Presentation is loading. Please wait.

Presentation is loading. Please wait.

ESSnet on linking of micro-data on ICT usage

Similar presentations


Presentation on theme: "ESSnet on linking of micro-data on ICT usage"— Presentation transcript:

1 ESSnet on linking of micro-data on ICT usage
Progress Report Mark Franklin UK Office for National Statistics Cologne: 27 October 2011

2 Agenda Context: Brief overview of project Some project issues Q & A
What is the project about? Where does the project sit in the statistical system? Brief overview of project Building on Feasibility Study Some project issues Q & A

3 Making better use of data existing in the statistical system:
Purpose of project Making better use of data existing in the statistical system: Produce new policy relevant indicators without the need to collect more data and without increasing the burden on enterprises. Re-use data for purposes beyond the initial objectives for collecting such data. Focus on economic impacts of ICT usage. But the methodology can be generalised to a range of policy issues and data sources.

4 Where does the project sit in the statistical system?
Project indicators are examples of distributed micro data or meso data Meso data sit between macro and micro forms of data Illustrate by the data-generating processes …

5 adding-up constraints
Macro: data process Surveys “Black box” Compilation process Macro Indicators Admin data Judgements, adding-up constraints etc Macro indicators (national accounts, trade, inflation, public finances etc) cannot be reproduced purely from survey data. Macro indicators are contingent on national accounts conventions (SNA, ESA), e.g. GFCF asset classes. Macro indicators are rich in structure and consistency (with other indicators, and other countries’ data), but poor in detail.

6 Micro: data process Published micro Indicators Clean data, Re-weight etc Run survey [Some NSIs] Micro dataset available to researchers in safe centre Micro indicators can in principle be reproduced purely from survey data. Micro indicators are contingent on survey design, e.g. E-Commerce survey. Micro datasets are rich in detail, poor in consistency and structure. In particular, cross-country analysis of microdata is difficult.

7 Meso: data process Meso Indicators, Country #1 Survey #1, Country #1 Common data-generating Code, Multiple Countries Meso Indicators, Country #2 Survey #2, Country #1 Meso Indicators, Country 3…. Survey #3 …, Country #1 Meso indicators can be reproduced purely from (micro-data versions of) survey data. Design is contingent on survey design, informed by policy relevance, e.g. ICT usage characteristics of firms by quartile of productivity; Cut survey data by industry, size class, whether multinational, young/old etc. Exploits richness of firm-level variation; yet consistent between countries.

8 Example: Should governments subsidise investment in broadband networks?
Evidence based policy making – need evidence on relationship between broadband access and firm performance across a group of countries. Could design a new survey to investigate the relationship (Q1:Do you have access to broadband? Q2: What is your growth of turnover/employment? …), but… Costly Time consuming Difficult to co-ordinate across countries Add to “red tape” burden on survey respondents What’s wrong with using ‘macro’ indicators? Not the same firms! What’s wrong with using ‘micro’ indicators? Cannot identify impacts of policy changes from a single country study Multi-country micro studies are rarer than hens teeth. Structural / “micro” policy also referred to as “supply-side” economics or “Reagonomics”. Concerned with improving the performance of individual firms, with getting more output from scarce resources, with improving productivity. [Bullet #1] [Bullet #2] – why a group of countries? Because we need to control for other changes, like a control group in a drugs trial. [Bullet #3] [Bullet #4

9 Meso: Indicators MexElec – Manufacturing exluding electicals.
This slide shows data from the industry/country datasets. In this case the MEAN of fast-internet penetration by quartile of the TFP distribution in a single year (2004). Ranked by average penetration. Interpretation: Higher productivity quartile of firms generally displays higher DSLPCT… But this is not always the case And there is substantial variation in the level of penetration across the sample.

10 Meso: analysis Meso – Micro plus re-allocation (including the dynamic process by which high performance firms grow at the expense of low-performing firms). Each dot is an observation on an industry/country/year, showing productivity against the share of workers with access to broadband. Regressions show positive and significant coefficients between a range of ICT indicators and productivity at the industry level.

11 Project Overview 15 NSIs. Steering group of 5 NSIs make recommendations to whole group 22 months: December 2010 – October 2012 2 contracted academic partners, plus liaison with other research bodies 7 Workstreams: Co-ordination and financial management (ONS) Metadata Review (ONS) (Lessons for) survey strategies (Stats Norway) Impact analysis (CBS) Dissemination (ONS) Technical infrastructure (Stats Sweden) Data dissemination (CBS) Outer circle – basics – Metadata stage, data assembly and cleaning, mapping to code, code execution, qa of outputs. Middle circle – workstream leaders (data sharing), research leads and champions/mentors. Draft sections of project report. Inner circle – project steering group – decision-making, strategic guidance.

12 Builds on 2006-08 Feasibility Study
Broader scope: More participants Longer runs of annual datasets New datasets, in particular the Community Innovation Survey Develop and generate meso indicators, and conduct some exploratory analysis of ICT impacts using these indicators Develop a schema for providing access to indicators Explore lessons for survey strategies.

13 Project Issues - 1 Choice of indicators and data boundaries (workstreams (b) and (d)) E-commerce variables: a range of different views across the project group over what variables are most relevant CIS variables: initial set of indicators agreed by analytical steering group, coded by academic contractor, being tested by steering group Cycling through metadata-indicators-analysis is a time-consuming process.

14 Project Issues - 2 Data sharing (workstreams (f) and (g))
Meso indicators are not micro-data, but are derived from micro-data, and hence subject to disclosure control Two dimensions to this issue: Internal: Secure FTP platform on which cross-country meso indicators are compiled. Access restricted to analytical steering group, subject to confidentiality agreements. External: Develop a protocol under which the cross-country meso indicators could be made available to outside researchers, and beyond the life of this project.

15 Any questions?

16 Economic Interpretation Division Office for National Statistics
Mark Franklin Economic Interpretation Division Office for National Statistics +44 (0) This work contains statistical data from ONS which is Crown copyright and reproduced with the permission of the controller of HMSO and Queen's Printer for Scotland. The use of the ONS statistical data in this work does not imply the endorsement of the ONS in relation to the interpretation or analysis of the statistical data. This work uses research datasets which may not exactly reproduce National Statistics aggregates.

17 Blank slide

18 Feasibility study on national survey strategies
Workstream C - led by NOR. Objectives to carry out a feasibility study on redesign of national survey strategies, including a study of the existing practices. to present strategies for improving data representativeness including their cost-benefit analysis. The study will cover linked datasets provided by the participating NSIs. Components Analysis of existing surveys and practices to improve representativeness of linked data. Presentation of the main challenges to data linking. Ways to improve representativeness of the linked data.

19 Project time line The approach comes in three main stages.
Metadata checking: Metadata is simply “data about data”. For example some of the metadata collected about variables held by countries include quotes of survey questions, numerical range of the variable, unit of variable, information on the source survey (frequency, sampling frame etc). This is a crucial process because if the same piece of code is to be run in every country then it must know what the characteristics of the data are. More information later. Code development and running: A discussion process with project members. Most productive development takes place at face to face meetings. The project has had a number of meetings where representatives from all of the countries and the project academic gather to make decisions on the content of core code and themes. For example what analysis should the project be focussing on? What variables should be merged? What are the key variables to be studied? The code is distributed to the project members who then run it on their individual firm level data. Results are reported to central location via a secure server. Developing the code is a cyclical process, in that after the code is run there is a stage of fixing bugs and re-releasing the code (sometimes with further analytical powers). For example we are now using version 2.3 of the code. Analysis and reporting of data


Download ppt "ESSnet on linking of micro-data on ICT usage"

Similar presentations


Ads by Google