Outline I. Trends: Linear vs. Exponential II Using Trending Variables in Regression III. Detrending Interpretation IV. R‐Squared with trending y V. Seasonality
12. Trends and Seasonality Time Series Analysis Read Wooldridge, (2013) Chapter 10.5
I. Trends 12. Trends and Seasonality . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat
• An economic time series have a common tendency of growing overtime. Thus, some time series contain time trend.
•
• Models capturing trending behavior:
•
{et } is an independently, identically distributed (i.i.d.) sequence of unobservable.
II. Using III. Detrending IV. R-Squared V. Seasonality
What is 1? Let et = 0
yt = yt – yt-1 = 1 • •
12. Trends and Seasonality . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat
Write a time series {yt } as
yt = 0 + 1t + et
– (1) Linear time trend – (2) exponential trend
I. Trends
2
Linear time trend
I. Trends
I. Trends
II. Using III. Detrending IV. R-Squared V. Seasonality
12. Trends and Seasonality . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat
3
If 1 > 0, then yt has an upward trend (growing overtime) If 1 < 0, then yt has a downward trend. (shrinking overtime)
I. Trends I. Trends
II. Using III. Detrending IV. R-Squared V. Seasonality 12. Trends and Seasonality . Quantitative Methods of Economic Analysis . 2949605 . Chairat Aemkulwat
4
Linear time trend
Usefulness of a linear time trend: Example: Linear time trend Let baht$ be the exchange rate (bath/$) $ = 16.0 + .4143t s.e. (1.37) (.054) t‐stat [11.64] [7.61]
• yt = 0 + 1t + et • E(yt) is linear in t. yt = 0 + 1t + et E(yt) = 0 + 1t
n=43 (1960‐2002), R2=.585749, R2‐bar=.575645
• If 1 > 0, then yt has an upward trend (growing overtime) • If 1 0; 2 0 3