0% found this document useful (0 votes)
66 views38 pages

Seasonality: T T T T

This document discusses seasonality in time series data. Seasonal components (St) are repetitive cycles over a calendar period like a year. Seasonality can be deterministic (predictable) or stochastic. Examples of seasonal time series include unemployment, gasoline sales, and housing starts. Seasonal adjustment removes the seasonal component (St) to focus on trend and cycle. Deterministic seasonality can be modeled using seasonal dummy variables in a regression. This estimates the magnitude of the seasonal effect for each period.

Uploaded by

tamann2004
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
66 views38 pages

Seasonality: T T T T

This document discusses seasonality in time series data. Seasonal components (St) are repetitive cycles over a calendar period like a year. Seasonality can be deterministic (predictable) or stochastic. Examples of seasonal time series include unemployment, gasoline sales, and housing starts. Seasonal adjustment removes the seasonal component (St) to focus on trend and cycle. Deterministic seasonality can be modeled using seasonal dummy variables in a regression. This estimates the magnitude of the seasonal effect for each period.

Uploaded by

tamann2004
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 38

Seasonality

• Recall that we said that it can be useful to 
describe the mean of a time series as the sum 
of components
μt = Tt + St + Ct
where  St is the seasonal component.
• The seasonal component  St  is a repetitive  
cycle over the calendar year
• Seasonality  St can be deterministic 
(predictable) or stochastic
Seasonality – Examples
• Gasoline consumption rises in summer due to 
increased auto travel
• International airline prices rise in summer due to 
increased tourism
• Natural gas consumption and prices rise in winter due 
to heating
• Electricity consumption increases in summer due to air 
conditioning
• Construction activity and jobs decrease in winter in the 
Midwest 
• Consumer spending increases in November and 
December due to holiday shopping
Deterministic vs Stochastic Seasonality
• If the seasonal pattern repeats year after year, it is 
deterministic and predictable.
– Christmas is always in December
• If the seasonal pattern roughly repeats itself, but 
evolves over the years, it is stochastic and only partially 
predictable
– Holiday shopping as a percentage of income is not a fixed 
constant
• Seasonal patterns can change dramatically as the 
economy evolves
– The spread of air conditioning shifted the seasonal pattern 
of residential electricity consumption from winter to 
summer
Seasonal Adjustment
• Most economic indicators reported by the government 
are seasonally adjusted. 
• Roughly, the component  St is estimated, and then 
what is reported is
yt* = yt − St
= Tt + Ct
• The idea is that seasonality distracts from the main 
reporting purpose
– Seasonally adjusted data allows users to focus on trend 
and business cycle movements
• Seasonal adjustment by central statistical agencies is 
sophisticated, allowing for evolving seasonal patterns.
Examples of Seasonal Time Series
• First Example:
• U.S. Unemployment Rate
– Men, 20+ years
– 1948‐present
– Not seasonally adjusted
U.S. Unemployment Rate
Men, 20+ years, 1948‐2009
Unemployment Rate, 1990‐2001
Unemployment Rate, by year
Example 2
U.S. Gasoline Sales Volume
Gasoline Sales, by year
Example 3
U.S. Housing Starts
(New Privately Owned Housing Units)
Housing Starts, by year
Deterministic Seasonality
• If seasonality is constant and deterministic then St
is simply a different constant for each period
• For example, for monthly data
⎧ γ1 if t = January
⎪γ t = February
⎪ if
St = ⎨ 2
⎪M M
⎪⎩γ 12 if t = December
• Seasonality is a constant which varies by the 
calendar period (quarter, month, week, day, or 
time of day)
Fitted Values and Forecasts
Pure Deterministic Seasonality
• In the simple pure deterministic seasonality 
model, fitted values and forecasts are the 
simple seasonal pattern
Example – Housing Starts

January 91
February 95
March 127
April 144
May 150
June 148
July 142
August 140
September 132
October 137
November 114
December 96
Seasonal Dummy Model
• Deterministic seasonality  St can be written as a 
function of seasonal dummy variables
• Let  s  be the seasonal frequency
– s=4 for quarterly
– s=12 for monthly
• Let  D1t , D2t , D3t ,…, Dst be seasonal dummies
– D1t= 1  if s  is the first period, otherwise D1t= 0
– D2t= 1  if s  is the second period, otherwise D2t= 0
• At any time period  t, one of the seasonal 
dummies D1t , D2t , D3t ,…, Dst will equal 1, all the 
others will equal 0.
Seasonal Dummy Model
• Deterministic seasonality
⎧ γ1 if t = January
⎪γ t = February
⎪ if
St = ⎨ 2
⎪M M
⎪⎩γ 12 if t = December
s
= ∑ γ i Dit
i =1

a linear function of the dummy variables
Estimation
• Least squares regression
s
yt + h = ∑ γ i Dit + et
i =1
s −1
= α + ∑ β i Dit + et
i =1
• You can either
– Regress  y on all the seasonal dummies, omitting the intercept, 
or
– Regress  y on an intercept and the seasonal dummies, omitting 
one dummy (one season, e.g. December)
• You cannot regress on both the intercept plus all seasonal 
dummies, for they would be collinear and redundant.
Interpreting Coefficients
• In the model
s −1
St = α + ∑ β i Dit
i =1

the intercept  α=γs is the seasonality in the 
omitted season. 
• The coefficients  βi=γi ‐γs are the difference in 
the seasonal component from the s’th period.
STATA Programming
• If the time index is  t and is formatted as a 
time index, you can determine the period 
using the commands
generate m=month(dofm(t))
generate q=quarter(dofq(t))
for monthly and quarterly data, respectively
(See dates and times in STATA Data manual)
Creating Dummies
• If  m is the month (1 for January, 2 for February, etc.), then
– generate m1=(m==1)
– This creates a dummy variable “m1”  for January
– Then
– regress y m1 m2 m3 m4 m5 m6 m7 m8 m9 m10 m11
or
– regress y m1 m2 m3 m4 m5 m6 m7 m8 m9 m10 m11 m12, noconstant
• Easier
– Type “b12.m” in the regressor list
– regress y b12.m
– This includes dummies for months 1 through 11, omits 12
– Same as mechanically listing the eleven dummies, but easier.
– It is important that “m” be the numerical month (1 for January, 2 for 
February, etc.)
Estimation
Estimated Seasonality – Housing Starts

January 91
February 95
March 127
April 144
May 150
June 148
July 142
August 140
September 132
October 137
November 114
December 96
Housing Starts, by year, and estimated 
seasonality
Predicted Values
Example 1
Unemployment Rate
Unemployment Rate, by year, and 
estimated seasonality
Predicted Values
Example 2
Gasoline Sales
Gasoline Sales, by year, and estimated 
seasonality
Predicted Values
Application – Weekly Data
• Unemployment Insurance Claims
• Department of Labor
• Issued Weekly
• Important indicator for unemployment
Unemployment Claims
Not Seasonally Adjusted
Unemployment Claims
Official Seasonally Adjusted Series
Estimation
Estimated Seasonal Process
Seasonally Adjusted 
(by Dummy Variable Method)
Other types of seasonality
• Daily data
– Day of the week
– Handle by including dummy variables for each day
• High‐frequency data
– Include hourly or time‐of‐day indicators
• Holiday effects
– Flower sales big on Valentines Day, Mothers Day, Easter, 
yet these days can move around
– Trading‐day/business‐day variation 
• Number of trading days/business days varies across months
• Can divide by number of trading days, or include as a regressor

You might also like