The purpose of this paper is to provide the workshop participants with an overview of the concepts of Computer Assisted Personal Interviewing (CAPI) and to provide sufficient detail and references to allow participants to gather information about the advantages and disadvantages of the technology, taking into account the technical infrastructure in their own countries and their survey requirements.  The paper provides detailed background information as a resource in addition to the workshop presentation on CAPI.

• an introduction to Computer Assisted Personal Interviewing (CAPI)
• the benefits of CAPI
• issues to consider
• broad architecture of a CAPI system
• some feedback from the ABS experience with CAPI
• summary

Computer Assisted Interviewing (CAI) involves the use of a computer to collect, store, manipulate and transmit data relating to interviews conducted between the interviewer and respondent(s).  Computer Assisted Personal Interviewing (CAPI) is one component of CAI.    It involves the conduct of face to face (household or business) interviews between the interviewer and respondent.  Other components include Computer Assisted Telephone Interviewing (CATI) and Computer Assisted Self Interviewing (CASI).

Internationally, significant developments have occurred over the last two decades on CASIC ( Computer Assisted Survey Information Collection ) methodologies, beginning with computer assisted telephone interview, computer assisted personal interviewing, touchtone data entry, voice recognition, FAX transmission and electronic data interchange.  With the advent of the Internet, some statistical offices (eg US Bureau of Labour Statistics) have trialled the use of the World Wide Web for collecting business data.  Increasingly CASIC methodologies are used in combination with experimental designs to increase the capability and power in testing questionnaires.  Some benefits of CASIC are data quality improvements through editing at source, and controlling errors caused by inaccurate question skipping through computer driven branching.  There is ample evidence in the literature that these methods have offered improvement in data quality, timeliness and reduced costs. 

OMS and IWMS are ABS terms for the systems providing the functionality described above; other organisations using CAPI may have different terms for these systems.

In the ABS, many initiatives have already been taken to improve the quality of the data we collect.  In any agency, a new collection and processing system should support past initiatives, making them easier to implement and maintain.  The introduction of more automated methods using new technologies has the capacity to significantly increase what is possible in some areas of data collection. 

Although it can be difficult to measure tangible benefits, improvements in data quality can be expected to be substantial under CAPI.  Improvements in data quality are possible through the use of technology to simplify some of the difficult parts of data collection, and provide a more consistent approach to interviewing.  Potential Quality benefits of CAPI may include:

• On-line edit checks.  Elimination of sequencing errors and reduced item non-response
• Improved statistical methodology (e.g. reduced biases)
• Interviewer instructions available from the CAPI system when required
• Standard approach to prompting
• Better quality data from use of Dependent Interviewing in continuous or longitudinal surveys
• Ease of changes to surveys for testing purposes and facilitation of methodological research.
• Timestamping options of CAPI systems
• Ability to ask more detailed questions
• Ability to better target population groups
• Potential to reduce sample size because of improved data quality
• Possible better response to sensitive questions (although the use of a computer can also raise privacy concerns with respondents)

The most significant impact on data quality arises because the editing of data takes place at the interview and therefore, supplementary information may be sought from respondents.  If the CAPI instrument is correctly implemented it eliminates errors caused through incorrect sequencing of the questionnaire.  The number of missing values should be reduced as questions are not missed through incorrect sequencing, and item non-response only occurs in situations where the respondents are unable or unwilling to answer the questions.  The ability to use complex sequence guides, or to sequence based on derived items, also assists with reducing respondent load.  (But on the other hand, testing of the instrument must be done carefully, as the automatic sequencing can have some undesirable effects if the coding has been done incorrectly!)

One of the major problems perceived with many surveys can be the length of time between collection and release of survey data.  A high priority for any new system would be to reduce the time lag between collection and publication.  New systems should also have the potential to reduce the total amount of time spent in developing and processing surveys. 

One of the major problems perceived with many surveys can be the length of time between collection and release of survey data.  A high priority for any new system would be to reduce the time lag between collection and publication.  New systems should also have the potential to reduce the total amount of time spent in developing and processing surveys. 

• Elimination of separate data entry phase
• Use of field coding systems (either in-built or agency-developed coding facilities) to reduce respondent load and improve coding reliability
• On-line derivation of many output data items (which can be edited or used in sequencing)
• Electronic transmission ensures that the data get back for processing more quickly

Quality methods of survey development require that the clarification of all output requirements of a survey is done early in the collection development phase, well before interviewing takes place.  CAPI encourages this type of survey development as, ideally, derived items and edits should be programmed into the survey instrument and examined as part of the testing of the survey.  Nevertheless CAPI technology does not force this to be done, only management commitment to it does.  Testing should uncover where questionnaires and derivations require more work and, if problems are fixed during the development stage, much of the reworking of data items currently done after data collection can be eliminated.

Quick counts and summary statistics may be available earlier through use of a CAPI instrument than would be the case with traditional data capture methods.

CAPI has the potential to reduce the costs of surveys compared with other non-CAI methods, depending upon procedures used.  From the research literature, it seems clear that CAPI systems require less resource on an ongoing basis, but that the set-up costs (mostly equipment costs) can be considerable.  Reductions in ongoing costs are possible through:

• improved coding cost effectiveness
• reduced respondent load by asking only the questions necessary and thereby decreasing the time required for each interview
• elimination of separate data entry (less applicable if OMR or other non-key entry method is already being used)
• streamlined processing as data is captured and edited, and derived items are created in one step
• reduced rework during processing, as the derived items and edits will have been specified and checked during the test phase
• reduction in clerical procedures, such as printing and distribution of forms, by replacing with automated methods using communication technology
• reduced maintenance of "old" systems and hardware is possible, for example, eventual elimination of the OMR equipment

• Set up costs (especially hardware) for CAPI can be considerable
• The costs for development of a CAPI instrument are usually greater than for a  paper form
• Requirements for transmission of data back to processing area - there is a need to look at the technical and communication infrastructure and budget to decide what is best
• Blaise (the DOS version) does not have the graphics facilities to present questions as they appear on paper questionnaires.  In particular, font size cannot vary, and italics and underlining are not available.  Therefore, certain changes to paper conventions on the presentation of questions, word emphasis etc. are necessary.  Such changes need to be clearly outlined at the training sessions.  This issue might be resolved with Blaise for Windows.
• Approach to system development - an agency contemplating moving to using CAPI for population surveys should determine an approach to evaluation and development of systems to ensure a controlled development and successful implementation.  One known problem area is to try to handle too much too soon - if resources are not sufficient to develop and support the number of surveys required to be implemented with CAPI, then problems will arise in the development of facilities in time for smooth deployment in the field, and many more support problems will be encountered.  The approach proposed is for an initial small scale test followed by implementation of the system for a full test; that is, it is aimed at a staged approach to the development task.

• Interviewer training takes longer - usually one additional day for small to medium surveys and two additional days for more complex ones (such as the ABS Household Expenditure Survey).  Interviewers to be trained for CAPI surveys should be experienced in household-based surveys.  This will enable them to become familiar quickly with the concepts of the survey instrument, and appreciate the efficiencies which the CAPI system provides.   Those interviewers who have undertaken a CAI survey within the last 6 months, and are required to enumerate another CAI survey, are given a half day Refresher training course.  Interviewers who have not used a CAPI instrument before are given 2 days of initial training. 
• Interviewers might have difficulty in understanding the overall structure of the questionnaire compared with a paper form (that is, the structure appears to be hidden from the interviewer's point of view).
• Perception of privacy problems by some respondents (although the use of a computer can appear an aid to privacy for other respondents) - interviewers need to be trained to deal with queries about the use of the computer and the data collected through it.
• Keycoach, a typing tutor program, can be set up on the computer so that interviewers can improve their typing skills.  This is not essential, but the interviewers appreciate the opportunity to improve their skills and decrease the time required to capture data for each interview.  An alternative to providing a typing tutor program would be to recruit interviewers who were already experienced and fast typists, but it is not actually critical to have high typing speeds - in fact some interviewers who are high-speed typists find that the instrument is too slow for them.

• Occupational Health and Safety (OHAS) issues are those which relate to the health and safety of staff in the work environment.  For interviewers, working out in the community contacting members of the public, OHAS presents different issues from those operating in an office, factory or other workplace environment.  For the interviewer, OHAS and personal safety are key aspects.  In this regard, it is the overriding policy of the ABS that interviewers should not put themselves at risk in the performance of their duties collecting survey data.    Some of the  main issues that relate to the use of a computer for interviewing are 
• Overuse Injuries, which can be sustained from incorrect posture and handling of personal computers
• Trip and Fall Hazards, which can happen whilst walking, climbing stairs, or tripping on uneven surfaces  and
• Interviewing with a Computer - Interviewers need to be trained in and aware of the procedures that must be followed to conduct an interview with a computer.

• The introduction of CAPI would have a significant impact on the survey development process and on the work of many agency staff which would need to be considered as part of any CAPI evaluation.  The main areas of impact identified are:
• Development timetables and responsibilities
• Potential effects on interviewers and permanent staff
• Effects on respondents and privacy/security issues
• Processing system implications
• Impact on statistical series
• Other operational issues

• Data security is an issue that needs to be taken very seriously.  The interviewers need to sign for their computers and be aware that all activity with those computers is their responsibility.  The notebook computer contains confidential information.  Interviewers need to understand that they have been entrusted to ensure the confidentiality of this information.  This means that:
• key security identifiers such as Personal Identification Numbers and passwords must be kept secret, and should not be written down or stored near the computer;
• key security identifiers must not be given to or shared by anyone.

The interviewers must be instructed that they must not hand their computers to any person other than their Supervisor or other authorised agency Officer.  This includes respondents and members of the interviewers' families, even if they are 'helping' with a technical problem.

• provision of a support hotline number which is attended during normal office hours and on weekends, to provide assistance for transmission, hardware or questionnaire problems 
• support for more complex technical problems
• transmission or re-transmission of instruments
• distribution of software patches and upgrades 
• transferring interviews between interviewers
• handling incomplete workloads

The description following is based on the ABS CAPI system, but many of the features would be common to CAPI implementations in general.  Some detail is given on hardware and software considerations.

CAPI requirements lead to the development of a new sample and workload formation system.  Workload formation is the allocation of households to interviewers.  These facilities are not, for the most part, CAPI specific, and they can be used for the general household survey program.  However the redeveloped system provided a convenient way to "add" electronic generation and transmission of workloads to CAPI machines.  The sample selection facilities are based on standard RDBMS client-server technologies and are integrated with the standard ABS applications environment.  These facilities were designed to provide the basis for later standardised estimation components.

An Office Management System (OMS) is required to handle the non-field aspects of CAPI management. Processing required before dispatch and after receipt requires a range of office facilities, including instrument development, initial notebook configuration, resolution of edits, additional editing, reallocation of incomplete interviews, backup and recovery, etc. 

The ABS has implemented an Office system in Blaise, which handles a major part of the requirements, and also a client/server based system using the ABS' Oracle RDBMS and the standard processing system interface (SPEED).  These other processes such as workload formation and Interviewer Assignment are also used for non-CAI surveys such as Monthly Population Survey (MPS).

The CAI OMS essentially involves 4 major processes:

• A generic front end interface that allows selecting of a survey and its cycle and selecting of a required function.
• The CAI data movement functions which include:
• preparing and sending data to interviewers 
• receiving of data from interviewers
• moving response status data to Speed systems and moving final data to a state file for  further processing
• final handover of final clean data for further processing 
• The "cleansing" of data within the office including limited coding. 
• Management information on response status & interviewer progress

There is a variety of different ways in which the data collected in the field can be returned to the processing area.  If electronic communications are difficult or unreliable, it may be most efficient to download the data onto a floppy disk, and hand-deliver or post this to the regional or central office.  Another possibility might be to return the whole computer to the processing area, but this has obvious implications for speed, efficiency and security.

If the communication infrastructure is good, then the most efficient ways of returning data would be by using an encrypted modem or via the internet, depending on the sophistication of the communication infrastructure and data security requirements.

In the ABS we use encrypted modems to return the completed workloads to the office.  We use a product called File Transfer Protocol (FTP), in conjunction with security software called SecurID.  The interviewers are provided with a modem which plugs into their ABS telephone line and laptop computer.  Details on connecting the modem and instructions on how to receive and send data are covered in interviewer training.

The transmission to and from the field uses encrypting modems, authentication fobs and specialised secure servers.  On top of this hardware layer, a simple but robust software system provides a convenient transmission scheme.  The field notebooks have automated dialup connection from the PC to a server (various states have a server).  To establish a connection one must not only identify oneself, but provide a 'password' generated by and frequently changed by an authentication 'fob'.  This essentially checks that the dialup party has a currently authorised fob in their possession, as well as being a recognised field operative using an encrypting modem.  Having established a connection, we transfer the content of a person-specific directory in each direction (using FTP).  Any special 'automatic execution' files delivered to the PC will execute after closing the link.  The server provides for periodic transmission of collected files to the ABS network.

A small portable computer is required for CAPI enumeration.  Portable computers currently take the form of Laptops or Notebook PCs, Handheld PCS (HPCs) and Palmtop PCs.    The choice of which of these is used depends on their price, size (most notably weight), processing power, ability to support the software used, ease of data entry, level of technical expertise for support of the computers and field conditions.

Any business case for CAPI needs to cost the initial outlay for computers and associated facilities (modems, encryption software, security software, additional batteries, interviewing stands and so on) and also needs to consider the replacement cycle for the hardware.  It is necessary to plan to replace the computers about every 4 years.  If the computers are kept in service for longer than this then problems can arise with malfunctions, hard disk size becoming too small and more frequent support required.

Support costs are also major issue and will vary with the arrangements within each agency.  For the ABS, the cost of a computer is $7,000 for 3 years ($3,000 for lease costs, $3,000 for support and $1,000 for infrastructure costs). 

The computers used for ABS CAPI applications are IBM colour notebooks, capable of running Windows, but as Blaise for Windows was not available at the time the ABS application was developed, the CAPI system itself requires only DOS.  The notebooks are pre-loaded with the CAPI instrument, and cannot be used for other purposes.  The computers are reasonably sturdy, however, they will not cope with being dropped, filled with crumbs, or having liquid spilled onto them, so interviewers need to be trained to look after them carefully.

The machines include disk encryption and access control software, and the modems plug in 'at home'.  Extra batteries and 'door stop' interviewing stands are included.  The stands are rarely needed - most interviewers are invited in and often permitted to use mains power.

Whether notebooks or palmtop computers are used, there are some features which should be looked at with regard to the hardware requirements, such as:

• Screen visibility and readability - the screen should be readable in the field environment taking into account a range of likely lighting conditions.
• Computer Speed and memory - speed and memory requirements are driven mostly by software needs.  Software techniques may be employed to improve performance.  The performance will be affected by the nature and complexity of the particular survey and its implementation.
• Battery charging - batteries are a major support issue.  Interviewers should try to avoid the situation where the batteries run out mid-interview - they should check the battery status before the interview, and save the interview when they first hear the low battery warning beeps.  If the batteries do run out mid-interview, and the interview hasn't been saved, the entire interview or any changes made since the last save for that household could be lost.  Interviewers should be encouraged to remember to save the interview regularly and always save before changing batteries.

The choice of a software product for building CAPI instruments depends on a number of factors, including the availability of products in the country, the type of hardware being used and the complexity of the particular surveys.  The type of hardware may be the more important consideration as this will govern what software can be used efficiently and effectively.

In the ABS, Blaise has proved to be a useful and powerful environment for CAPI applications (as well as some straight data entry applications eg diary systems).  Its entry, coding, sequencing and editing capabilities are well suited to the needs of household CAPI.  It is also used for field management software (contact lists, workloads), as well as household instruments and individual instruments.  This gives substantially the same look and feel for most operations, reducing training and support loads. 

Other software can be loaded and used on the notebook, for instance some specialised software was used during the Survey of Mental Health and Well-being.  The basic survey instruments were generally developed by subject matter processing people, though specialist IT programmers wrote, integrated and maintained substantial Blaise programs.  Over time, particular 'styles' for writing Blaise programs evolved that both avoided some common problems, and reused standard ways of doing certain types of operation.  These would have (or will) become better documented 'standards' if Blaise continued to be used.

While the ABS has used Blaise, there are other options.  A decision on the package or packages to be used should be made after looking at available software, gaining opinions from other agencies which have used particular packages and measuring the software against the evaluation criteria which are deemed to be the most important. 

Essential criteria for CAPI software products are that the product must:

• Provide the necessary functionality for a CAPI system according to stated requirements (eg skip logic, editing functionality, speed of operation).
• Be easy to use by survey developers, interviewers in the field and programmers.  There are issues for all of these, but in terms of importance ease of use by survey developers and interviewers takes priority, as programmers will always find a way to make the system work.
• Have adequate support and training in the system provided by the vendor.
• Be recommended by other statistical organisations (in that they have not had significant problems with the system or vendor).

The following diagram shows the core processes used during CAPI survey operation.  This is followed by a more narrative account of the main workflow path.  Various precursor activities (such as instrument development, field trials, and interviewer training) and later processing steps are not covered.

This process deals with taking the sample selected and allocating work to interviewers in accordance with certain rules.  ABS-standard processes in ABS-standard processing environments were used (although the software was redeveloped), with additional processes to prepare the information for electronic dispatch to CAPI machines.  Instruments can be in the field for various lengths of time - up to a year, and workload allocation has to take into account field trials of proposed instruments.  The workload for each interviewer may need to be merged with already captured information for some interviews, for example for incomplete response or multiple interview instruments.  Problems usually occurred when the workloads were not quite standard - eg. when the household system was 'adapted' for the longitudinal person-based Survey of Employment and Unemployment Patterns.

Instrument distribution used two techniques, either preloading instruments to disk before dispatching to interviewers, or transmitting the new files via the modem based file transfer system.  It is desirable to cleanly separate the transmission of software and the data.  Retransmission of an instrument is sometimes necessary, and care needs to be taken that previously captured data is not lost if this occurs.  A register of software (and data) present on each field machine would simplify matters.

A workload is a set of interviews to be performed by an interviewer.  An interviewer may have several workloads at the same time.  Transmission is by the modem based file transfer mechanism, though care needs to be taken when handling multiple workloads and/or surveys.  Mechanisms that risk overwriting already collected information should be avoided, and the central servers should be capable of acting as a remote backup device.  However data separation need not be complete as on occasion it is desirable to use data collected by one instrument in processing another. 

Patching and upgrading is not desirable when instruments are in the field, but it is sometimes necessary.  When needed, we use the file exchange mechanism and 'automatic execution' feature.  Synchronising software release with necessary data patches is particularly error prone.  Some means of 'switching off' an instrument till the completion and success of software installation has been checked would be useful.

A Blaise based system is provided, but its use is not mandatory - some interviewers usei it, some don't.  Typically, newer interviewers find it useful but more established interviewers have already established effective work practices.

Case selection is provided by a Blaise-based front end program, the use of which is mandatory - essentially it launches the Blaise instrument on the appropriate record(s).  This part of the system records the basic status information and can hold commentary useful for subsequent visits/contacts, office editing, etc.

The Interviewer Workload Management System (IWMS) as the name suggests allows interviewers to manage their workloads for each of their surveys.  The system will allow them to carry out a range of organisational procedures, all of which they have previously used paper for.  They are able to view their workloads, select a household for interview, make appointments, view appointments, view previous calls to the household, make comments, set the final response status etc. 

Field capture is essentially by direct entry at time of interview, but sequencing is always fully automated and this provides the ability to field instruments with much more complex sequencing paths (although it causes some 'instrument validation' problems).  For coding, picklists are used where practical, otherwise either Blaise based trigram coding is used, or free text captured for later resolution.  Verification is at-time-of interview when possible, improving data quality and reducing downstream problems, but edit override and later editing are also enabled on some fields.  Remarks can be inserted against any fields, and post-interview editing can make use of these comments.

A help desk number is provided, with third line support available for the relatively few 'technical' problems.  Most problems are with instrument interpretation and relatively simple technical problems.  Incorrect early 'finalisation' of an interview (blocking further data entry and editing) is a common non-technical problem.  Transmission problems are the most common technical problems.  There are many possible causes of these problems, such as:

• interviewer inexperience
• phone line failure
• hardware problems
• network unavailability in the office
• wrong workload given
• transmission software problems

Returning data uses the above file exchange mechanism.  Before dialling up the server a program is run to extract data to be transmitted, the data is 'zipped' and encrypted.  The ability to take a full snapshot of status at transmission time proves a useful backup and recovery mechanism.  Data integrity needs careful thought when the data can reside on central servers and one or more remote machines.  Unlocking field records for further work requires care that the office record (and any changes made to it) be properly handed.  Transferring interviews between interviewers has the potential to cause similar integrity problems particularly in longitudinal surveys, but also in incomplete response situations, or when an interviewer could not complete a workload.

Initially the original Blaise instrument was used and  progressively extended over time to incorporate derivations, new edits etc.  Sometimes it was useful to do 'batch' operation opening and reprocessing/resaving each interview record, although Blaise had some limitations in this regard being oriented to on-line processing.

Manipula (a standard part of the Blaise environment) is mostly used to extract data.  Typically, a SAS based environment is the target for further processing.  Blaise field naming conventions cause a few problems in transferring the data to SAS, but the process is quite manageable.

• two waves of the longitudinal Survey of Employment and Unemployment Patterns (1996 and 1997)
• the Survey of Mental Health and Well-being (1997)
• the Survey of Disability, Ageing and Carers (1998)
• the Household Expenditure Survey (1998-99) and 
• the Housing Survey (1999 - in the field at present).

The CAPI instrument for the Household Expenditure Survey was very large because of the size and complexity of the questionnaire, thus the system was slow in use at times and required the use of mains power as much as possible.

The CAPI software (Blaise) has also been used for straight data entry of Time Use diaries and Household Expenditure Survey diaries.   The 1999 Housing Survey will be the last survey to use CAPI for the time being - the next possibility is the 2002 Household Expenditure Survey, depending on management decisions from the new Business Case for CAPI currently being discussed by ABS Senior Management.  This business case focusses on the conclusions drawn from the experiences over the last 3 years and is based on gains from data quality and timeliness rather than cost reductions as such.

One problem that the ABS will face if there are no CAPI surveys for the next 1-2 years is the loss of CAPI expertise from the survey design and application support areas.  This will mean effectively starting again in some ways, although the current systems have been well documented for future reference.

CAPI in a Blaise/Notebook environment has provided significant gains in the complexity and nature of instruments that can be fielded, and in the quality of data captured.

One of the largest cost savings from moving to CAPI is usually the reduction in the cost of data keying.  In the case of the ABS however, migration some years ago to OMR for data capture from paper forms has meant that the business case for CAI has been much harder to make than was the case for example in both Canada and the UK, where CAPI replaced large scale data entry facilities.

Moreover, the time and cost savings from CAPI occur mainly in the back-end processing and  these savings need to be sufficient to offset additional costs in front-end development and infrastructure investment.  Because these latter additional costs are largely fixed whereas the back-end savings are variable, net savings are most likely to be realised when CAPI is used on large scale, continuous surveys.  It may not be surprising therefore that CAPI has proved an expensive option for ABS when applied to only a few, comparatively small scale, non-continuous surveys.  In fact, many overseas agencies have introduced CAPI first either on a large, continuous survey and set the infrastructure costs against that before implementing it on other surveys, or they have introduced it simultaneously across all, or at least the great majority of, surveys that they carry out.   In the current ABS context, it is difficult to make the busines case for CAPI on cost savings grounds alone.

The case for CAPI on cost reduction grounds alone is, in any case, not always clear cut.  Some overseas agencies report substantial cost benefits, others are more equivocal.  However, all state that there are other substantial advantages for CAPI in improvements in speed and quality of a survey.

The experience that setting up a CAPI survey requires an additional 3 months up front can be argued to be a benefit since it imposes an important discipline on the survey clients and designers in getting the specification and development work at the beginning of the process, where it should be.   Because CAI requires the time to be spent more at the development and testing stage at the beginning of a project there is less likelihood of problems that will require extensive post fieldwork reworking and correction.

One of the concerns expressed from several areas has been the growing complexity of survey content. Particularly in respect of complex survey instruments, the quality dividends from CAPI are already well known  as it supports automated derivations, eliminates routing errors, and builds range and consistency edit checks into the actual interview process.  This provides an important argument in favour of CAPI since it enables more complex collection instruments to be used with less likelihood of failure or breakdown.

The major issue for the ABS in terms of future use of CAPI technology involves establishing a sound business case.  For such a case to be successful, it must clearly demonstrate that the benefits overall from adoption of CAPI are going to be sufficient to justify the substantial costs involved in deploying the necessary hardware and systems to support its introduction and ongoing use.  In this respect, it is important not to define the benefits only in monetary terms.

In the ABS, consideration is being given to future CAPI options including notebook or palmtop based options,  SSS or Labour Force fleet size, and doing without CAPI for a while.  A recent review of the household survey program in the ABS canvasses the better exploitation of the wide range of information held in administrative datasets.  It also questions whether the marginal benefits in terms of data better suited to user needs warrants the added complexity that has increased costs and time taken to develop and process each survey.  However the review recognises the desirability of a range of different data collection options including computer assisted interviewing.

A new  business case currently being discussed by management supports continuation of CAPI but notes that the business case is not based on reduced costs as such (as the ABS has already made significant cost reductions by moving from key entry to OMR some years ago), but on the improvements in data quality and timeliness possible through using CAPI instruments.

Savings will arise principally from shorter interviewing times (also reducing respondent load), reductions in related clerical activities undertaken during and after field enumeration, savings from process re-engineering, forms printing, mainframe costs and data capture costs.  Savings are also expected by appropriately reducing the sample size of surveys to offset the improvement in the Mean Squared Error (ie sum of Sampling Variance, Response Variance and Bias Squared) of survey estimates brought about by CAI.

See Attachment 3:  Sample Business Case, for further details.

For the current ABS CAPI implementation, three facilities would, with the benefit of hindsight, been very useful.

The first, providing "remote debugging", would provide the ability to control a PC from a remote site (ie. the in-field machine from a central help desk).  This would have greatly simplified problem resolution at times.

The second is the provision of fairly (Windows-)standard news/mail/groupware facilities.  At the time Notes and Windows were available but the training overheads, particularly given the difference between BLAISE/DOS and Notes/Windows, were though too high.  With Blaise for Windows now available, and a better appreciation of the potential benefits of additional facilities, the decision would probably be reversed today. 

The third is the provision of additional management information systems - particularly around tracking hardware location, help desk queries and hardware failure records, software instruments downloaded to notebooks, and some exception reports - treatment of special dwellings etc.

Blaise for Windows has been trialled for the Housing Survey Office Management System, but problems have been encountered because the actual Housing Survey instrument is in DOS.   Blaise for Windows processes multiple commands at a time, whereas the DOS version processes commands one by one, so it appears that the instrument would need to be re-written for Windows.  As the Housing Survey is the last CAPI survey being undertaken by the ABS for the time being, the effort required to re-write the instrument for Windows cannot be justified at this time.

Using CAPI for the capture of population data can deliver significant benefits in terms of:

• data quality 
• timeliness
• cost savings

CAPI also provides the opportunity to apply experimental design principles to survey testing, and the flexibility to enable collection of important policy related data for difficult topics.  Benefits such as enabling the targeting of specific population groups through the use of sophisticated question sequences would not be possible using a PAPI questionnaire.

There is a need to look carefully at survey/census requirements and investigate the advantages and disadvantages of each type of capture technology before deciding on what is appropriate for the requirements of the organisation. 

The cost/benefit calculations will depend on factors such as:

For further information or answers to any questions that might arise from this paper, please contact

• rob.edmondson@abs.gov.au
• rosemary.crocker@abs.gov.au

• CAI Training Manual Contents
• Paper on Blaise for Diary and Office Processing 
• Sample CAPI Business Case