This document discusses telephone sampling issues and advancements in the Middle East and North Africa region. It outlines the increased importance of telephone surveys using CATI methodology due to high mobile phone penetration. The total survey error framework is applied, highlighting coverage errors from lack of landlines and issues with open numbering plans. Methods for sampling include using auxiliary lists and screening samples to check connectivity for improved representation. Country case studies show the suitability of dual frame sampling but also issues reaching populations in unstable areas.
Regression analysis: Simple Linear Regression Multiple Linear Regression
Telephone Sampling in the Middle East: Issues and Advancements
1. Telephone Sampling in the MENA:
Issues and Advancements
Carsten Broich
3MC Conference 2016
2. Why this research?
7/27/2016
3MC Conference 2016 – Telephone Sampling in the MENA: Issues and Advancements
2
• Importance of Sampling in CATI surveys
• Increase knowledge of sampling options in MENA region
• Outline advantages and issues (Change F2F Phone)
• Collect past experience from sampling projects
• Create a follow-up agenda with issues and possible solutions
3. Scope of the research: MENA & CATI
7/27/2016
3MC Conference 2016 – Telephone Sampling in the MENA: Issues and Advancements
3
Data Collection Modes:
Face-to-Face Surveys
Mail Surveys
Online Surveys
Telephone Surveys
SMS Surveys
Mixed Mode etc.
4. Why CATI surveys in the MENA region?
7/27/2016
3MC Conference 2016 – Telephone Sampling in the MENA: Issues and Advancements
4
Country Mobile
Subscriptions [%]
Landline
Subscriptions [%]
Egypt 114.31 7.57
Iran 87.79 39.06
Kuwait 218.43 14.20
UAE 178.06 22.26
Libya 161.12 11.30
Oman 157.75 9.56
Qatar 145.76 18.41
Saudi Arabia 179.56 12.33
Syria 63.86 16.57
Tunisia 128.49 8.54
USA 110.20 39.83
UK 123.58 52.35
Germany 120.42 56.89
5. Methodology: Total Survey Error
Coverage Error
Numbering Plan
7/27/2016
3MC Conference 2016 – Telephone Sampling in the MENA: Issues and Advancements
5
Target Population
Sampling Frame
Designated
Sample
Final Sample
Coverage Error
Sampling Error
Nonresponse
Error
Sample Creation
Screening of Sample
Fieldwork Results
Errors of Representation
6. Coverage
7/27/2016
3MC Conference 2016 – Telephone Sampling in the MENA: Issues and Advancements
6
• Target Population:
National Representative / Area Sampling
• Coverage Error: No Mobile / Landline
1. Bureau of Statistics
2. Face-to-Face Survey
3. Estimation
7. Coverage
7/27/2016
3MC Conference 2016 – Telephone Sampling in the MENA: Issues and Advancements
7
• Methods for estimating landline ownership
• Landline / Mobile Penetration statistics from ITU
• Example KSA:
• 4.1M LL subscriptions - Average HH size: 6.4
• Assuming 10% business subscriptions
Estimate 23.7 M people 76.5% (79% in fact)
8. Numbering Plan
7/27/2016
3MC Conference 2016 – Telephone Sampling in the MENA: Issues and Advancements
8
• Open / Closed Numbering Plan
• Open Numbering Plan: SN Varying Length
• Documentation differs per country
CC
NDC
SN
Name of Region Area Code Numbering Range
East Cairo (Region 1) 0 2
23100200-23100399
23130000-23131999
23132000-23137499
24660000-24664159
24690000-24690511
West Cairo (Region 3) 0 2 25466000-25466649
9. Numbering Plan: List-assisted
7/27/2016
3MC Conference 2016 – Telephone Sampling in the MENA: Issues and Advancements
9
• Open numbering plan / billions of possible combinations
• Creation of auxilary sampling frame:
A. Seed numbers (landline & mobile)
B. Provider lookup (mobile only) List-Assisted
Database
Web Directories
Social Media
Web Mining
Past-Surveys
10. Numbering Plan: List-assisted
7/27/2016
3MC Conference 2016 – Telephone Sampling in the MENA: Issues and Advancements
10
0
5
10
15
10 30 50 70 90
Block Density Analysis
Al Madar Libyana
• Open numbering plan / billions
of possible combinations
• Creation of auxiliary sampling
frame:
A. Seed numbers (landline &
mobile)
B. Provider lookup (mobile only)
Seed numbers: Social Media,
directories, web mining
11. Sample Creation
7/27/2016
3MC Conference 2016 – Telephone Sampling in the MENA: Issues and Advancements
11
• Simple Random Sampling
• Stratified Random Sampling
• Geocoding
• Case Syria:
• Stratified Random Sample
• Geocoding City / Fraction
12. Screening of Sample
7/27/2016
3MC Conference 2016 – Telephone Sampling in the MENA: Issues and Advancements
12
Mobile
Provider Lookup
Landline
Dialer
• Mobile Screening very
effective: check for active SIM
cards
• Landline numbers have issues
with connectivity – possibility
to exclude numbers that are
working
Include batch of non-working
sample batches
• People with multiple phones
Sampling Error
13. Fieldwork Results: General SAPS
7/27/2016
3MC Conference 2016 – Telephone Sampling in the MENA: Issues and Advancements
13
Ratios Mobile vs. Landline
Response rates
~25% of sample unclear Disposition
Influence on RR.
Structured Sampling Frames
Mobile screening accurate
Issues remain with LL
Country Oman Bahrain Qatar Kuwait
Completes N=1008 N=1005 N=1002 N=999
Net Effective Incidence 10,8% 6,1% 6,4% 5,3%
Response Rate 3,8% 5,3% 4,9% 5,0%
Cooperation Rate 7,0% 7,5% 7,5% 6,4%
Refusal Rate 50,9% 65,2% 60,8% 72,5%
Contact Rate 56,7% 72,3% 67,9% 80,3%
Country Mobile Landline
Bahrain 70% (n=700) 30% (n=300)
Kuwait 60% (n=600) 40% (n=400)
Oman 50% (n=500) 50% (n=500)
Qatar 50% (n=500) 50% (n=500)
14. Fieldwork Results: Countries in turmoil
7/27/2016
3MC Conference 2016 – Telephone Sampling in the MENA: Issues and Advancements
14
Province Total Contact % Sampling %
Aleppo 11 0,9 23,8
Alhasakah 153 12,4 7,5
Al-Quneitra 8 0,6 0,4
Al-rakkah 4 0,3 4,7
Al-Swedaa 78 6,3 1,8
Damascus 357 28,5 13,4
Damascus city 42 3,4 9,1
Dara 63 5,1 5,0
Edleb 6 0,5 7,4
Hamah 122 9,8 8,2
Homs 2 0,2 9,0
Latakia 212 17,1 5,2
Quneitra 4 0,3 0,4
Tartous 191 15,4 4,1
Grand Total 1253 100,0 100
• RDD Dual Frame suitable?
• Mobile provided good coverage
• Landline: Affected city issues
40% of respondents have LL
Landline Mobile
Disconnected 9491 96,2% 18023 62,1%
Unknown 327 3,3% 608 2,1%
Working 43 0,4% 10399 35,8%
9861 29030
Sample Results: Libya
Sample Results: Syria
15. Looking ahead and Takeaways
7/27/2016
3MC Conference 2016 – Telephone Sampling in the MENA: Issues and Advancements
15
• Sampling Frames not an issue but numbering screening /
disposition codes remain challenging. Predictive Dialer
• Information on numbering plan varies: Assigned numbers,
screening options, geocoding possibilities
• Increased importance of mobile phones
• Separate Fieldwork and Sampling
• Request Documentation (prevent clustering, public listings etc)
• Central Database: Information on Sampling
Editor's Notes
Thank you
Interesting previous presentation
Glad to look on an international level
15 minutes + 5 minutes for questions
Almost decade worked on telephone sampling frame
Importance of SAMPLING: Documentation
There are sampling frames for mobile and landline
No sampling frames exist: Data Collection Challenges I
GeographicBoundaries
CATI surveys -> Dual Frame RDD Sample
-> possible for SMS surveys
Explain phone penetration developments:
-> Increase of cell phone penetration
-> Stagnation of landline phone
-> Household size: Qatar 8.7, Germany 2.1 Iran: 4.6
Situation in the Middle East : Turmoil, War, Autoritarian Regimes
Online Surveys: not suffcient coverage – facebook and google are working on that
Face-to-Face surveys: Dangerous for interviewers / difficult to ask political questions
-> Situation in Libya: one out of 2 mobile phone provider was taken over by rebels -> large coverage error
-> Important to connect mobile / landline to understand population that is excluded -> what are the characteristics
CC: Country Code
NDC: National dialing code
SN: Subscriber Number
Geocoding: Extra variables: rural / urban / settlement size / governing region / language /
Libya Single Frame Cell phone suitable
Large coverage for Mobiles >160%
Low amounts of contacts: response / working number rate
40% of Cell Phone respondent stated that they have a landline number – clearly we missed it