15
Improving of Household Sample Surveys Data Quality on Base of Statistical Matching Approaches Ganna Tereshchenko Institute for Demography and Social Research, Kyiv, Ukraine The European Conference on Quality in Official Statistics Rome, 8-11 July 2008

Improving of Household Sample Surveys Data Quality on Base of Statistical Matching Approaches

  • Upload
    pegeen

  • View
    25

  • Download
    0

Embed Size (px)

DESCRIPTION

Improving of Household Sample Surveys Data Quality on Base of Statistical Matching Approaches. Ganna Tereshchenko Institute for Demography and Social Research, Kyiv, Ukraine. The European Conference on Quality in Official Statistics Rome, 8-11 July 2008. - PowerPoint PPT Presentation

Citation preview

Page 1: Improving of Household Sample Surveys Data Quality on Base  of Statistical Matching Approaches

Improving of Household Sample Surveys Data Quality on Base of Statistical Matching Approaches

Ganna TereshchenkoInstitute for Demography and Social Research,Kyiv, Ukraine

The European Conference on Quality in Official Statistics

Rome, 8-11 July 2008

Page 2: Improving of Household Sample Surveys Data Quality on Base  of Statistical Matching Approaches

Measurement of Employment and Unemployment

Main source is The State Sample Survey of Economic Activity of Population (LFS) :

is conducted by State Statistics Committee of Ukraine by ILO methodology, according to international standards population in the age of 15–70 years is surveyed is conducted since 1995: in 1995–1998 once a year, in 1999–2003 – quarterly, since 2004 – monthly LFS sample cover all regions of Ukraine by type of

settlements: urban area (cities, towns) and rural area size of monthly LFS sample is 32,5 thousands of surveyed

households

Page 3: Improving of Household Sample Surveys Data Quality on Base  of Statistical Matching Approaches

Reliability of unemployment rate annual estimates

0

10

20

30

40

50

60

Region of Ukraine

Co

eff

icie

nt

of

vari

atio

n,% LFS,2003 LFS rural area,2003

Page 4: Improving of Household Sample Surveys Data Quality on Base  of Statistical Matching Approaches

Improvement of reliability of LFS indicator estimates for rural area based on statistical matching approach

Using of two probability stratified two stage samples: sample of LFS and sample of household agricultural activity survey (AAS)

Sample design in AAS is differ from LFS: In AAS households are selected in the second stage with probability

proportionally to their area of agricultural allotment, in LFS – on base of the procedure of systematic selection

The size of monthly LFS sample in the rural area makes approximately 3,6 thousand households

The size of AAS sample of households which have to be interviewed under LFS questionnaire is 7,4 thousand households

Total size of monthly sample for interview under LFS questionnaire in the rural area due to AAS has increased three times and is equal to11,1 thousand households

Page 5: Improving of Household Sample Surveys Data Quality on Base  of Statistical Matching Approaches

Rates of employment and unemployment by regions of Ukraine, February, 2007

0

2

4

6

8Crimea

VinnytsiaVolyn

Dnipropetrovsk

Donetsk

Zhytomyr

Zakarpattya

Zaporizhya

Ivano-Frankivsk

Kyiv

Kirovograd

LuganskLvivMykolaiv

Odesa

Poltava

Rivne

Sumy

Ternpoil

Kharkiv

Kherson

Khmelnytsky

Cherkasy

ChernivtsiChernigiv

Rates of unemployment AAS Rates of unemployment LFS

0

20

40

60

80

100Crimea

Vinnytsia

Volyn

Dnipropetrovsk

Donetsk

Zhytomyr

Zakarpattya

Zaporizhya

Ivano-Frankivsk

Kyiv

Kirovograd

LuganskLvivMykolaiv

Odesa

Poltava

Rivne

Sumy

Ternpoil

Kharkiv

Kherson

Khmelnytsky

Cherkasy

Chernivtsi

Chernigiv

Rates of employment AAS Rates of employment LFS

Page 6: Improving of Household Sample Surveys Data Quality on Base  of Statistical Matching Approaches

Composite estimation

)AAS(unun

)LFS(ununun

)AAS(emem

)LFS(ememem

Y)ˆ1(YˆY

Y)ˆ1(YˆY

Page 7: Improving of Household Sample Surveys Data Quality on Base  of Statistical Matching Approaches

Calculation of optimal weights coefficients and

where – standard error of estimate of employed population number on LFS sample;

– standard error of estimate of employed population number on AAS sample;

– standard error of estimate of unemployed population number on LFS sample;

– standard error of estimate of unemployed population number on AAS sample,

is the bias of estimate of number of employed population by data of AAS, calculated as average of biases for current and the two previous months,

is the bias of estimate of number of unemployed population by data of AAS, calculated as average of biases for current and the two previous months.

em

,)ˆ()ˆ()ˆ(

)ˆ()ˆ(ˆ

; )ˆ()ˆ()ˆ(

)ˆ()ˆ(ˆ

)(2)(2)(2

)(2)(2

)(2)(2)(2

)(2)(2

personsunemployedforYBYSEYSE

YBYSE

personsemployedforYBYSEYSE

YBYSE

AASun

AASun

LFSun

AASun

AASun

un

AASem

AASem

LFSem

AASem

AASem

em

)ˆ( )(AASemYSE

)ˆ( )(LFSemYSE )(ˆ LFS

emY)(ˆ AAS

emY)(ˆ LFS

unY)(ˆ AAS

unY)ˆ( )(LFS

unYSE

)ˆ( )( AASunYSE

)ˆ( )( AASemYB

)ˆ( )( AASunYB

un

Page 8: Improving of Household Sample Surveys Data Quality on Base  of Statistical Matching Approaches

Calculation of coefficients for adjustment of the resulted employed and unemployed persons weights in rural area

On the first stage value of is calculated for employed and unemployed persons in rural area by the formula:

The corrected statistical weights of employed and unemployed persons in rural area of each region are calculated by the formula:

.)ˆ1(

;)ˆ1(

sampleAASonpersonunemployedfor

sampleAASonpersonemployedfor

sampleLFSonpersonunemployedfor

sampleLFSonpersonemployedfor

k

un

em

un

em

i

,iii kww

ik

Page 9: Improving of Household Sample Surveys Data Quality on Base  of Statistical Matching Approaches

Calculation of coefficients for adjustment of the resulted economically inactive persons weights in rural area

On the second stage value of is calculated for economically inactive persons in rural area for each region by the formula:

where – total number of able-bodied population in rural area of region, calculated on external data;

– estimate of employed population number on LFS sample in view of corrected statistical weights ;

– estimate of employed population number on AAS sample in view of corrected statistical weights ;

– estimate of unemployed population number on LFS sample in view of corrected statistical weights ;

– estimate of employed population number on AAS sample in view of corrected statistical weights ;

– estimate of economically inactive population number on LFSP sample in view of corrected statistical weights ;

– estimate of economically inactive population number on AAS sample in view of corrected statistical weights

)()(

)(')(')(')('7015

ˆˆ)ˆˆˆˆ(

AASei

LFSei

AASun

LFSun

AASem

LFSem

iYY

YYYYNk

7015N)('ˆ LFS

emYiw

)('ˆ AASemY

iw

iw

iw

iw

iw

)('ˆ LFSunY

)('ˆ AASunY

)('ˆ LFSeiY

)('ˆ AASeiY

Page 10: Improving of Household Sample Surveys Data Quality on Base  of Statistical Matching Approaches

Reliability of employment rate monthly estimates in rural area before and after statistical matching of the LFS data, February, 2007

0

2

4

6

8

10

12

14

Region of Ukraine

Coeff

icie

nt

of

variation,%

LFS LFS&ASS

Page 11: Improving of Household Sample Surveys Data Quality on Base  of Statistical Matching Approaches

Reliability of unemployment rate monthly estimates in rural area before and after statistical matching of the LFS data, February, 2007

0

20

40

60

80

100

Region of Ukraine

Coe

ffic

ient

of

varia

tion,

% LFS LFS&ASS

Page 12: Improving of Household Sample Surveys Data Quality on Base  of Statistical Matching Approaches

Potential problem with comparability of unemployment rate estimates by regions

.onlydataLFSbyreceivedareіregionforestimatesif,1

;datamatchedASS&LFSbyreceivedare

іregionforestimatesif,0

Di

Share of incomparable estimates

where – number of regionsN

24,025

61 N

Di

R

N

ic

Region Type of data

Crimea 1 LFSVinnytsia 0 LFS&ASSVolyn 0 LFS&ASSDnipropetrovsk 0 LFS&ASSDonetsk 0 LFS&ASSZhytomyr 0 LFS&ASSZakarpattya 0 LFS&ASSZaporizhya 0 LFS&ASSIvano-Frankivsk 1 LFSKyiv 0 LFS&ASSKirovograd 0 LFS&ASSLugansk 1 LFSLviv 0 LFS&ASSMykolaiv 0 LFS&ASSOdesa 0 LFS&ASSPoltava 0 LFS&ASSRivne 0 LFS&ASSSumy 0 LFS&ASSTernpoil 0 LFS&ASSKharkiv 0 LFS&ASSKherson 0 LFS&ASSKhmelnytsky 1 LFSCherkasy 1 LFSChernivtsi 1 LFSChernigiv 0 LFS&ASS

iD

Page 13: Improving of Household Sample Surveys Data Quality on Base  of Statistical Matching Approaches

Relative efficiency of matching procedure by regions, February, 2007

0,0

0,2

0,4

0,6

0,8

1,0

Crim

ea

Vin

nyt

sia

Vo

lyn

Dni

pro

pe

tro

vsk

Don

ets

kZ

hyt

om

yrZ

aka

rpa

ttya

Za

por

izh

yaIv

an

o-F

ran

kivs

k

Kyi

vK

irov

og

rad

Lug

ans

k

Lvi

vM

yko

laiv

Od

esa

Po

ltava

Riv

ne

Su

my

Te

rnpo

ilK

har

kiv

Kh

erso

nK

hm

eln

ytsk

yC

herk

asy

Che

rniv

tsi

Che

rnig

iv

Region of Ukraine

reff

for employment rate for unemployment rate

41,0)ˆ(

)ˆ()(

)&(

LFSem

AASLFSem

emYV

YVreff 48,0

)ˆ(

)ˆ()(

)&(

LFSun

AASLFSun

unYV

YVreff

Page 14: Improving of Household Sample Surveys Data Quality on Base  of Statistical Matching Approaches

Conclusions

Statistical matching of the labour force survey data, received on samples with different design has allowed improving the reliability level of employment and unemployment indicators estimation in rural area.

At the same time there is a potential problem with providing of data comparability

It is necessary to take into account that the volume of the information for processing grows and estimation procedures are complicated

Page 15: Improving of Household Sample Surveys Data Quality on Base  of Statistical Matching Approaches

Thank you for attention!

Ganna Tereshchenko

Institute for Demography and Social Research

of National Academy of Sciences of Ukraine

Kyiv, Ukraine

[email protected]