Lecture 1.4: Nature of Time Series Data, Assumptions under Regression Analysis Asymptotic Properties of OLS in Time Series • Theorems on asymptotic properties that holds for random sampling may not hold for random processes, such as the time series. • The key is whether the correlation between the variables at different time periods tend to zero quickly. • Otherwise, that is, time series may have substantial temporal correlation require special attention in regression analysis. Stationary and Weakly Dependent Time Series Stationary and Weakly Dependent Time Series • These are key concepts that are needed to apply the large approximations in regression analysis with time series data. Stationary Time Series A stationary time series process is one whose probability distributions are stable over time in the following sense: If we take any collection of random variables in the sequence and then shift that sequence ahead h time periods, the joint probability distribution must remain unchanged. Stationary Time Series Definition: The stochastic process 𝑥𝑡 ∶ 𝑡 = 1,2, … is stationary if for every collection of time indices 1 ≤ 𝑡1 < 𝑡2 < ⋯ < 𝑡𝑚 , the joint distribution of 𝑥𝑡1 , 𝑥𝑡2 , … , 𝑥𝑡𝑚 is the same as that of 𝑥𝑡1+ℎ , 𝑥𝑡2+ℎ , … , 𝑥𝑡𝑚 +ℎ for all integers ℎ ≥ 1. Stationary Time Series Implications: • 𝑥𝑡 has the same distribution for all 𝑡 = 1,2, … • In other words, the sequence 𝑥𝑡 ∶ 𝑡 = 1,2, … is identically distributed. But stationarity requires even more… including subsets of the sequence • Note: no restrictions yet for the relationship between 𝑥𝑡 and 𝑥𝑡+ℎ . Stationary Time Series Implications: • If a stochastic process is not stationarity, then is nonstationary process. Stationary Time Series Definition: The stochastic process 𝑥𝑡 ∶ 𝑡 = 1,2, … is covariance stationary if (i) 𝐸 𝑥𝑡 is constant (ii) 𝑉𝑎𝑟 𝑥𝑡 is constant and (iii) for any 𝑡, ℎ ≥ 1, 𝐶𝑜𝑣(𝑥𝑡 , 𝑥𝑡+ℎ ) depends only on ℎ and not on 𝑡 Stationary Time Series Implications: • Thus, covariance stationarity is a weaker form of stationarity and requires only that the mean and variance are constant across time, and the covariance just depends on the distance across time. • To emphasize that stationarity is a stronger requirement than covariance stationarity, the former is referred to as strict stationarity. Stationary Time Series How is stationarity used in time series econometrics?
Technical level: Stationarity simplifies statements of
the law of large numbers and the central limit theorem Stationary Time Series How is stationarity used in time series econometrics?
Practical level: If we want to understand the
relationship between two or more variables using regression analysis, we need to assume some sort of stability over time. Stationary Time Series How is stationarity used in time series econometrics?
Practical level: If we allow the relationship between
two variables (say, 𝑦𝑡 and 𝑥𝑡 ) to change arbitrarily in each time period, then we cannot hope to learn much about how a change in one variable affects the other variable if we only have access to a single time series realization. Weakly Dependent Time Series Definition: A stationary time series 𝑥𝑡 ∶ 𝑡 = 1,2, … is weakly dependent if 𝑥𝑡 and 𝑥𝑡+ℎ are “almost independent” as ℎ increases. Weakly Dependent Time Series Implications: • If for a covariance stationary process 𝐶𝑜𝑟𝑟(𝑥𝑡 , 𝑥𝑡+ℎ ) → 0 “sufficiently quickly” as ℎ → ∞, we’ll say this covariance stationary process is weakly dependent. (asymptotically uncorrelated) • In other words, as the variables get farther apart in time, the correlation between them becomes smaller and smaller. Weakly Dependent Time Series Implications: • Is weak dependence really important for regression analysis? Essentially, it replaces the assumption of random sampling in implying that the law of large numbers (LLN) and the central limit theorem (CLT) hold. Weakly Dependent Time Series Implications: • If for a covariance stationary process 𝐶𝑜𝑟𝑟(𝑥𝑡 , 𝑥𝑡+ℎ ) → 0 “sufficiently quickly” as ℎ → ∞, we’ll say this covariance stationary process is weakly dependent. (asymptotically uncorrelated) • In other words, as the variables get farther apart in time, the correlation between them becomes smaller and smaller. Weakly Dependent Time Series Example 1: Moving average process of order one [MA(1)]: 𝑥𝑡 = 𝛼1 𝑒𝑡−1 + 𝑒𝑡 𝑡 = 1,2, … , 𝑛 where 𝑒𝑡 ∶ 𝑡 = 1,2, … is an independent and identically distributed sequence with mean 0 and 2 variance 𝜎𝑒 . Weakly Dependent Time Series Example 1: Moving average process of order one [MA(1)]: 𝑥𝑡 = 𝛼1 𝑒𝑡−1 + 𝑒𝑡 𝑡 = 1,2, … , 𝑛 Is this stochastic process stationary? Is it covariance stationary? Is it weakly dependent? Weakly Dependent Time Series Example 2: Autoregressive process of order one [AR(1)]: 𝑦𝑡 = 𝜌1 𝑦𝑡−1 + 𝑒𝑡 𝑡 = 1,2, … , 𝑛 where 𝑒𝑡 ∶ 𝑡 = 1,2, … is an independent and identically distributed sequence with mean 0 and 2 variance 𝜎𝑒 . We also assume that 𝐸 𝑦0 = 0. Weakly Dependent Time Series Example 2: Autoregressive process of order one [AR(1)]: 𝑦𝑡 = 𝜌1 𝑦𝑡−1 + 𝑒𝑡 𝑡 = 1,2, … , 𝑛 Is this stochastic process stationary? Is it covariance stationary? Is it weakly dependent? Random Walk Model A case where 𝜌1 = 1: 𝑦𝑡 = 𝑦𝑡−1 + 𝑒𝑡 • A random walk is defined as a process where the current value of a variable is composed of the past value plus an error term defined as a white noise (a normal variable with zero mean and variance one). Random Walk Model A case where 𝜌1 = 1: 𝑦𝑡 = 𝑦𝑡−1 + 𝑒𝑡 • In a random walk model, the series itself is not random. • However, its differences—the changes from one period to the next—are random. • This type of behavior is typical of stock price data. Random Walk Model A case where 𝜌1 = 1: 𝑦𝑡 = 𝑦𝑡−1 + 𝑒𝑡
• In practical terms: If series follows
random walk, original series offers little or no insights. • It would be hard to forecast future trends Weakly Dependent Time Series Example 3: A time series with trend: 𝑦𝑡 = 𝛼0 + 𝛼1 𝑡 + 𝑒𝑡 𝑡 = 1,2, … , 𝑛 where 𝑒𝑡 ∶ 𝑡 = 1,2, … is an independent and identically distributed sequence with mean 0 and 2 variance 𝜎𝑒 . We also assume that 𝐸 𝑦0 = 0. Weakly Dependent Time Series Example 3: A time series with trend: 𝑦𝑡 = 𝛼0 + 𝛼1 𝑡 + 𝑒𝑡 𝑡 = 1,2, … , 𝑛 Is this stochastic process stationary? Is it covariance stationary? Is it weakly dependent? Weakly Dependent Time Series However, if we take the deviation from the mean, 𝑦𝑡 − 𝐸 𝑦𝑡 where 𝑦𝑡 = 𝛼0 + 𝛼1 𝑡 + 𝑒𝑡 𝑡 = 1,2, … , 𝑛 can be stationary time series process, weakly dependent. Weakly Dependent Time Series The previous is an example of: Trend- Stationary Process, that is, a series that is stationary about its time trend as well as weakly dependent.
Thus the random shocks to the process are
transitory and the process is ‘mean reverting’, with the mean Weakly Dependent Time Series How about getting the first difference? ∆𝑦𝑡 = 𝑦𝑡 − 𝑦𝑡−1 where 𝑦𝑡 = 𝛼0 + 𝛼1 𝑡 + 𝑒𝑡 𝑡 = 1,2, … , 𝑛 Can this be a stationary time series process, and weakly dependent?