators is a well-established empirical fact. Economic equilibrium is often understood as a relationship between several variables, "pushing" some linear combination of these variables to zero so strongly that deviations from zero are very small. Thus, this linear combination of non-stationary variables becomes stationary, and the original variables are co-integrated. Although the concept of co-integration is very natural, the economic methods required for working with it significantly differ from classical economic principles used in microeconomic theory. The differences in methods are so significant that the following paper may surprise a reader familiar with classical economic theory. We begin by noting that most classical regression analysis is based on the concept of exogeneity, whereas co-integration regressions provide consistent estimates even if all variables are endogenous. Moreover, direct and reverse regressions give almost identical results—a thing impossible in microeconomic theory. The complexity of working with co-integration lies in the fact that familiar economic statistics converge to non-standard asymptotic distributions and require non-standard critical values. Engle and Granger show that the desire to avoid these complexities by transitioning to first differences of variables is an erroneous step and leads to significantly biased errors. The bias in estimates arises because the stationary linear combination of non-stationary variables is a necessary regressor in the regression of first differences. This regression is called the error correction model. The authors consider the issue of two-step estimation of the error correction model and the issue of testing co-integration. The ideas and paper of Engle and Granger have separated macroeconomic theory and time series analysis into a separate branch of economics. Robert Engle is also known for his work on stochastic volatility (ARCH and GARCH models), which were named in the official Nobel Committee announcement. Clive Granger is the author of the well-known concept of "Granger causality." The authors worked at the University of California, San Diego for about 30 years before retiring in 2003. Clive Granger passed away in 2009. A. E. Mukuweba Applied Econometrics, 2015, 39 (3), pp. 107–135. Applied Econometrics, 2015, 39 (3), ppIn 2003, the Royal Swedish Academy of Sciences awarded the Nobel Prize in Economics to Robert Engle and Clive Granger for their methods of analyzing economic time series with a common trend, known as co-integration. Their paper laid the foundation for this field and changed the approach of applied macroeconomists in analyzing data. The concept of co-integration is a natural extension of the idea of economic equilibrium, considering the non-stationarity of most macroeconomic variables. While stationary time series values remain close to their mean, non-stationary variables have an infinite expected return to the mean and can drift far from it. The non-stationarity of most macroeconomic indicators is a well-established empirical fact. Economic equilibrium is often understood as a relationship between several variables, "pushing" some linear combination of these variables to zero so strongly that deviations from zero are very small. Thus, this linear combination of non-stationary variables becomes stationary, and the original variables are co-integrated. Although the concept of co-integration is very natural, the economic methods required for working with it significantly differ from classical economic principles used in microeconomic theory. The differences in methods are so significant that the following paper may surprise a reader familiar with classical economic theory. We begin by noting that most classical regression analysis is based on the concept of exogeneity, whereas co-integration regressions provide consistent estimates even if all variables are endogenous. Moreover, direct and reverse regressions give almost identical results—a thing impossible in microeconomic theory. The complexity of working with co-integration lies in the fact that familiar economic statistics converge to non-standard asymptotic distributions and require non-standard critical values. Engle and Granger show that the desire to avoid these complexities by transitioning to first differences of variables is an erroneous step and leads to significantly biased errors. The bias in estimates arises because the stationary linear combination of non-stationary variables is a necessary regressor in the regression of first differences. This regression is called the error correction model. The authors consider the issue of two-step estimation of the error correction model and the issue of testing co-integration. The ideas and paper of Engle and Granger have separated macroeconomic theory and time series analysis into a separate branch of economics. Robert Engle is also known for his work on stochastic volatility (ARCH and GARCH models), which were named in the official Nobel Committee announcement. Clive Granger is the author of the well-known concept of "Granger causality." The authors worked at the University of California, San Diego for about 30 years before retiring in 2003. Clive Granger passed away in 2009. A. E. Mukuweba Applied Econometrics, 2015, 39 (3), pp. 107–135. Applied Econometrics, 2015, 39 (3), pp