Response to the Consultation on Access to Business Register and Employment Survey (BRES)

Response to the Consultation on access to Business Register and Employment Survey (BRES) estimates on Nomis

Response to the consultation on access to Business Register and Employment Survey (BRES) estimates on Nomis

November2016

Background

The Office for National Statistics (ONS) has undertaken a review of how disclosive data are disseminated across the organisation, as well as the statistical disclosure control methods applied to business survey outputs.

As detailed levels of aggregation for the Business Register and Employment Survey (BRES) are made available, some of the information may potentially be disclosive. Currently those wishing to access this data need to apply for a Chancellor of the Exchequer’s Notice (CEN). We havedecided that this process can no longer continue and potentially disclosive estimates can no longer be put into the public domain in this manner.

Weproposed having 2different levels of access to BRES data on Nomis (public access and safeguarded), both of which involved the use of rounding to address issues of confidentiality. While the safeguarded estimates will give more accuracy, being rounded to a lower base, users will still be required to sign up to certain conditions regarding their use of the data.

This rounding of data was to be introduced in September 2016 when the latest set of BRES estimates is due for publication. It was proposed the use of CENs will also cease in September 2016.

The initial consultation was published on 24 February 2016, both on our website and Nomis. In addition, all CEN holders were directly emailed a copy of the consultation document by Nomis. The consultation document was also debated at the meeting of the Central and Local Government Information Partnership (CLiP) held on 3May 2016 in London. The consultation itself ran from 24 February 2016 to 5 April 2016, although as we were keen to listen to as much user feedback as possible comments from the CLiP meeting were also considered.

2. Responses received to the consultation

Responses received directly from users

We received 23 responses into the proposed new arrangements for obtaining BRES microdata.

There were 6 from central government, 10from local government and 7from the private sector.

Some were focused on issues of clarification about the new arrangements only.

The main messages coming out from the responses were the following:

iMany users expressed concern that the introduction of rounding to the BRES estimates would lead to a reduction in the usefulness of the data due to a fall in quality. One spoke of a “significant detrimental impact on the scope and the quality of analysis that the team produces”. Another stated that they were already aware of the disclosure aspects and were taking responsible precautions against release of disclosive results.Moreover, there was a fear that the rounding could introduce “substantial error” into the calculations and hence undermine the results obtained from user analysis. The same user was worried that the way the rounding was implemented may mean that more severe rounding from 1part of the dataset would be mixed with less severe from another and lead to a situation which was complicated and “the worst of both worlds”.Although 1of the users felt that the rounding proposal would probably not undermine quality too much, there was a request from that user that information on the impact of quality be released. Finally, a user expressed relief at not having to be concerned about disclosure any more but had lingering worries over a potential loss in quality that rounding brought.There were some concerns that aggregated estimates would be obtained from rounded lower-level data. This is not the case, since aggregates will be derived from unrounded lower-level estimates and then rounded after the aggregation has taken place.

ii The UK Commission for Employment and Skills (UKCES) collaborates with Warwick University to produce the Working Futures dataset.Since BRES data were used in the model for all years, there was some concern over the potential need to destroy past data and replace the inputs with the new safeguarded estimates, and if this were not the case would ongoing distribution to users of Working Futures be allowed.It has been agreed that this dataset should be allowed to use the existing data in the past but only safeguarded data going forward.This principle will be followed for any such models making use of BRES data for years prior to the introduction of the new arrangements.

iiiMost users felt that there was little or no benefit to them in the release of the open access dataset.

However, it should be noted that the respondents were all keen UK users of the unrounded dataset. More occasional users will almost certainly benefit from this, including those from overseas who were previously excluded from this kind of access.

iv There were general clarification requests on how the new arrangements will be implemented, which emphasises the need for clarity in all communications on this matter.

Responses received from CLiP

Local authority representatives voiced disquiet regarding BRES users no longer having access to unrounded estimates, should rounding be introduced. The loss of this granular detail was of particular concern to local authority planning departments, who felt it represented a loss of data quality and integrity. CLiP representatives asked, given rounding had to be introduced, if the proposed level of rounding could be revisited and possibly reduced.

Although unrounded disclosive aggregate BRES data would no longer be available via Nomis, CLiP members accepted that access to the actual unrounded BRES microdata would still be possible via the Virtual Microdata Laboratory (VML). To gain access, prospective users will have to make an application to the Microdata Release Panel (MRP), explaining the reasons for requesting the data and enabling the panel to check the capability of the researcher and understand the research project. It was noted that the MRP only meets once a month, so there may be a delay in the timing of the application and access being granted. It was also noted that local authorities would need a Public Services Network(PSN) or Government Secure Intranet(GSI) connection to access the VML.CLiP members asked if it were possible to delay the removal of the notice level unrounded estimates on Nomis planned for September 2016 while access to a PSN or GSI connection was confirmed or arranged.

The overall summary of user views obtained from the consultation was that:

• there were concerns over the loss of the detailed unrounded data and its impact on quality
• there were concerns on exactly how the rounding of estimates would be implemented, including the stage at which it occurs
• there were questions about how, exactly, the new arrangements would operate

Conclusion

While we note the serious concerns about the loss of the unrounded dataset, it is felt that the present way of dissemination of detailed BRES data cannot continue due to the potential for release of disclosive data, and the need to bring BRES into line with the rest of the department in the release of disaggregated data.

Many of the users who responded to the consultation would be able to obtain the unrounded dataset from the Virtual Microdata Laboratory (VML). These users would be those from the public sector and academics, provided certain criteria are abided by. We are willing to point such users in the right direction so that this can be achieved.

We have reviewed the rounding plans in the light of concerns expressed. Our methodologists have come up with revised rounding rules which are designed to maintain quality of estimates, even at detailed levels, and these should be reassuring to users. Full details of the new rounding rules are shown in Annex A.

We are grateful for the comments received from users and are keen to help users adjust to the new arrangements. In recognition that this is a big change for users of BRES data, we have agreed to extend the application of the Chancellor’s Notice method of release of data for 1 more year; the release of data by this method in September 2016 will be the last time that it occurs.

Annex A

This note outlines how access to BRES data changes in September 2016 and provides details on rounding rules to be implemented.

Implementing Public, Safeguarded and Notice access

1. Overview

Currently users have access to 2BRES datasets on Nomis:

• Business Register and Employment Survey (2009 to 2014)

• Business Register and Employment Survey public/private sector data (2009 to 2014)

A third BRES dataset is available, Business Register and Employment Survey 2010 consistent with earlier years. For the purposes of this year's change, we consider this as a historical dataset so propose that it is left unchanged at this time and reviewed when the changes are applied to historical datasets.

When new 2015 and revised 2014 estimates are released on 28 September 2016,2changes will be made:

• introduce a new public access BRES dataset

• the existing BRES dataset will be amended to allow users to request outputs from it to be disclosure controlled to either Notice, Safeguarded or Open access levels

2. Open access

The new open access dataset will offer users identical data selection options (such as geographies, industries) as the current restricted access dataset but outputs will be rounded, as detailed in section 4 below. The differences users will see are:

• on dataset access, no page requiring users to select a Notice and identify themselves
• outputs will be rounded;no flagging of confidential cells as all outputs will be public

The name of the new dataset will be:Business Register and Employment Survey: open access.

By referring to the dataset as "open access" rather than "public access" we will avoid confusion with use of the word ”public” in the existing Business Register and Employment Survey public/private sector dataset. This dataset is already open access.

3. Restricted access

The existing BRES dataset name will be changed by appending the text “restricted access”:Business Register and Employment Survey: restricted access.

Users obtaining a Notice will get access to this restricted access dataset as they do under the current system. The dataset would be modified to include a new dimension called access level. This year there will be no separate application process for just Safeguarded level of access. The purpose of providing Safeguarded access at this point is only to allow users to evaluate the impact of getting outputs at that level on their work, ahead of the withdrawal of Notice level access in 2017.

Figure 1 illustrates how the new access level option will appear.

The 3levels will allow users to choose whether output is unrounded (Notice level) or rounded to Safeguarded orPublic levels. By selecting the Public access option users would get the same outputs as produced from the public access dataset. The access level options will be implemented as radio buttons, so that users can only choose 1access level at a time. Outputs will be clearly flagged to indicate the level of rounding and disclosure restrictions that apply.

By implementing the change in this way, the process of sharing data becomes much simpler and less liable to accidental disclosure, as users should no longer need to do their own secondary disclosure control. Users can request Notice or Safeguarded level data for their own purposes then if they wish to share this data they can simply change the access level option to Public to get a version which they can share.

Following the introduction of the safeguarded only system (September 2017), users will continue to use the same dataset. After the changeover, the access level option will have the Notice level removed so users only see the Safeguarded and Public levels.

4. Rounding methodology

Size of estimate / Public access / Safeguarded (EUL) access
Non-disclosive cell / Disclosive cell / Non-disclosive cell / Disclosive cell
1 to 49 / 5 / 10 / 5 / 5
50 to 249 / 25 / 50 / 10 / 25
250 to 499 / 50 / 100 / 25 / 50
500 to 999 / 100 / 200 / 50 / 100
1,000 to 2,499 / 250 / 500 / 125 / 250
2,500 to 4,999 / 500 / 1,000 / 250 / 500
5,000 + / 1,000 / 2,000 / 500 / 1,000

Relaxed rounding will be applied to non-disclosive cells, which are defined as cells containing at least 3 enterprises and no dominant contributors.

Dominance will be calculated using the current methodology, which is at least 3 enterprise groups in a cell and the total of the cell less the 2 largest local units must be greater than or equal to 10% of the value of the largest local unit.

5. Percentages

In the existing BRES dataset, percentages (that is, percentage of jobs in a particular industry) are always calculated using unrounded data and output to 1 decimal place.

Following the update in September, the required method is to calculate percentages from the figures as they appear in outputs. For Notice level outputs this means percentages will continue to be based on the unrounded outputs, but for Safeguarded and Public access outputs percentages will be calculated from the rounded data.

Where a user requests percentages in outputs at Safeguarded or Public level, a new footnote will need to be added to make clear that the percentages are calculated based on rounded values.

We may also want to advise users that this may also mean that percentages for the same figure may differ depending on whether they are looking at Notice, Safeguarded or Public access figures.

Wehave given careful consideration to the views expressed by users of BRES estimates on Nomisandhave decided that it would be beneficial if users are effectively given “a year’s grace” before the rounding of BRES estimates is enforced. In addition, the level of rounding to be applied to the BRES estimates has been reviewed, particularly with regard to where the size of the estimate is below 500.

When users access the BRES estimates on Nomis following the publication of new and revised data on 28 September 2016 they will be able to access3forms of the data:

• continued access to unrounded BRES estimates (2009 to2015) via the usual Chancellor of the Exchequer's Notice (CEN) application procedure and agreeing the associated Data Access Agreement (DAA)
• access to a new “open access” rounded BRES dataset (2009 to 2015) that will be freely available to all on a free-to-view basis
• access to a new “safeguarded” rounded BRES dataset (2009 to 2015) that will be available to all users who apply via the usual CEN application procedure and the associated DAA;this dataset can only be accessed via the CEN application procedure and as such will be “linked” to applying for unrounded estimates

It is hoped this change will give users both a “year’s grace” before the switch to just “open access” and “safeguarded” rounded BRES datasets (which will be implemented in September 2017) and time to explore obtaining access to the BRES microdata help on the Virtual Microdata Laboratory(VML) should they wish to do so.

Taking into consideration users concerns regarding the perceived loss of unrounded granular detail, we will amend the rounding boundaries in order to provide a greater degree of detail at lower-level estimates. The changes to the rounding can be seen in Tables 1 and 2:

Table 1: Originally proposed rounding boundaries

Size of estimate / Rounding in public access dataset / Rounding in safeguarded dataset
Below 500 / Round values to the nearest 100 / Round values to the nearest 50
Above and including 500 and below 1,000 / Round values to the nearest 200 / Round values to the nearest 100
Above and including 1,000 and below 2,500 / Round values to the nearest 500 / Round values to the nearest 250
Above and including 2,500 / Round values to the nearest 1,000
In addition, a very small number of cells above 5,000 may be suppressed where it is felt that additional protection is required / Round values to the nearest 500
In addition, a very small number of cells above 5,000 may be suppressed where it is felt that additional protection is required

Table 2: Amendedrounding boundaries

Size of estimate / Rounding in public access dataset / Rounding in safeguarded dataset
Below 50 / Round values to the nearest 10 / Round values to the nearest 5
Above and including 50 and below 250 / Round values to the nearest 50 / Round values to the nearest 25
Above and including 250 and below 500 / Round values to the nearest 100 / Round values to the nearest 50
Above and including 500 and below 1,000 / Round values to the nearest 200 / Round values to the nearest 100
Above and including 1,000 and below 2,500 / Round values to the nearest 500 / Round values to the nearest 250
Above and including 2,500 / Round values to the nearest 1,000
In addition, a very small number of cells above 5,000 may be suppressed where it is felt that additional protection is required / Round values to the nearest 500
In addition, a very small number of cells above 5,000 may be suppressed where it is felt that additional protection is required

In addition, it has been decided that for the 3surveys prior to BRES –the Annual Business Inquiry (ABI) (1998 to 2008), the Annual Employment Survey (1991 to 1997) and the Census of Employment (1971 to 1991) –rounding will notbe applied in September 2016 and access will continue to be via the usual CEN application procedure only. It is intended that, for now, rounding and open or safeguarded access will be applied at some point after the main BRES update in September 2016.

The pre-BRES datasets are all at least 8 years old, in the case of ABI, and 19 or more years old, in the case of the other 2. However, a dominant business in a particular sector might have been around for a significant length of time and be more visible, in comparison to individuals and households information where the data age offers considerable protection. For this reason rounding will need to be applied at some point before the loss of the CEN application procedure in September 2017.

Annex B: Organisations which responded to the consultation

Central government

UK Commission for Employment and Skills (UKCES)

Scottish government

Welsh government