The spreadsheet contains data from 13 local authorities (LAs) in the West Midlands: Birmingham, Coventry, Dudley, Herefordshire, Sandwell, Solihull, Staffordshire, Stoke, Telford, Walsall, Warwickshire, Wolverhampton and Worcestershire. Initially, data was also provided by Rutland but was not included in the study due to very small numbers. The table reports on data routinely collected by LAs about children who were on a CPP or were LAC on 31st March 2012.  LAs were asked to provide the following anonymised data for each child:  postcode (of family home or from which they entered care), age or date of birth, gender, ethnicity, whether disabled or not, reason for being on a CPP, legal status if LAC and whether they were unaccompanied asylum seekers. Geographical data is presented at MSOA level; MSOA refers to Middle Layer Super Output Areas.  MSOA and LSOA (Lower Super Output Areas)  were originally released in 2004 for England and Wales to improve the reporting of small area statistics, all of which have a unique nine character code in line with all statistical geographies provided by the ONS (ONS, 2014).

Data cleaning

Data was asked for approximately one year after LAs had submitted this routine data which has to be recreated for the purpose of request.  Once returned, some data had been corrected and in a couple of cases this had led to very substantial changes, particularly in the CIN numbers. LAs were asked to check and explain any differences when numbers varied by more than 2%. Responses to these inquiries demonstrated the difficulties in ensuring precise equivalence in data even when national criteria are used; a number of reasons for discrepancies emerged (see Bywaters et al., 2014).  Therefore, after the first stage of data cleaning, the largest differences between published and sample data were for CIN (approximately a 10% difference), and the large majority of that difference resulted from changes in one LA (a difference of around 3,000 cases). Due to a number of reasons, it is impossible to say whether the published data or our sample can be viewed as a more accurate account.   

During the second stage of data cleaning a number of issues were encountered when matching the LSOAs submitted against the LSOAs revised for the 2011 Census. Some LAs provided their returns using 2011 LSOAs; others were based on the previous boundaries established in 2001. Although the majority of LSOA boundaries had remained the same, some 2.6% nationally had changed.  Finally, due to four factors, some data was excluded from the study.  Firstly, 5.3% of children did not have postcodes or LSOA. Secondly some postcodes or LSOA given were outside of the LAs reporting the data which raised the concern that we might have been given LAC placement addresses rather than home addresses of origin. Thirdly, the date of birth was not given or was unknown for some children (542 children = 1.6%); in most cases, these were unborn children about whom child protection concerns had already been raised. We also found that different LAs reported over 18s in different ways; 2,970 of the cases concerned CIN over 17, but with highly varied rates between LAs. Finally there were 111 likely duplicate cases, mostly from a single LA and we assume these result from recording errors.  Therefore, the data detailed above was excluded from the analysis. This information has been taken from Bywaterset al., (2014); please refer to this paper for further detail.

Data presentation

The data presented in the table contains a total 696 MSOAs from the 13 LAs stated above.  LAs are listed in column A with the corresponding MSOA name in column B, and MSOA code in column C.  Column E shows the IMD (Index of Multiple Deprivation) decile for corresponding MSOAs.  The IMD is calculated by combining (and using appropriate weights) 38 separate indicators across 7 different domains and is calculated at LSOA level (DCLG, 2011).  The data is presented as numbers of children within each MSOA. The first columns of the table show the population of children in each MSOA, divided by gender, age and ethnicity, taken from the 2011 Census (columns F-AS). This enables you to calculate CIN, CPP and LAC rates per 10,000 children.  The table then breaks down the data by CIN (columns AU – DN), CPP (columns DP – GI) and LAC (columns GK – IF) numbers, again each split by gender, age and ethnicity.  The data when broken down by CIN, CPP and LAC numbers contains some ‘unknown’ ethnicity data, and for some CIN and CPP cases gender is ‘unknown’.  This means that in these particular instances no ethnicity or gender was stated in the data received from LAs.

Using the data

When using this data please acknowledge the research team and the Nuffield Foundation who commissioned the research.  The research team comprises Professor Paul Bywaters, Dr Geraldine Brady, Professor Tim Sparks and Elizabeth Bos, Coventry University.  Please contact Professor Paul Bywaters should you have any questions about the data:

We would be interested to hear about any use of the data whether for publication or not.



 Queen’s Award for Enterprise Logo
University of the year shortlisted
QS Five Star Rating 2020