SHORT-TERM ECONOMIC STATISTICS EXPERT GROUP (STESEG)
28 – 30 JUNE 2004
DATA PRESENTATION AND SEASONAL
ADJUSTMENT
- DATA AND METADATA PRESENTATION
TERMINOLOGY -
1
TERMINOLOGY
Background
• Provision of metadata by national agencies and
I/Os essential to enable users to assess
relevance of data to their needs.
• Metadata in international context is essential to
compare data and practices across countries.
• Terminology is a key element of metadata
2
TERMINOLOGY
What is “terminology”?
• Refers to information providing meaning to terms
used in a particular subject.
• In context of statistics this means information
about concepts, variables, etc.
3
TERMINOLOGY
What information is required?
•
•
•
•
Concept label
Concept definition (provides meaning)
Source (seldom provided)
Context (appropriate use, background,
limitations)
4
TERMINOLOGY
How is terminology developed?
Stovepipe approach in national agencies and I/Os
Nat.
accounts
Concept A
Prices
Concept A
Int.
trade
Concept A
Labour
market
Concept A
Exchange
rates
Concept A
5
TERMINOLOGY
Int. glossaries (OECD, Eurostat, MCV)
Concept A
Corporate glossaries
Concept A
Nat.
accounts
Prices
Int.
trade
Labour
market
Exchange
rates
6
TERMINOLOGY
- FEEDBACK ON RECOMMENDATIONS -
Recap on feedback
• Almost all agree on the central role of
terminology and the need to provide clear
standards for key data presentation concepts
• Some commented on the need to include clear
definitions of concepts, especially growth rates in
publications to inform users
• Some of the suggested definitions required
rewording to make them clearer
7
Task Force Recommendations on
Terminology
Agreement
Modificatn
required
Year-on-year growth rate
Annualised growth rate
Linear approximation of the annualised figure
Calendar or working day adjustment
Moving average
Preliminary / provisional
Cycle (in a time series)
Oscillation
Delete term
Seasonal variation
Time series
Trend
Trend-cycle
Calendar effects component
Irregular component
Seasonally adjusted component or series
8
Year-on-year growth rates (changes)
are rates expressed over the
corresponding period (month or quarter)
of the previous year. Such rates
(changes) may be expressed as Qt/Qt-41 or Mt/Mt-12-1 (Qt-Qt-4 or Mt-Mt-12)
Year-on-year growth rates are rates of
change expressed over the
corresponding period of the previous
year. Such rates may be expressed as
Qt/Qt-4-1 or Mt/Mt-12-1 (Qt-Qt-4 or MtMt-12)
Month-on-month growth rates are
rates of change expressed over the
previous month. Such rates may be
expressed as Mt/Mt-1-1
Quarter-on-quarter growth rates are
rates of change expressed over the
previous quarter. Such rates may be
expressed as Qt/Qt-1-1
Annual growth rates (annual change)
are rates of change expressed over the
previous year. Such rates (changes) may
be expressed as Yt/Yt-1-1 (Yt-Yt-1).
Annual growth rates (annual change)
are rates of change expressed over the
previous year. Such rates (changes) may
be expressed as Yt/Yt-1-1 (Yt-Yt-1).
9
Annualised growth rates show the
value that would be registered if the
rate of change measured for a
month or quarter were maintained
for a full year, i.e. . [((Qt/Qt-1)4)-1],
[((Mt/Mt-1)12)-1]. Such rates
facilitate comparison of data for
different time periods (e.g. years
and quarters).
Annualised growth rates show the
value that would be registered if the
rate of change measured for a
month or quarter were maintained
for a full year. Such rates facilitate
comparison of data for different time
periods (e.g. years and quarters).
The term “Annualised growth rate” is
The term “Annualised growth rate” is sometimes used to described the
sometimes used to described the
quarterly growth rate multiplied by
quarterly growth rate multiplied by
four as opposed to compounding
four as opposed to compounding
the quarterly growth rate. This is
the quarterly growth rate. This is
more appropriately referred to as
more appropriately referred to as
“linear approximation of the
“linear approximation of the
annualised figure”.
annualised figure”.
10
Seasonal adjustment is a statistical
technique to remove the effects of
seasonal calendar influences operating
on a series. Seasonal effects usually
reflect the influence of the seasons
themselves either directly or through
institutional factors or social conventions.
Seasonal adjustment is a statistical
technique to remove the effects of
seasonal calendar influences operating
on a series. Seasonal effects usually
reflect the influence of the seasons
themselves either directly or through
institutional factors or social conventions.
Other types of calendar variation occur
as a result of influences such as the
number of days in the calendar period,
the accounting or recording practices
adopted or the incidence of moving
holidays (such as Easter).
Other types of calendar variation occur
as a result of influences such as the
number of days in the calendar period,
the accounting or recording practices
adopted or the incidence of moving
holidays (such as Easter).
No change
11
Calendar adjustment refers to the
correction for calendar variations other
than seasonal factors, e.g. number of
days in the calendar period, the
accounting or recording practices
adopted or the incidence of moving
holidays (such as Easter).
Calendar adjustment refers to the
correction for calendar variations other
than seasonal factors, e.g. number of
days in the calendar period, the
accounting or recording practices
adopted or the incidence of moving
holidays (such as Easter).
The terms “calendar adjustment” and
“working day adjustment” are often used
interchangeably. However, there is a
subtle difference between the two terms
as working day adjustment is merely one
type of calendar adjustment, along with
an adjustment for say, new recording
practices.
The terms “calendar adjustment” and
“working day adjustment” (also known as
“trading day adjustment”) are often used
interchangeably. However, there is a
subtle difference between the two terms
as working day adjustment is merely one
type of calendar adjustment, along with
an adjustment for say, new recording
practices.
Do new recording practices addressed as
part of calendar adjustment represent a
substantive program change that should be
addressed explicitly in their own right?
12
A cycle in a time series refers
to smooth variations around the
trend revealing a succession of
phases of expansion and
recession. The cyclical
component can be viewed as
those fluctuations in a time
series which are longer than 1½
years but shorter than those
attributed to the trend.
A cycle in a time series refers
to smooth variations around the
trend revealing a succession of
phases of expansion and
contraction. The cyclical
component can be viewed as
those fluctuations in a time
series which are longer than 1½
years but shorter than those
attributed to the trend.
Recession is a specialised term in business
cycle analysis. Possible to have economic time
series that have cycles that are not the same as
business cycle in timing. Better to use the more
general term “contraction”.
13
A time series is a set of ordered
observations on a quantitative
characteristic of an individual or
collective phenomenon taken at
different points of time.
A time series is a set of timeordered observations on a
quantitative characteristic of an
individual or collective
phenomenon taken at different
points of time.
The trend is the component that
represents the long-term
variations in a time series. Trend
can be viewed as those
variations of very low
frequencies
The trend is the component that
represents the long-term
variations in a time series. In the
frequency domain, trend can be
viewed as those variations
corresponding at very low
frequencies
14
The calendar effects component is the
component that represents the calendar
variations in a time series, such as
trading days, moving holidays and other
calendar effects (such as leap year). The
effects of the normal length of a month
are assigned to the seasonal
component.
The calendar effects component is the
component that represents the calendar
variations in a time series, such as
trading days, moving holidays and other
calendar effects (such as leap year). The
effects of the normal length of a month
or quarter are assigned to the seasonal
component.
A seasonally adjusted component or
series is the result of the extraction of
the seasonal component and the
calendar effects component from a time
series. If neither seasonal nor calendar
influences are present in the raw data,
the seasonally series is given by the raw
data. For series with no identifiable
seasonal variations but with identifiable
calendar variations, the seasonally
adjusted series is given by the calendar
adjusted series.
A seasonally adjusted component or
estimate is the result of the extraction of
the seasonal component and the
calendar effects component from a time
series. If neither seasonal nor calendar
influences are present in the raw data,
the seasonally series is given by the raw
data. For series with no identifiable
seasonal variations but with identifiable
calendar variations, the seasonally
adjusted series is given by the calendar
adjusted series.
15
The calendar effects component is the
component that represents the calendar
variations in a time series, such as
trading days, moving holidays and other
calendar effects (such as leap year). The
effects of the normal length of a month
are assigned to the seasonal
component.
The calendar effects component is the
component that represents the calendar
variations in a time series, such as
trading days, moving holidays and other
calendar effects (such as leap year). The
effects of the normal length of a month
or quarter are assigned to the seasonal
component.
A seasonally adjusted component or
series is the result of the extraction of
the seasonal component and the
calendar effects component from a time
series. If neither seasonal nor calendar
influences are present in the raw data,
the seasonally series is given by the raw
data. For series with no identifiable
seasonal variations but with identifiable
calendar variations, the seasonally
adjusted series is given by the calendar
adjusted series.
A seasonally adjusted component or
estimate is the result of the extraction of
the seasonal component and the
calendar effects component from a time
series. If neither seasonal nor calendar
influences are present in the raw data,
the seasonally series is given by the raw
data. For series with no identifiable
seasonal variations but with identifiable
calendar variations, the seasonally
adjusted series is given by the calendar
adjusted series.
Issue is with idea of calling a calendar adjusted series
with no identifiable seasonal variation a “seasonally
adjusted series”
“Estimate” highlights nature of series as an
analytical product based on original data
16 and which
are subject to estimation errors
The trend cycle is the component
that represents the variations of low
frequency in a time series, the high
frequency variations having been
filtered out. This component can be
viewed as those variations with a
period longer than a chosen
threshold (usually 1½ years). In
practice, statistical agencies
estimate trend-cycle by filtering the
seasonal and irregular component
The trend cycle is the component
that represents the variations of low
frequency in a time series, the high
frequency variations having been
filtered out. This component can be
viewed as those variations with a
period longer than a chosen
threshold (usually 1½ years). In
practice, statistical agencies
estimate trend-cycle by estimating
and removing the seasonal and
irregular component
Is there some ambiguity in the current
definition in not accounting for
fluctuations in a time series of more
than a year but less than 1½ ?
17
SUMMARY OF WRITTEN COMMENTS
Future work
Data and Metadata Reporting
and Presentation Manual
Terminology
Growth rates
Seasonal
adjustment
18
TERMINOLOGY
Future work
• Final definitions will be incorporated into OECD
Glossary of Statistical Terms – close relationship
with CODED
• Terminology will also link into work of the SDMX
partnership with Eurostat, IMF, ECB, BIS, World
Bank, UNSD
• Will be incorporated as required into the SDMX
Metadata Common Vocabulary (MCV)
19