Using Register information to estimate (early) monthly unemployment rates for EU aggregates

Size: px

Start display at page:

Download "Using Register information to estimate (early) monthly unemployment rates for EU aggregates"

Samson Maxwell
6 years ago
Views:

1 Slide 1 of 22 Session 2: Register-based Social Statistics Using Register information to estimate (early) monthly unemployment rates for EU aggregates Carsten Olsson Eurostat F2 "Labour Market Statistics carsten.olsson@ec.europa.eu

2 Slide 2 of 22 Content Introduction Compiling EU unemployment rates Data sources available to Eurostat Method Register data content compared to target What then? Conclusion

3 Slide 3 of 22 Exercise Doomsday book Napoleon Danish government

4 Slide 4 of Introduction! Background " Personal experiences " Professional experiences! What/Where is the limit for using Registers in statistics, or! Why Registers may give the right answer to the wrong question! Case study: Estimation of EU unemployment rates

5 Slide 5 of Compiling monthly EU unemployment rates Aim: Produce comparable unemployment rates # Coverage: whole population # Labour status: as defined by ILO # Frequency: monthly # Delay: short

6 Slide 6 of Compiling monthly EU unemployment rates Main sources for monthly EU unemployment rates: Labour Force Survey (LFS) $ Coverage: whole population* $ Labour status: as defined by ILO Frequency: quarterly Delay 12 weeks (~3 months) Registered with Employment Agencies (EA) Coverage: not well defined population Labour status: according to national rules and needs $ Frequency: monthly $ Delay: less than one month (~3-28 days)

7 Slide 7 of Compiling EU monthly unemployment rates General approach to harmonised monthly unemployment rates by Eurostat: 1st step: Get basic data, i.e. quarterly LFS and monthly registered Unemployment levels from EA (totals broken down by sex and two age-groups) 2 nd step: Benchmark monthly EA data against quarterly LFS (infra-quarterly development) 3rd step: Forecast harmonised rates for most recent data using EA unemployment levels 4th step: Aggregate Member State data to EU totals

8 Slide 8 of 22 EA Registered and LFS time series Czech Republic, Total Unemployment Data Jan-00 Jan-01 Jan-02 Jan-03 Jan-04 Jan-05 Jan-06 Jan-07 Registered LFS

9 Slide 9 of 22 EA Registered and LFS time series Czech Republic, Total Unemployment Data Benchmark Jan-00 Jan-01 Jan-02 Jan-03 Jan-04 Jan-05 Jan-06 Jan-07 Registered LFS Benchmarked estimate

10 Slide 10 of 22 EA Registered and LFS time series Czech Republic, Total Unemployment Data Benchmark Forecast Aggregate 0 Jan-00 Jan-01 Jan-02 Jan-03 Jan-04 Jan-05 Jan-06 Jan-07 Registered LFS Benchmarked estimate Current estimate

11 Benchmarking Ratio of Registe d / LFS (2006) Slide 11 of LFS = Effect of benchmarking is to adjust underlying definitions IE BE SI CZ HU AT BG ES PL PT DK DE LV FR SK LT CY SE MT UK EE RO

12 Benchmarking Ratio of Registe d / LFS (2006) Slide 12 of LFS = IE BE SI CZ HU AT BG ES PL PT DK DE LV FR SK LT CY SE MT UK EE RO

13 Slide 13 of Problems in Eurostat method Problems " LFS not monthly and timely " EA data not same definitions and methodology Consequence " Infra-quarterly development is purely based on EA Registered Unemployment " Most recent development is based on EA Registered Unemployment (-> Revisions)

14 Slide 14 of LFS Characteristics Labour Force Survey: LFS is a big survey! Usually household interviews! About 1.5 million persons covered per quarter! Beyond labour status also: age, marital status, educations, nationality, duration of unemployment, hours worked, work outside normal hours, temporary or contract work, self-employed, seasonal workers, unpaid family workers, lay-offs, etc.! Would be sufficient if monthly, timely, and perfect!!! ". But it is not

15 Slide 15 of LFS Characteristics Typical LFS problems (apart from frequency and timeliness) " Sampling effects " Sample selection (bias), sample size (variance) " Survey mode, response willingness, and nonresponse handling " Estimation/grossing-up procedures

16 Slide 16 of EA data Characteristics Is EA data perfect? Typical Employment Agencies characteristics " Willingness/incentive to register Mean to find a job, but Other means may be just as efficient students, specialist areas, mouth-to-mouth, news papers, internet, Has a job, buy may want a better job Better pay, working conditions, Part-time vs. full-time, Money unemployment and social benefits, but Temporary lay-off (e.g. due to bad weather) Not necessarily able to work (e.g. health) Not necessarily willing to work (in general or in a specific period) " Timeliness and Stickiness

17 Slide 17 of EA data Characteristics Is EA data perfect? Typical Employment Agencies characteristics " Willingness/incentive to register Mean to find a job, but Money unemployment and social benefits, but " Timeliness and Stickiness Register before unemployed (expect need) Do not un-register when starting new job

18 Slide 18 of Definition overlap ILO unemployment level (LFS) Registered with Employment Agency

19 Slide 19 of Definition overlap EU-ILO definition EU-ILO unemployed (LFS) EU-ILO unemployed and registered with Employment Agency Registered Registered with Employment Agency

20 4. Getting monthly EU-ILO unemployment rate right Slide 20 of 22 Ideally apply EU-ILO definition using a Census, but not possible in practice! Instead EA data used! At Macro-level: EA data can be as an (early) indicator/co-variable to support estimation of EU-ILO unemployment level (e.g. Eurostat)! At Micro-level: EA data can be used for pre- or post-stratification (e.g. Nordic countries)

21 Slide 21 of Way forward! Eurostat " Making full use of Registers given the limitations Legal limitations: No access to micro-data of Registers, only macro results Practical limitations: NA " May improve the modelling that links Register data with LFS data " All-in-all: Cannot improve significantly

22 Slide 22 of Way forward! Member States " Produce and deliver monthly LFS data with aim of compiling better EU aggregates Sample design allow monthly estimates Publication (???) " Where access to micro-data of Registers, then do as Nordic countries (pre- and poststratification) " (General trend: More usage of Registers)

23 Slide 23 of Conclusion Shown that Registers produce reliable information for the purpose that they are designed for Comparable statistics may need specific Surveys and cannot rely on general purpose Registers Harmonised Monthly EU-ILO Unemployment rates " EA Registers may improve timeliness of LFS estimates but cannot replace survey results " If access to micro-data, then may improve stratification of a survey Way forward " Monthly surveys (LFS) with appropriate sample design " Supplemented by Register information for stratification purposes where possible

January 2010 Euro area unemployment rate at 9.9% EU27 at 9.5%

January 2010 Euro area unemployment rate at 9.9% EU27 at 9.5% STAT//29 1 March 20 January 20 Euro area unemployment rate at 9.9% EU27 at 9.5% The euro area 1 (EA16) seasonally-adjusted 2 unemployment rate 3 was 9.9% in January 20, the same as in December 2009 4.