You are in: Home › Census › 2001 Census › A review › Data quality › Census Variables › Results and Conclusions ›

2001 Census Variables

4. Results and Conclusions

Many Census items are well answered with little requirement to edit values supplied by form-fillers or impute values when not supplied.  For example, discounting records added by the One Number Census (ONC), 0.35 per cent of values of sex were not as collected on a Census form.  If ONC cases are included, then 4.3 per cent of records had a value for sex that were not as collected on a Census form.  At the other extreme, the equivalent figure (excluding ONC cases) was over 20 per cent for the following cases.

• The indicator of whether a person had a place of work or study.
  26 per cent of persons finally classified as 'not working or studying'; were not classified as such on the form. There were lower rates (6 and 13 per cent respectively) for persons classified as travelling to work or study. Including ONC cases gives rates of missingness of 30, 10 and 18 per cent respectively.
  
  
• The destination of persons travelling to place of study.
  This rate was 23 per cent (28 per cent including ONC cases). The equivalent figures for travellers to work were 14 and 17 per cent.
  
  
• Whether the person ever worked.
  This rate was 20 per cent (24 per cent including ONC cases). 

Full details are in the tables below for household items, relationships and person items. The tables consist of summary rows taken from each of the linked reports for individual variables.

Appendix A - Household Reports and Tables

Appendix B - Relationship Reports and Tables

Appendix C - Person Items Reports and Tables P2 to P9

Appendix D - Person Items Reports and Tables P10 to P17

Appendix E - Person Items Reports and Tables P19 to P34

To assess the effectiveness of Imputation would entail follow up work with persons affected to get a 'true' response to compare with the imputed one. This has not been done. Instead the final distribution of each item has been compared with data from other sources and generally found to match given the characteristics (e.g. survey non response) of the source. These results show that Imputation is generally needed where you would expect it. They (with any future work) also show where - on what categories of household or person - efforts should be concentrated in order to minimise the need to impute.