ESCB experiences with breaks in labour market time series Colm Bates Economist-Statistician European Central Bank ESCB experiences with breaks in labour market time series LABOUR FORCE SURVEY WORKSHOP 2018 Reykjavik; 17 - 18 May 2018
Overview 1. Motivation 2. Overview of responses 3. Visualisation of examples given 4. Methods used to overcome breaks 5. Conclusions
Motivation Framework Regulation Integrating European Social Statistics anticipates breaks NSIs concerned about ability to address breaks Potential interruption to regular monthly unemployment rate Interruptions to MIP indicator and other headline indicators Dedicated session at LFS Methodology Workshop ECB could present a user point of view Questionnaire launched
Overview of questionnaire and responses Series: e.g. LFS.Q.AT.N.UNEMPL.TOTAL0.15_99.T Description of break: Austrian LFS Data have a break between 2003 and 2004 Use of series/intended use of series: descriptive statistics or analysis of microdata. For BMPE, LFS data are not relevant with the exception of unemployment data How did you detect the break: Information by Statistics Austria (substantial change in sampling method, starting with 1Q2004) Treatment of data: Partly, we account for the breaks with backward calculation with growth rates, partly we use EUROSTAT series, which are already adjusted (e.g. NC_UNE_RT_Q.S_ADJ.NSA.AGE.TOTAL.UNIT.THS_PER.SEX.T.GEO.AT) Overview of questionnaire Completed between 03.01.2018 and 26.01.2018 Sought diverse examples of breaks in series Including description of the break How the break was detected How the break was treated Openness was encouraged Similar questionnaire also circulated at ECB. Overview of responses 21 NCBs responded to the WG GES written procedure. 13 NCBs provided examples using the questionnaire. 8 NCBs replied in writing only Those NCBs who replied in writing only specified close working relationships with NSIs as the leading cause to not have problems with breaks in series. ECB replies were mostly focused on national data Types of series and uses A wide range of series Mainly sourced from LFS Regular economic analysis Forecasting Estimation of underutilisation/underemployment Comparison with other data; descriptive statistics
Visualisation of examples given: Reclassifications
Visualisation of examples given: Level shifts
Visualisation of examples given: Missing Values
Visualisation of examples given: Mixed
Methods used to overcome breaks Data interpolation Use of conversions tables and micro data Use of overlapping quarters No intervention Use of alternative sources Use of additive and multiplicative factors Dummy variables
Methods used to overcome breaks Use of overlapping periods to preserve annual growth rates
Methods used to overcome breaks Use of ARIMA modelling and additive factors 3 2 Technicians and associate professionals, Ireland - Does not preserve levels 1
Methods used to overcome breaks Joint ARIMA modelling, with additive factors, based on conversion tables
Methods used to overcome breaks Use of alternative series
Conclusions Þakka þér fyrir! Breaks are an inconvenience to users, they are able to manage those breaks. The methods used are heterogeneous and case specific. Introduction of new Implementing Regulations could disrupt and overwhelm users. There is a good relationships between NCBs and NSIs in the field of labour market statistics, NSIs and NCBs are encouraged to work at a national level to ensure data needs understood and met. Þakka þér fyrir!