Transcript D3.1
International Seminar on Timeliness,
Methodology and Comparability of
Rapid Estimates of Economic Trends
The availability, timeliness and quality
of rapid estimates
Michel Girard – Statistics Canada
May 27, 2009
Overview
2
Indicators
Consultation with stakeholders
Flash estimates
Quality
Framework for indicators
Qualitative statistics
Databases and data management
Statistics Canada • Statistique Canada
2016-03-31
Indicators
Business registry
Indicators
•
•
•
•
•
3
Industry
Prices
Employment (lagging)
Financial
Administrative data
Qualitative statistics
Opinion surveys or sentiment indicators (leading)
National accounts statistics
Leading indicators
Statistics Canada • Statistique Canada
2016-03-31
Indicators – Multidimensional needs
Infra-annual and seasonal adjustment
Administrative data to fill data gaps
Turnover and prices (survey)
Sales and financial statistics - Private sector data e.g.
credit card sales, Financial Reuters and Bloomberg
GDP
Opinion survey or sentiment indicators
• For example, about construction or new orders
Loans and public debt
Linking financial and business sectors data
4
Statistics Canada • Statistique Canada
2016-03-31
Consultation with stakeholders
STC – Concerns about stakeholders’ needs
• Adequacy of data packages to analysts / policy makers
Large number of indicators with a lag of 42 to 60 days
• Preliminary actions - Highlighting the right data
Emphasis on financial data products
Ratios to facilitate analysis
Awareness of the financial accounts and balance sheet
products
– workshops and analytical paper
5
Statistics Canada • Statistique Canada
2016-03-31
Consultation process
Formal consultation process
• Departments, B.o.C, academic, forecasters, advisory
committees, business analysts
• Users are not comfortable with the STC making
unilateral decisions on the product line
• NSO should not rush into short-term collection with
testing
• Short term ad hoc measures will not be useful unless
there is a time series continuity
• Continue to focus on long term investments in quality
improvements and new data products
6
Statistics Canada • Statistique Canada
2016-03-31
Consultation outcome
Short-term
•
•
•
•
•
•
Investment survey
Feasibility study: Monthly transactions in international securities
Investigate capacity to advance some releases
Development of a module (LFS) about labour market transitions
Discussions about the potential effects of regulatory reform
Long-term
Metadata
Workshops
Web site
7
Statistics Canada • Statistique Canada
2016-03-31
Flash estimates
A flash estimate is an early estimate produced and published as soon
as possible after the end of the reference period, using a more
incomplete set of information than the set used for the traditional
estimates – EUROSTAT
Flash estimates are to be released 30 days after reference period Korea
Differences between “flash” and “traditional” estimates are defined
according to 5 dimensions: timeliness, accuracy, coverage, sources and
methods
EUROSTAT paper presents the estimation procedures behind the:
• Quarterly GDP
• Monthly Union Index of Consumer Prices
Paper deals with issues related to the lack of basic information: use
movement of a smaller sample, historical and partial information as well as
the use of regressions and other mathematical methods
8
Statistics Canada • Statistique Canada
2016-03-31
Flash estimates
Revision analysis to assess the performance of flash estimates
New information do not add significantly to flash estimates;
improvements to timeliness do not necessarily come at the expense
of lower reliability
Working on a 30-60-90 days schedule framework
• Compare to the current schedule that extend to 123 days.
• Flash GDP (30 days)
• Main expenditure, output and income components of GDP (60
days),
• Complete set of quarterly accounts, including financial and nonfinancial accounts for institutional sectors (90 days)
9
Statistics Canada • Statistique Canada
2016-03-31
Flash estimates
Similar conclusion about the quarterly GDP in Korea
Using econometric techniques, administrative data, rapid survey indicators
Flash estimate released after 25 days is very close to the one released after
70 days
Would be interested to know if the GDP releases by EUROSTAT and Korea
are accompanied by a text explaining which components of GDP are
responsible for the fluctuations. BEA provide such detail with the “advance”
GDP release.
Canada has a more integrated approach where complete set of accounts is
available after 75 days after reference period, monthly and quarterly GDP
are available 60 days after reference period.
10
Statistics Canada • Statistique Canada
2016-03-31
Quality of signals
To what extent, when facing incomplete information, extrapolation
techniques or other (assumptions) are capable of capturing the
correct movement in a severe changing economic context
Issue well covered in the BIS paper when it refers to the extrapolation of previous months’ year-over-year changes
in the CPI to gain timeliness
BIS mentioned the necessity to conduct research to understand the new reaction:
•
The market reaction to bank rescue packages
•
Given the severity of current crisis and synchronised economic downturn, historical models and data will
prove to be poor reference for the future as future business cycles will certainly be different.
US paper about the advance estimates of GDP:
•
Revisions must be judged against average growth and volatility in growth.
•
Seasonal factors largest source of revisions
•
Quality assessed on direction (98%), acceleration or deceleration (74%) above/near/below trend (60%)
•
My interpretation is that larger revisions could be acceptable as long as long as the story does not change
•
•
11
What is the quality of the signal in a rapid changing economic environment?
Judgemental trend in the area of services – How to ensure consistency over time?
Statistics Canada • Statistique Canada
2016-03-31
Incomplete information
Issue of low response or simply data source availability
Widespread approach is to prepare flash estimates using series that
are representative of an overall; introduce new information gradually
in the calculation
Use of extrapolation techniques or other mathematical techniques
Need for clear notes about the methodologies
Present statistics on the basis of reported data and at country group
level and/or main players – UNCTAD
Flash estimates are generally for aggregates
12
Statistics Canada • Statistique Canada
2016-03-31
Qualitative statistics
Early Warning System – Korea
To identify symptoms of an economic crisis in
advance
EWI is compiled by transforming a composite
index into a scale from 0 to 5
• Normal: 0-1
• Caution: 3-4
• Danger: 4-5
13
Statistics Canada • Statistique Canada
2016-03-31
Qualitative statistics
Similar approach exist in the Netherlands
Long-term trend and cycle-trend of a series of
indicators are calculated
Examine the cycle to see if:
• Above or below long-term trend
• Increasing or decreasing
14
Statistics Canada • Statistique Canada
2016-03-31
15
Statistics Canada • Statistique Canada
2016-03-31
Framework of indicators
Indicators should be given a context
Business Cycle Tracer – 15 macroeconomic indicators
Export Condition Monitor – 6 short-term indicators
10 additional indicators putting the BCT and ECM in a
more general context
•
•
•
•
•
16
2 financial indicators (stock exchange and loans),
2 indicators of raw material prices,
2 indicators in the field of industry and construction,
2 indicators in the field of commercial services,
2 indicators for the real estate market
Statistics Canada • Statistique Canada
2016-03-31
Databases
Paper from the World Bank reported on the GDDS and
SDDS databases.
• To facilitate dissemination in standards formats of quarterly
external debt data
• To support macroeconomic analysis and cross-country
comparison.
• Public and publicly-guaranteed external debt
Interagency Group on Economic and Financial Statistics
• BIS, ECB, Eurostat, IMF, OECD, UN, World Bank
• Make data more accessible/timely via common Website
• New data needs
17
Statistics Canada • Statistique Canada
2016-03-31
Data management
Common definitions, concepts
Harmonized classifications
Introduce new information gradually
Automated data exchange
Sharing methodologies - rapid estimates
Creating processing capacity – UNCTAD
Creating analytical capacity to confront statistics quickly before
release, even if a flash estimate
• IT component
• Human resources component
18
Statistics Canada • Statistique Canada
2016-03-31