SEMINAR AT HIVA RESEARCH INSTITUTE FOR WORK AND SOCIETY Friday, 18 - - PowerPoint PPT Presentation

This project has received funding from the European Union’s Seventh Programme for Research, Technological Development and Demonstration under Grant Agreement No 312691

www.inclusivegrowth.be

SEMINAR AT HIVA

■ RESEARCH INSTITUTE FOR WORK AND SOCIETY

The use of f auxiliary and event data in in tr tracking an in inhomogeneity of f substantive data in in lo longitudinal stu tudies.

The e cases es of

• f ESS Rou
• und 7th

th and EWCS CS Wave e 5th th

Teresa Żmijewska-Jędrzejczyk Ins nsti titute of

• f Phil

hilosophy and and So Soci ciology at t the the Poli

• lish Academy of
• f Sci

Sciences PhD hD Stu Student: : Uni niversi sity ty of

• f War

arsa saw

Ques uestio ions: ter eresa.zmij ijewska@if ifispan.waw.pl

Friday, 18 November 2016

2

Agenda

• Homogeneity of a data in longitudinal crossnational studies –

why important?

• The necessity to track transitory factors in trend analysis
• Why EWCS and ESS data?
• EWCS and ESS data – results of the study
• Summary and problems to be solved
• Disscussion

3

Why im important?

• Standardization as a key premise of survey
• The challenges of measurement validity in

broad sense: non-response bias, measurement equivalence

»Ref.: Jowell 2004, Billiet 2015 »Ref.: Commission de sondages

• Statistical analysis of hierarchical models

4

Surv rveys are prone to the dif ifferent effects

Among long list of the different surveys’ effects, three types are rele levant to

• th

the le length of

• f th

the su survey peri riod

• Con
• ntext Effects
• Surv

Survey clim climate (R (Ref.: Loo Loosveldt & & Jo Joye 2016 2016)

• Su

Sudden an and lon

• ng-lasting events’ effects
• In

Interv rviewer Effects

• Fati

tigue Effects (in interv rviewer lose

• ses foc
• cus,

, be become tir ired)

• Lea

Learnin ing Effects (in interv rvie iewer kn knows by y hea heart a a qu questi tionnaire)

• Effects con
• nnted wi

with th the (in (in)stabili ility of

• f th

the attitudes

• Mod
• de effects
• Man

any ot

• thers (R

(Ref.: Weis isberg H. . F. . 2005 2005)

5

Pattern rns of f the an in inhomogeneit ity of f su substantive data - example les

6

The necessity to track transitory ry factors

• In

In tr trend an analysis is fr from lon longitudin inal stu tudies th the pos

• stulates
• f
• f hom
• mogeneity of
• f th

the data

• As a result of the different conditions under which cross-country

and longitudinal studies are carried out, it is necessary to collect contextual data about the economic situation, the level of democratization, political regime, political events like the imminence of general elections

• Th

That im impose als also th the necessity to

• control

l th the im impact of

• f tr

tran ansitory ry factors - events

• Su

Such factors mig ight bias iased th the lon long-term tr trends in in attitudes!

7

Rationale for selecting data

Cross-country with long tradition Longitudinal and long fieldwork period High quality data and different topics

8

Procedure

1 step

• All

ll varia iable les

• All

ll countries

2 step

• Se

Sele lect th the countries for fu further analy lysis

• Varia

iable le prone to be an in inhomogeneous

3 step

• Marked variables and countries
• Additional controlled variables

European Work rkin ing Co Condit itio ions Su Survey 5th th wave (2 (2010)

Representative sample le based on the whole country pop

• pulation 15+ or 16+ (Spain, UK, Norway)

Co Coverage: EU27, Norway, Croatia, FYROM, Turkey, Albania, Montenegro and Kosovo Main in subject: provide an overview of the every eryday rea ealit ity

• f
• f men

en and women at t wor

• rk

Frequency: since 1990 every ery 5 yea ears

Aim im: deriv ive those varia iables and those countrie ies prone to be an in inhomogeneous (red fla lags)

1st step All ll varia iable les and all ll countrie ies

12

Method and assumptions

• DV: lis

list of

• f 39

390 0 var aria iable les - all for which the scale level was know were used (ISCO occupation classification excluded)

• IV

IV: date of

• f th

the fie fieldwork (month)

• No weights applied
• =.001
• The estimation method (and test stat) for the variables
• Metrics: OLS (F-test),
• Ordinal: ordered probit (chi2)
• Nominal: multinomial logit regression (chi2)
• No control var

aria iable les

Sem Vandekerckhove

13

Results: : red fla flags

Country red flags Turkey N=2100 64 France N=3037 50 Albania N=1000 48 Belgium N=3995 40 Portugal N=1000 30 Luxembourg N=999 27 Latvia N=1001 22 Croatia N=1100 21 Germany N=2133 17 Slovenia N=1404 17 Greece N=1037 15 Country red flags UK N=1575 14 FYROM N=1100 14 Malta N=1000 13 Ireland N=1003 12 Norway N=1085 11 Montenegro N=1041 11 Kosovo N=1018 10 Bulgaria N=1014 8 Estonia N=1000 8 Netherlands N=1017 8 Austria N=1003 7 Poland N=1497 7 Country red flags Sweden N=1004 5 Spain N=1007 4 Italy N=1500 4 Hungary N=1006 4 Romania N=1017 4 Czech Rep. N=1000 3 Cyprus N=1000 3 Lithuania N=1004 3 Denmark N=1069 1 Slovakia N=1002 1 Finland N=1028 1

14

Results: : red fla flags by scale and topic

Country Total Nominal Ordinal Metrics Topics

Turkey N=2100 64 31 30 3

All

France N=3037 50 29 18 3

All

Albania N=1000 48 21 24 3

All

Belgium N=3995 40 23 14 3

All

Portugal N=1000 30 11 14 5

All

• Different sample size: Belgium, France vs. Albania, Portugal
• Differnt suvey climate across Europe
• Variable prone to be an inhomogeneous are from different topics

European So Socia ial l Su Survey 7th th round of f ESS SS (2015/2016) an academic icall lly driv riven

Sample based on the whole country pop

• pula

lation 15+ + Co Coverage: crosseuropean – in 2015/16: Austria, Belgium, Czech Republic, Denmark, Estonia, Finland, France, Germany, Hungary, Ireland, Israel, Latvia, Lithuania, Netherlands, Norway, Poland, Portugal, Slovenia, Spain, Sweden, Switzerland, United Kingdom Main in subject: since 2001 every ry 2 years measures the attitudes, beliefs and behavior patterns of diverse populations across Europe

16

Method and assumptions

• DV:

: Sub ubstantiv ive que questio ions selected from the core questionnaire, e.g.: trust in public institutions, assessment of the public institutions (incl. the European Parliament and the attitude to European integration), political activity, subcetive well-being, happiness

• IV

IV: da date of

• f the

the fi field ldwork rk (days)

• No weights applied
• The estimation method (and test stat) for the variables

– 11 point numeric scales: linear regression – ordinal variables: ordinal logistic regression (PLUM) – dichotomous variables: binary logistic regression

• No
• con
• ntrol vari

ariable les

17

Results for France

***p=<0,001 **p<0,05 *p<0,1

b Trust: parliment

0,013***

Trust: legal system

0,007**

Trust: police

0,011***

Trust: politicians

0,009***

Trust: political parties

0,010***

Trust in the European Parliament

0,010***

How satisfied with the national government

0,013***

How satisfied with the way democracy works in country

0,016***

State of health services in country nowadays

0,005*

18

Summary

• Varia

iable le prone to

• be an

an in inhomogeneous ar are fr from dif ifferent top

• pics an

and th they have more country sp specific patterns

• Th

Thus fie fieldwork peri riod is is an im important element in the proces of assessment the data homogeneity, both in terms of measurement equivalence and statistical analysis (e.g.: hierarchical models) in cross-country comparisons

• Tracing the homogeneity of the data is still a challenge for
• researchers. Why? In cross-countries data sets it may be

dif ifficult to

• for
• rmulate even prelimin

inary ry hyp ypotheses what kin kind of

• f

tr transit itory factors, effects or

• r events ar

are rele levant an and what var ariable les mig ight be affected

The im impact of f tr transitory and non-transitory ry factors

Where’s varia iable le quali lity hid idin ing?

20

Other effects – example fr from the ESS

De Demographics

• Ge

Gender

• Age

ge

• Education
• Ru

Rural vs.

• s. urb

rban ar area Event data to

• control th

the im impact of

• f tr

tran ansitory ry factors

21

Charlie Hebdo terrorist attack

Fr France, 7.

7.-9. . 01.2015 r.

Char arlie ie Hebdo sa satir irical l weekly magazin ine.

The mag agazin ine has has be been the target of

• f two te

terroris ist attacks, s, in 201 2011 an and 201 2015. Bot Both wer ere pr presumed to to be be in n res esponse se to to a a nu number of

• f controversia

ial Muhammad car artoons it t pu

• published. In

In the sec second of

• f these

se attacks, , 12 12 peo people wer ere kill killed, , including pu publis ishing di director

• r Ch

Charb an and se several ot

• ther pr

prominent car artoonists.

22

• London
• Paris

Charlie Hebdo solid lidarity acts

The Je suis Charlie ("I am Charlie") slogan became an endorsement of freedom of speech and press.

23

Charlie Hebdo - > > in infl fluence on the fi field ldwork

The front cover of edition of 14 January 2015, with a cartoon in the same style as 3 November 2011 cover, uses the phrase "Je Suis Charlie". (Headline translation: "All is forgiven.")

ESS 7 in France

• Fieldwork: 31.
• 31. 10.
• 10. 14

14 – 03.

• 03. 03.
• 03. 20

2015 15

• Sample: N=1 917
• RR=50.9%

24

Method and assumptions

• DV: sa

same su substantive questions

• IV

IV: gender, ag age, education (5 5 category), plac lace of

• f resid

idence: : 1 1 - big ig cit city or

• r its

its su suburbs (26 (26.4%) roots ou

• utsid

ide Fr France: 1 1 - resp spondent, mother/father bor

• rn ou
• utsid

ide (24 24.6%) %) in intervie iew aft fter th the attack on

• n Char

arli lie Hebdo: 1 1 - yes: N = = 27 275 5 in interviews started on

• n Jan

January 7 7 by y 14 14 or

• r la

later (14 14.3%) %) 0 - no:

• : N =

= 16 1642 42

• De

Design weig ight ap applie lied

25

Results for France

***p=<0,001 **p<0,05 *p<0,1

b

Trust: parliment 0,586*** Trust: police 0,268 Trust: politicians 0,475*** Trust: political parties 0,563*** Trust in the European Parliament 0,396** European unification go further or gone too far 0,007* How satisfied with the national government 0,801*** How satisfied with the way democracy works in country 0,913*** State of health services in country nowadays 0,176

26

Comparison: : results wit ith and wit ithout CV CV

***p=<0,001 **p<0,05 *p<0,1

b b without CV

Trust: parliment 0,586*** 0,013*** Trust: police 0,268 0,009** Trust: politicians 0,475*** 0,011*** Trust: legal system

• 0,007***

Trust: political parties 0,563*** 0,010*** Trust in the European Parliament 0,396** 0,010*** European unification go further or gone too far 0,007* 0,007* How satisfied with the national government 0,801*** 0,013*** How satisfied with the way democracy works in country 0,913*** 0,016*** State of health services in country nowadays 0,176 0,005*

27

Summary

• In many cases differences are significant
• Analyses indicate that the event not related

thematically, influence the process of formulating

• pinion on various topics
• Again, tracking the effects of contextual current

events is an important element of the assessment of their value in terms of homogeneity

DV: : MIP IP QUESTION for r EWCS EVENTS/C /CLAIMS for r ESS CV: : WEEK OF THE FIE IELDWORK and oth ther control variables for r a gr group of f countr tries

Furt rther r analy lysis is

This project has received funding from the European Union’s Seventh Programme for Research, Technological Development and Demonstration under Grant Agreement No 312691

www.inclusivegrowth.be

SEMINAR AT HIVA

■ RESEARCH INSTITUTE FOR WORK AND SOCIETY

Dank je!