The Feasibility of linking household survey and administrative record data: New evidence for Britain

Figures & data

Table 1 Record Linkage Rates (%) for ISMIE Respondents

			All who gave
		ISMIE	consent to data	Supplied	Did not
		sample	linkage	NINO	supply NINO
Criterion and matching variables		(1)	(2)	(3)	(4)
Independent matching
1. NINO		48.2	62.1	70.0	–
2.	Sex, date of birth, postcode	49.7	64.0	64.3	61.5
3.	Sex, date of birth, forename, family name	47.9	61.7	62.6	55.0
4.	Sex, postcode, forename, family name	41.7	53.7	54.4	48.4
5. Sex, forename, family name, address line 1		33.7	43.4	44.3	36.3
Pooled matching: at least one of the above		57.3	73.8	74.5	68.1
Hierarchical matching
1 followed by 2, or 2 followed by 1		56.4	72.6	74.1	61.5
2 followed by 3, or 3 followed by 2		56.1	72.1	72.7	68.1
N		1033	802	711	91
(as % of all who gave consent)			(100)	(88.7)	(11.3)

Note:

The table includes potential mismatches (see section ‘Linkage Rates’ in the text). NINO, National Insurance Number; ISMIE, Improving Survey Measurement of Income and Employment.

Table 2 Linkage Outcomes among Consenting ISMIE Respondents

	All who gave consent to data linkage		AAll who gave consent and supplied a NINO
Linkage outcomes^a	Frequency	Percentage	Frequency	Percentage
00000	210	26.2	181	25.5
00100	7	0.9	3	0.4
00101	2	0.3	0	0
01000	16	2.0	4	0.6
01110	20	2.5	7	1.0
01111	49	6.1	18	2.5
10000	11	1.4	11	1.6
10010	1	0.1	1	0.1
10011	1	0.1	1	0.1
10100	47	5.9	47	6.6
10101	10	1.3	10	1.4
11000	68	8.5	68	9.6
11110	74	9.2	74	10.4
11111	286	35.7	286	40.2
All	802	100.0	711	100.0

Notes:

^aOutcomes for criteria 1, 2, 3, 4, and 5 (in that order), with ‘0’ meaning not matched, and ‘1’ meaning matched. For example ‘10010’ means respondent matched by criteria 1 and 4, but not by 2, 3 or 5. Note: The match criteria are defined in the text and summarised in Table 1. The table includes potential mismatches (see section ‘Linkage Rates’ in the text).

Table 3 Estimates of Linkage Accuracy

		False-positive rate			False-negative rate
Matching method		%	(N)	%		(N)
Independent matching
1.	NINO	2.2–11.6	(498)	30.9		(304)
2.	Sex, date of birth, postcode	0	(513)		23.9–27.3	(289)
3.	Sex, date of birth, forename, family name	0–10.9	(495)	30.6		(307)
Hierarchical matching
1 followed by 2		1.9–9.9	(583)	4.1		(219)
2 followed by 1		0.5–8.6	(583)	4.1		(219)
2 followed by 3		0–8.1	(579)	4.7		(213)
3 followed by 2		0–9.3	(579)	4.7		(213)
Pooled matching
Match by at least one of 1–5		0–8.6	(592)	0		(210)

Notes:

Independent, hierarchical and pooled matching defined in the text. False-positive rate for criterion m = percentage of matches by m that were mismatches according to criteria other than m. False-negative rate for criterion m = percentage of non-matches by m that were genuine matches according to criteria other than m. Estimates of false-negative rates assume that all 210 cases with match pattern ‘00000’ were not benefit recipients (see text). N refers to the number in the denominator of the relevant rate calculation.