Before using resources for one-to-one mapping, please ensure that you are certain that a one-to-one mapping is most suitable for your purposes. Further discussion can be found on my LFS SIC Mapping page.

1. Office for National Statistics (ONS) one-to-one mapping between SIC 1992/SIC 2003 and SIC 2007

The ONS realised that a proportional mapping based on the IDBR weighted correspondence between SIC 1992 (2003) and SIC 2007 (discussed here) would not be suitable for many purposes using LFS data, where those purposes required a mapping at individual level that was consistent over time. The ONS LFS team therefore devised a modified direct, one-to-one, mapping. It is this direct mapping that is used to generate the ongoing industry information based on SIC 1992 after the changeover to SIC 2007 in 2009 Q1. The QLFS data from 2009 Q1 use a SIC07-SIC92 mapping from SIC 2007 Classes (4-digit level) to SIC 1992 Divisions, Sections and Sectors (2- and 1-digit levels).

The reverse SIC92-SIC07 would be useful for time series analysis if SIC 2007 classification is wanted throughout. Sean Milburn of the LFS Research team at the ONS Social Survey Divison has very kindly made available his coding to map SIC 1992 Classes to SIC 2007 Divisions or Sections.

• SIC92 Class (4-digit) to SIC07 Division (2-digit) one-to-one mapping, implementing a recode in a STATA do-file.
• SIC92 Class (4-digit) to SIC07 Section (1-digit) one-to-one mapping, implementing a recode in a STATA do-file.
• ONS document prepared for the LFS Steering Group in October 2010 on the effect of the change to SIC 2007, including information about the mapping method used in the LFS.

2. Jennifer Smith's one-to-one mapping between SIC 1980 and SIC 1992

Using two alternative methods, I have created direct mappings between SIC 1980 and SIC 1992.

The first method is based on overlapping information provided in the LFS Autumn 1993 dataset on SIC80 Activity and SIC 1992 Section.

• SIC80 Activity (4-digit) to SIC92 Section (1-digit) one-to-one mapping based on overlapping data provided in the Seasonal Quarterly LFS dataset Autumn (September-November) 1993, implementing a recode in a STATA do-file.
• Excel spreadsheet containing SIC80-SIC92 overlap-based correspondences and giving information about the overlap-based one-to-one mapping.

The second method is based on individuals remaining in the same job (stayers) between Autumn 1993 and Winter 1993/94. Autumn 1993 data give SIC 1980 Activities and Winter 1993/94 data give SIC 1992 Divisions and SIC 1992 Sections.

• SIC80 Activity (4-digit) to SIC92 Division (2-digit) one-to-one mapping based on job stayers between Autumn 1993 and Winter 1993/4, implementing a recode in a STATA do-file (REVISED 13 October 2016 to include previously omitted SIC codes when using recode based on ONS codes).
• SIC80 Activity (4-digit) to SIC92 Section (1-digit) one-to-one mapping based on job stayers between Autumn 1993 and Winter 1993/4, implementing a recode in a STATA do-file.
• Excel spreadsheet containing SIC80-SIC92 stayer-based correspondences and giving information about the stayer-based one-to-one mappings (REVISED 13 October 2016 to include previously omitted SIC codes when using recode based on LFS codes).

This STATA do-file implements both versions of the 1-to-1 map and effectively explains how the Excel spreadsheets and recodes are derived.

3. Jennifer Smith's one-to-one mapping between SIC 1968 and SIC 1980

With the aid of invaluable advice from Sean Milburn at the ONS, and the spreadsheet of correspondences available from the LSE RLAB, I have created a direct mapping between SIC 1968 and SIC 1980.

• Information about the mapping method:

  • The LSE RLAB spreadsheet of correspondences matches, so must be derived from, ONS documents showing SIC68-SIC80 correspondences available on the ONS website of Archived SICs. See, for example, this pdf document.
  • Unfortunately, published documents contain only correspondences between SIC 1968 Minimum List Headings and SIC 1980 Activity Headings. Where there is more than one SIC 1980 Activity matching a given SIC 1968 MLH, there are no published proportions or other guidance on which is the dominant correspondence.
  • There is no LFS dataset in which both SIC 1968 and SIC 1980 classifications are simultaneously observed.
  • In these circumstances, all mapping methods are likely to be flawed.
  • The method I choose generates proportions for each SIC 1980 Activity based on their relative frequencies in the 1981 LFS dataset (which is the first one classified according to SIC 1980). This method is not ideal, because it is based on aggregate frequencies of each SIC 1980 Activity. If it were possible to simultaneously observe SIC 1968 and SIC 1980, a better alternative might be to use frequencies of each SIC 1980 Activity within each SIC 1968 MLH (which would mirror the technique used for my stayer-based SIC80-SIC92 mapping described above).
  • Classification error will tend to be lower, the higher the level of aggregation chosen for the final classification.

I have mapped 3-digit SIC 1968 to 2-digit and 1-digit SIC 1980. Specifically - using the original nomenclature for each SIC aggregation level, nomenclatures that confusingly differ from those used in later SIC classifications - I have mapped each 3-digit "Minimum List Heading" (the lowest level of aggregation available for SIC68, corresponding to Group level in later SIC classifications) to a single SIC 1980 2-digit "Class" or SIC 1980 1-digit "Division" (where in terms of aggregation SIC80 "Classes" correspond to Divisions of later classifications, and SIC80 "Divisions" to later classifications' Sections or Sectors).

• SIC68 Minimum List Heading (3-digit) to SIC80 Class (2-digit) one-to-one mapping, implementing a recode in a STATA do-file.
• SIC68 Minimum List Heading (3-digit) to SIC80 Division (1-digit) one-to-one mapping, implementing a recode in a STATA do-file.
• Excel spreadsheet containing SIC68-SIC80 correspondences and giving information about the one-to-one mappings.