Tax linkage rates

View the most recent version.

While respondents may grant Statistics Canada permission to use their tax data, they are not asked for their Social Insurance Number (SIN). Without a SIN to identify SLID respondents on the tax file, it is necessary to perform a linkage operation to find a respondent's SIN. The generalized record linkage system (GRLS) developed at Statistics Canada is used to perform this linkage.

After preprocessing of both the tax file and the SLID file to ensure compatible formatting of all match variables, a direct match is performed using 7 key matching variables. These matching variables are: Sex, province, soundex² code for surname, surname, date of birth, postal code and first initial. The SLID record can have no missing data for key matching variables. Output for the direct match is manually reviewed for errors where a SLID record matches to more than one tax record, where more than one tax record matches to a SLID record, and where the first given name is not the same on the 2 sources (only first initial is used in the tax match). The match rate on the direct match is approximately 55 percent.

The unmatched records are then run through a statistical match. Pockets³ for matching are defined. The files are segmented into pockets with sex, province and surname soundex code defining a pocket. Every record within a pocket on the SLID file is compared with every record within the same pocket on the tax file. Factors of importance are assigned for full agreement, partial agreement, and disagreement. These factors are numeric values and are used to evaluate the likelihood that a pair of records (one from SLID and one from tax) represent the same person. Factors are defined for each of the matching variables. Thresholds are defined whereby records are determined to be definite matches if their total factor is greater than the upper threshold or definite non-matches if their total factor is below the lower threshold. Manual verification is done to ensure the quality of the matches. Figure 7.1 gives the percentage of the SLID sample giving tax permission for which a SIN can be found. Since some respondents who give tax permission have not filed a tax return not all cases for which a SIN is found will result in successful tax linkages. Figure 7.2 gives tax linkage rates for those in the SLID sample for which we were successful in finding a SIN.

Figure 7.1 SIN found for respondents giving permission (%)

Description

Description

Figure 7.2 Tax linkage rates where a SIN was found (%)

Table 7.1 compares the proportion of records from tax to those collected in the telephone interview. In total eighteen income variables are imputed during SLID income imputation. Many individuals require only partial imputation. Partial imputation is when one or more income items is imputed with some information being supplied by the individual.

Table 7.1 Income data coming from tax or interview (%)