Using EUROMOD to nowcast risk of poverty

in the EU

Jekaterina Navicke, Olga Rastrigina and Holly SutherlandISER, University of Essex

2013 EUROMOD research workshopLisbon, 2 October 2013

Motivation & aim Context Toolbox Results Further steps

Outline:

Problem: 2-3-year time lag in the production of EU-SILC statistics

Timely indicators would: Promote distributional issues when assessing current socio-economic

conditions Facilitate monitoring of current policy reforms/problems Help assess progress towards Europe 2020 target Not a substitute for more timely data collection and processing! As any forecast should be treated with caution

Aims: To predict what the EU-SILC will show when the data on current income

are available To develop methods that can be applied quickly and updated easily for

EU27 To estimate the direction and scale of movement of key income-based

indicators (median, risk-of-poverty, inequality, etc.)

Motivation & aim

Dec 2012: paper at the NetSILC2 conference develop the method tested on EU-SILC 2008 (2007 income) 8 countries: Estonia, Greece, Spain, Italy, Latvia, Lithuania, Portugal,

Romania nowcast for 2010-2012 Eurostat working paper:

http://epp.eurostat.ec.europa.eu/cache/ITY_OFFPUB/KS-RA-13-010/EN/KS-RA-13-010-EN.PDF

July 2013: EUROMOD working paper focus on validation https://www.iser.essex.ac.uk/publications/working-papers/euromod/em11-

13

By Dec 2013: SSM research note (work in progress) improvements to current methodology application on EU-SILC 2010 (2009 income) more countries (+ Germany, Finland, …) nowcast 2011-2013

Context:

EUROMOD simulation

Adjusting EUROMOD to account for employment changes

Calibration to align EUROMOD and EU-SILC

No data adjustments to account for demographic changes

Ad hoc and country specific adjustments kept to the minimum

Toolbox:

EUROMOD - static tax-benefit microsimulation model for the EU: Unique: consistent results across 27 Member States Operates on anonymized EU-SILC cross-sectional micro-data Scope: income taxes, social contributions and cash benefits For details see EUROMOD Country Reports:

https://www.iser.essex.ac.uk/euromod/resources-for-euromod-users/country-reports

Simulation: Tax and benefit policies simulated up to 2012 (as of June 30th); Non-simulated benefits and original incomes are updated from 2007 to

2012 using indexes (earnings, CPI etc.) plus official projections. Updating disaggregated where possible (e.g. earnings by sector).

Toolbox (1): EUROMOD simulation

Fig 1 Nominal proportional changes in average gross employment income (EUROMOD and EU-SILC) and compensation per employee (AMECO), EUR

Notes: Chain growth. EU-SILC numbers are lagged by one year to correspond to the income reference year. Statistics on compensation per employee obtained from the annual macro-economic dataset of DG ECFIN (AMECO).

Adjusting EUROMOD input data (SILC 2008) for employment changes (2008-2012)

Based on LFS data: Trade-off between more up-to-date and more detailed data We use published LFS employment figures

(in 2012: annual up to 2011 & rolling quarterly average for 2012) Concepts do not align perfectly between SILC and LFS => = > Aim is not to align LFS and SILC, but model relative changes

Steps: modelling employment transitions

(net changes in employment rates modelled within 18 stratum by age, gender, educational status: random selection + 200 replications for more robust results)

modelling share of long-term unemployment to capture changes in eligibility for benefit receipt (similar method)

adjusting labour market characteristics in the EUROMOD data & simulating benefits.

Toolbox (2): Employment Adjustments

Fig 2 Employment rates in the LFS, EU-SILC and in EUROMOD before and after labour market adjustments

Notes: EU-SILC numbers are lagged by one year to correspond to the income reference year.

Estimates based on EUROMOD diverge from EUROSTAT even in the baseline year. Sources of discrepancy include (Figari et al. 2012):

Version of the SILC data Slightly different definition of disposable income Non take-up or leakage of means-tested benefits; tax evasion. Reporting errors in the data or reference time period mismatches Simulation error due to low quality or lack of information in the data EUROMOD adjusts household composition to correspond to income year

(babies born since income reference period are dropped)

Calibration: Household-specific calibration factor Factor is calculated based on 2007 income data and applied to 2008-2012 Calibration on average improves predictions of both levels and changes

Toolbox (3): Calibration to EU-SILC

Results (1): POVERTY RISKFig 3 EUROMOD 2007-2012 and EU-SILC 2007-2010: At risk of poverty rates (using 60% median as the threshold)

Notes: EU-SILC numbers are lagged by one year to correspond to the income reference year

We focus on direction and scale of movement in indicators relative to the latest available EU-SILC estimates (not on 2012 levels).

3 main reasons for this: Discrepancies between the EUROMOD and EU-SILC estimates still

remain after adjusting for employment and calibration. Wide confidence intervals around AROP point estimates in the EU-

SILC: (standard errors vary from 0.4 pp for IT, ES to 0.9 pp for LT). Nowcasts of direction and scale of change are more reliable:

reduction in the standard errors due to covariance in the data.

Results (2): NOWCAST

Results (3): NOWCASTED CHANGEChange in 2010-2012 (i.e. since the income year of latest SILC statistics)

Notes: *** p < 0.001, ** p < 0.01, * p < 0.05. Information on the sample design of EU-SILC 2008 used for calculations was derived following Goedemé (2010) and using do files Svyset EU-SILC 2008 provided at: http://www.ua.ac.be/main.aspx?c=tim.goedeme&n=95420. Standard errors around AROP indicators are based on the Taylor linearization using the DASP module for Stata.

Poverty rates (60% of median)

 Median Income All Males

Females

Children

(<18)Prime-

ageElderly (65+)

Estonia 13.3%*** 0.62 -0.77 1.80*** -1.77* -1.57** 9.79*** Greece -21.0%*** 1.38** 1.85** 0.91 3.94*** 4.08*** -8.95*** Spain -3.2%*** 0.32 0.65* 0.00 1.89*** 1.47*** -4.54*** Italy 1.6%*** -0.19 -0.22 -0.16 -0.10 -0.05 -0.15 Latvia 15.6%*** 1.38** 0.23 2.35*** -0.28 -0.78 9.20*** Lithuania 9.2%*** 0.98 0.81 1.13* 2.82 0.00 1.81*** Portugal -3.0% -0.52 -0.49 -0.55 -0.02 -0.29 -2.25*** Romania 1.8% -0.63* -0.71* -0.55* -0.80 -0.84 -0.40

Results (4): NOWCASTED LEVELSWhat EU-SILC 2013 will show (2012 income)

Poverty rates (60% of median)

 

Median Income

(€ per year) All Males FemalesChildren

(<18)Prime-

ageElderly

(65+)

Estonia 6,340 18.1 16.8 19.2 17.7 14.3 22.9

Greece 8,683 22.8 22.8 22.8 27.6 22.7 14.7

Spain 12,111 22.1 21.8 22.4 29.1 21.9 16.3

Italy 16,225 19.4 18.1 20.6 26.2 19.1 16.9

Latvia 4,796 20.5 20.2 20.8 24.7 18.5 18.1

Lithuania 4,373 21.0 20.6 21.2 27.1 19.8 13.9

Portugal 8,155 17.5 17.1 17.8 22.4 14.7 17.7

Romania 2,155 21.6 21.2 21.9 32.1 21.0 13.7Notes:Household incomes are equivalized using the modified OECD scale. Median income in Euro per year.Change in 2010-2012 applied on the latest EU-SILC statistics.

Select cases for employment transitions based on estimated conditional probabilities of being in a particular employment status

occupational decision models (e.g. Habib et.al. 2010, Ferreira et.al. 2004) estimation based on the latest EU-SILC microdata Logit model for predicting employment/non-employment estimated

separately for those with low/high education, age frame 15-64: predictors: sex, age, years of education, occupational status as measured by

ISCO, dependency ratio in a household, participation rate, dummies for hh head, employed partner, small children under 4; squares, interactions.

combine with published LFS employment statistics (as currently)

Modeling new wages needs refinement: Currently based on average wage within the stratum Refining this using wage equations

Reweighting for changes in employment and/or demographics

Other?

Further steps (methodological improvements)

Thank you!Your comments are welcomed!

Jekaterina Navicke: jnavicke@essex.ac.uk

Fig 4 EUROMOD 2007-2012 and EU-SILC 2007-2010: Median equivalized household disposable income (EUR per year)

Note: SILC data corresponds to the income reference period.

Fig 5 EUROMOD 2007-2012 and EU-SILC 2007-2010: At risk of poverty rates (using 60% median as the threshold)

Notes:Confidence intervals for EUROMOD estimates are due to a random element in the simulation of employment transitions and do not account for sampling variability. Confidence intervals for EU-SILC estimates of at risk of poverty rates are constructed based on the standard errors provided in Comparative EU Intermediate Quality Reports for EU-SILC 2008-2010 (Available at: http://epp.eurostat.ec.europa.eu/portal/page/portal/income_social_inclusion_living_conditions/quality/eu_quality_reports).

Fig 6 EUROMOD 2007 -2012: At risk of poverty rates by household type (using 60% of the 2007 median as the threshold)

Note: The poverty threshold is 60% of median 2007 equivalised household income, indexed by the HCPI

10

15

20

25

30

35

40

45

2007 2008 2009 2010 2011 2012

Estonia

051015202530354045

2007

Total Children Elderly

15

20

25

30

35

40

45

50

2007 2008 2009 2010 2011 2012

Greece

15

20

25

30

35

40

45

50

2007 2008 2009 2010 2011 2012

Spain

15

20

25

30

35

40

45

50

2007 2008 2009 2010 2011 2012

Italy

15

20

25

30

35

40

45

50

2007 2008 2009 2010 2011 2012

Lithuania

20

25

30

35

40

45

50

55

2007 2008 2009 2010 2011 2012

Latvia

15

20

25

30

35

40

45

50

2007 2008 2009 2010 2011 2012

Portugal

5

10

15

20

25

30

35

40

2007 2008 2009 2010 2011 2012

Romania

Growth incidence curvesChange in real income by percentile, 2010-2012

-15%

-10%

-5%

0%

5%

10%

15%

20%

25%

6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 62 64 66 68 70 72 74 76 78 80 82 84 86 88 90 92 94

LV

EE

LT

RO

IT

ES

PT

-80%

-70%

-60%

-50%

-40%

-30%

-20%

-10%

0%6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72 75 78 81 84 87 90 93

GREECE

Note: based on re-ranked distribution