AUTOMATED DATA LINKING

Automated Data Linking (ADL) refers to the use of automated data linking processes to determine possible links between Census and PES data before any clerical matching process begins. ADL was introduced into the PES in 2011 for the first time and will be used again for 2016.

The approach to data linking is to generate large numbers of candidate links and then use a process of elimination to filter down to genuine matches. ADL employs probabilistic data linking techniques (Fellegi et al 1969), using a range of personal and address characteristics, to evaluate the chance that a PES record and a Census record are the same person. ADL therefore provides the opportunity to match persons that would have been too difficult to match using only clerical match and search procedures. The key gains in matching effectiveness and efficiency provided by ADL include:

• the ability to conduct a more comprehensive search for PES respondents than was possible from previous clerical matching processes;
• the ability to locate PES respondents at Census night addresses, regardless of whether PES respondents recall all the locations they may have been recorded on a Census form; and
• a reduced requirement for clerical matching resources.

ADL uses Freely Extensible Biomedical Record Linking (FEBRL) software developed at the Australian National University. A number of different linking runs will be used to compare PES and Census records, each focused on a slightly different combination of name, addresses and demographic variables.

Potential links are assessed by assigning weights that reflect the level of agreement on selected data items from the two records. Large positive weights indicate probable matches, while large negative weights are observed for probable non-matches.

Important to the effective use of ADL is a series of processes that are run after ADL output is obtained. The Collect, Analyse, Reduce, De-duplicate and Systematise (CARDS) process identifies and rates the most plausible links from each ADL run for all PES respondents. The process then combines the links from all ADL runs and removes any duplicates.

While ADL is the next step in the evolution and continual improvement of PES processing, it is important to note that ADL cannot entirely replace the clerical decision-making process that has previously been at the core of PES processing. Clerical judgment will always be required to resolve the more complex or ambiguous cases and be used as a means of quality assuring automated processes.

To clerically confirm or reject potential person links identified by ADL, it is efficient to handle links for all persons in the same PES dwelling, so as to examine the entire household and their corresponding Census form at the same time. For this reason the CARDS process concludes by creating dwelling links and assigning Dwelling Link Ratings (DLR), to determine which PES dwellings will be examined by the PES Match and Search System (MSS), and which can be confirmed as links without further review. Person-level links are grouped together according to the dwelling they were found in PES, and dwelling-level links are derived and rated on the basis of the number and quality of person-level links identified between the PES and Census dwellings.

PES uses three Dwelling Link Ratings:

• Platinum links are of sufficiently high quality that they are confirmed immediately and do not require clerical review;
• Silver links are of moderate level quality, for example high quality links between some persons in a dwelling and not others, and require clerical review to confirm or reject the quality of the link;
• Tin links are of low level quality and unlikely to be true matches, so clerical review is required to search for matches for these dwellings and persons without the assistance of ADL.

All PES dwellings with either Silver or Tin links are sent for clerical review. A small percentage of Platinum links are also clerically reviewed for quality assurance purposes.