SOURCE: Malaysian Communications and Multimedia Commission, Malaysia TITLE: Primary, Secondary and Administrative Data in Telecommunications

Transcription

1 INTERNATIONAL TELECOMMUNICATION UNION TELECOMMUNICATION DEVELOPMENT BUREAU Document 017-E 27 February 2009 Original: English 7 TH WORLD TELECOMMUNICATION/ICT INDICATORS MEETING, CAIRO, EGYPT, 3-5 MARCH 2009 SOURCE: Malaysian Communications and Multimedia Commission, Malaysia TITLE: Primary, Secondary and Administrative Data in Telecommunications

2

3 Primary, Secondary and Administrative Data in Telecommunications - Case of Malaysia - Koay Hock Eng Malaysian Communications and Multimedia Commission SKMM Data come in many forms Not only Primary data (including piggy-backs) Secondary data Administrative records But also as Guven (2001), Statistics Division ESCAP adds Monitoring data Computation methods Proxy data 1

4 Primary data Definition Data collected directly from statistical units in pursuance of research objectives Advantages Purpose built with all attendant advantages Disadvantages Costly Takes a while Subject to errors too some of which may not be measurable Piggy-backs Might be cheaper and quicker Reporting burden, lower overall response rate Secondary data Definition Statistics or data sourced from other surveys / censuses which have objectives other than the one at hand Crawford (1997) in Chapter 2 of Marketing Research and Information Systems published by FAO Regional Office for Africa, made a strong case for studying secondary data before engaging in primary research. Advantages May be available which is entirely appropriate and wholly adequate Not as costly Quicker 2

5 Secondary data Disadvantages Dissimilarities in definition Dissimilarities in reference period Dissimilarities in target population Measurement error Source bias Reliability Time scale Administrative records Definition Data routinely gathered for a particular administrative goal such as the registering of subscribers by service providers Marquis, Palacios & Wetrogan (1996) was of the opinion that in future government statistical systems, administrative records will be used widely. Demand for information will increase beyond the ability of surveys and censuses to meet it.. 3

6 Administrative records Advantages May be more accurate May be put to good use in exploratory phase Help in defining the population Not as costly Quick Reduce response burden on citizens and organisations More precise small area statistics Potentially more info is available unconstrained by questionnaire length Rates of missing data can be lower, fewer non-response adjustments, fewer imputations and potentially less nonobservation bias No interviewer effects Avoid human cognition effects such as forgetting Administrative records Disadvantages Generally compiled by people with no primary interest in statistics thus quality can be suspect e.g. Customs and Excise Difficult or impossible to get Costs may be surprisingly high: to locate, purchase, match, review, fix errors and produce estimates Coverage may not be complete Seldom perfect match of concepts Currency of info Matching problems in relation to estimate relationship between variables Privacy, confidentiality and disclosure issues 4

7 Administrative records Marquis et al tell us administrative records databases used successfully in Canada, (Standish and Ravindra, 1998) Finland, (Myrskyla 1991, and Lankinen, 1995) The case of Malaysia SKMM is the Malaysian Communications and Multimedia Commission. Statistical function in SKMM started since establishment with collection and use of administrative records of licensees. Today the SKMM embraces all 3 and more in addressing statistical needs. 5

8 Addressing statistical needs in the SKMM External users Internal users 2 o data? Administrative data? 1 o data? Purchase Market Research? Collection of primary data Began in earnest in 2004 with the establishment of SKMM CATI Centre The surveys done at SKMM CATI Centre includes: Hand Phone Users Survey (Annual since 2004) Household Use of the Internet Survey (2005, 2006, 2008) Mobile Broadband Survey (ad-hoc 2009) These 1 o surveys collect data on usage (such as SMSes, Internet banking etc) not afforded by administrative records. Are regulators well placed to do primary surveys? 6

9 A peek at SKMM CATI Centre A peek at SKMM CATI Centre 7

10 Collection of primary data Web-based primary data collection from 2005 Household Use of Internet Survey 2005 Annual Licensees Survey Caution against general uncontrolled deployment on the web that allows selfselection and abuse Examples of use of secondary data Censuses and household surveys courtesy of NSO Malaysia. Examined for similarity of concept, definition and reference period E.g. population for a given year Used as-is Percentage of household in possession of computers. Percentage of household in possession of Internet connections. Percentage of household in possession of TVs. Percentage of household in possession of radios. Percentage of household in possession of DELs. Long gestation period Regularity needs to be improved upon. 8

11 Use of administrative records Among the administrative records collected from licensees include: Subscriptions Exchanges International Gateways Points of interconnect (POI) Traffic (minutes and revenue) All services DELs Mobile cellular Broadband Record Keeping Rules Record Keeping Rules, Communications and Multimedia Act 1998;Section 268 The minister may make rules, to be published in the Gazette, to provide for record-keeping and to require one or more licensees or persons to keep and retain records. This requirement may alter the data fields recorded as administrative procedure or furnish the data in a primary data collection. Provision for records to be audited Enhances the scope, quality and uniformity of administrative data across the board 9

12 Synergy of primary, secondary and administrative records At present, primary data, secondary data and administrative records are collected and used synergistically. In planning primary surveys Licensees report from their administrative records, the number of subscriptions by network. RDD samples drawn with probability proportional to size of network. Estimates of number of users made. Estimates related to secondary demographic data from NSO to give various penetration rates Synergy of primary, secondary and administrative records Points Obviates the necessity for a frame for infinite population, single domain study. RSEs can even be calculated. Effective in a population that is almost saturated e.g. mobile cellular in Malaysia. Even the most basic data in administrative records would have an NRID number (which in turn has the birth date, and hence age, and gender encoded), address which then can be cross-checked with NSO database for urban-rural location. Reduce respondent burden and errors due to ignorance. 10

13 Example of effective use of administrative data Identification of underserved areas Licensees report from their administrative records, the number of subscriptions in a locality. Population of locality is obtained from the NSO. Penetration rate in locality is calculated. Localities classified as underserved if penetration rates fall below a benchmark. Example of effective use of administrative data Points The other alternative would be to do a census or sample every locality as the penetration rate of each must be enumerated. This would be too costly and would take too long. Required regularity of measurements for tracking progress of rollout favour use of administrative records over primary surveys. The accuracy with which the data is needed is not that exacting. If the benchmark is 14 % penetration, then whether a locality has 8.6 or 8.1 penetration, it is still underserved. 11

14 Conclusion Any discussion on the pros and cons of using 1 o, 2 o and administrative records will be inconclusive. 2 o data and administrative records can be very useful under the right circumstances depending on: Purpose Use Accuracy requirements To disregard both in favour of 1 o data might result in reinvention of the wheel. To dismiss administrative records totally would be akin to throwing the baby out with the bathwater. Conclusion 1 o data in the form of surveys are potentially more useful in measuring usage (such as actual usage of Internet, characteristics of Internet users, activities they do while online, etc) Administrative data is useful but can't measure actual usage (for example: mobile phone subscribers can be double counted because of SIMs and inactive prepaid SIMs, etc) However sample surveys are subject to sampling errors which may be assessed and quantified. More insidious is bias, coverage errors, non-response errors, cheating and deliberate lying, the extent of which may not be fully known. 12

15 References Marketing research and information systems Data Types and Sources of Environment Statistics TOWARDS A U.S. POPULATION DATABASE FROM ADMINISTRATIVE RECORDS Thank You 13