EC 572/472 Time Series Analysis and Forecasting
Winter 2017, 6:40-8:30pm TTH (CH-307)
Prof. K.-P. Lin (CH 241G, 725-3931)
Office Hours: 3:30-4:30 TTH & by appointment
(Last updated: 3/08/2017)
Case Study: Predicting U. S. Economic Growth
- Time Series Decomposition
(1.1,
1.2,
1.2a,
1.2b,
1.2c)
- Time Series Smoothing
(1.3,
1.3a,
1.3b,
1.3c,
1.4)
- Trend Regression and Forecasting
(1.1a,
1.1b,
1.1c,
1.1d)
- Stationarity
(1.7,
1.8,
1.9)
- Cointegration
(2.1,
2.2,
2.3)
- ARMA Analysis
(arma,
1.11,
1.12,
1.13,
1.14,
1.14a,
1.14b)
- ARCH Analysis
(dmbp1, dmbp2,
1.16)
- VAR Models
(2.4,
Notes)
Homework
This set of homework works on financial time series analysis.
There are many time series available for monitoring the financial sector of the economy.
In general, they are highly frequent chaotic time series but with embedded long-term structure
of the market. It is served as one of the important leading indicators for predicting the economic
fluctuations and performance. The popular Standard and Poor's 500 Index data will be used in
class for instructional demonstration and homework assignment (named the Index hereafter).
Given the high frequency nature of the Index series, both daily and monthly
(beginning or ending day of the month) series will be used.
For the following homework assigments, we will study the monthly series of Index.
- Analyze the Index series for its trend, seasonal, and random components
using exponential smoothing methods. Alternatively, using linear regression to
estimate the deterministic time trend (linear and quadratic) and seasonality.
Can you use the trend and/or seasonality to predict the Index itself?
Compare and evaluate the forecast performance for exponential smoothing and
linear regression approaches.
Due: January 26.
- Now we have two variables to work with: GDP (from the Case Study) and the Index
(from previous homework). First, the data series of GDP
is available in quarters while the Index is in months (although the daily series are available).
There are at least three methods to convert a monthly series to its quarterly equivalent:
(1) 3 months average, (2) middle month of the quarter, and (3) last month of the quarter.
Compute and test for unit roots for each converted quarterly Index series.
Does GDP cointegrated with the Index? Perform and explain proper cointegration
tests on the regression of GDP or GDP Growth with the Index (and lags if needed).
Due: February 9.
Reminder: Mid Term Exam February 16, in class.
- Identify and estimate the mean process (ARIMA) and variance process (GARCH) of
the Index time series which is now a converted quarterly series.
Hint: the best model you select and estimate must be able to demonstrate the best
forecast performance as well.
Due: March 2.
- Formulate and estimate a time series regression model capable
of predicting GDP with the Index for the next 6 months. Report and evaluate the forecast
performance of your model.
Due: March 16.
Reminder: March 21, Tuesday, final exam in class.