ONS has released interactive data visualisation tools to display the flow of commuters in local authorities across Great Britain. The data used to produce these tools come from the Annual Population Survey (APS). This document contains more information about the APS as well as some of their uses and limitations.
ONS has produced, for 2010 and 2011, three interactive data visualisation tools to help analyse commuting flows for local authorities within England, Wales and Scotland:
Commuter flows mapping tool;
Spider map of local authority commuter flows1;
Bar chart of local authority commuter flows.
These tools were produced using data from the Annual Population Survey (APS), a survey carried out by ONS on a sample of UK residents. The APS includes data from the Labour Force Survey2 as well as an additional boost sample (known as the local labour force survey) of addresses in England, Wales and Scotland. The sample aims to meet a target number of economically active adults at the local authority level within England, Wales and Scotland3.
Patterns of commuting are typically analysed using census origin-destination data using the following information:
Where an individual lives;
The address of that person's main job.
APS data provide a very rich source of information on commuting flows for all workers4. Other data sources providing information about commuting flows, such as census data, are analysed to create Travel-to-Work Areas (TTWAs) which provide geographic areas which are largely self-contained - ie that the majority of people who live in the area also work in the area, and the majority of jobs in the area are filled by people who live in the area5. The major drawback with using census data for analysing commuting patterns is that they become out of date.
The advantage of using APS data as a source for an update on commuting patterns is that the data are refreshed annually. The APS asks a sample of households, questions on their working patterns. The next section explores those questions asked in the APS, and considers sample sizes to determine whether it could be used to update commuting flows.
The Annual Population Survey (APS) is one of the UK’s largest surveys comprising responses from two waves1 of the Labour Force Survey (LFS), and local sample size enhancements in England, Wales and Scotland. Annual sample sizes are approximately 300,000 which is more than 1% of the working population.
The address of residence for all APS respondents is known precisely because this is a household survey and therefore the address of the household must be known in order for it to be sampled. The address of main place of work is determined by the responses to the questions:
What city, town or village is your place of work in?
What county is that in?
This populates an automatic coding frame, where the interviewer is presented with a list of possible geographies based on the answers given to these questions2. Then, the interviewer, together with the respondent, can establish the correct workplace from which the local authority can then be derived3. So, in terms of the available information, the origin-destination data available from APS is:
Origin: postcode or any small statistical geography up to local authority or region;
Destination: local authority or region.
Respondents are sometimes unsure of which local authority their workplace is in, particularly for workplaces in London where specific local authorities are less well known. In such cases, the interviewer asks further probing questions such as whether the respondent knows which train or tube station is closest to their workplace, to try and identify which local authority is most appropriate.
The interactive data visualisation tools now available use Annual Population Survey data to look at levels of self-containment in local authorities with respect to their working and resident populations. These interactive data visualisation tools can primarily be used to help answer the following questions:
Do the working residents in a particular area also work in that area?
Do the people who work in a particular area also live there?
These two different concepts are called residence self-containment and workplace self-containment respectively.
Using these tools it may be possible to identify some surprising or unlikely commuting links between local authorities. Part of the explanation for these commuting links might be because of anomalous commuting flow estimates resulting from grossing survey data to the full population. This can arise from a comparatively small sample size for a given local authority. Another possible reason for unusual commuting links is that some survey respondents may have a second home, close to their place of work, while their main home (the home where the survey took place) is a long way from their place work. As such, each estimate of the flow of commuters is also presented with confidence intervals which provide an indication about the degree of uncertainty around an estimate. These confidence intervals are available in the data tables as part of this release and are also displayed on the interactive bar chart.
All sample surveys are subject to sampling error, which means the sample is only one possible subset of those eligible for the survey. If a different sample was selected it would undoubtedly provide slightly different results. Estimates provided in the Commute-APS data visualisation tools are potentially subject to sampling error due to small sample sizes. For this reason, the counts data for the Isles of Scilly local authority have been removed from the commuter flows mapping tool. To aid in the interpretation of results confidence intervals have been provided in the interactive bar chart of local authority flows tool.
Additionally, data in these tools are subject to non-sampling error. There are a number of types of non-sampling but the two that are of key importance to these products are errors due to respondents being unable to answer a question accurately and error due to mistakes in the data collection process (miscoding of respondents’ answers). It is known that the Annual Population Survey (APS) routinely overestimates the number of people who work in the City of London (and other key urban areas) as respondents incorrectly report that they work there when in fact they work in neighbouring areas, the problem is especially acute when the respondent is taking part in the survey on behalf of another household member. It is also likely that some data collection errors occur and those workplaces are occasionally incorrectly coded.
The Commute APS products do not feature statistical tests to determine whether a local authority’s estimated commuting flow is significantly greater or smaller than another estimated flow. However, it is possible to perform such calculations from the raw data. One method of doing this is to carry out a t-test which assesses whether the means of two groups are significantly different from each other when taking their variability into account, as follows:
X1 = mean of local authority one;
X2 = mean of local authority two;
n1 = number of observations in sample of local authority one;
n2 = number of observations in sample of local authority two;
s1 2 = variance around the estimated commuting flow for local authority one;
s2 2 = variance around the estimated commuting flow for local authority two.
It is also possible to ‘eyeball’ whether or not the flows of two local authorities to or from a selected local authority are significantly different to one another. If the lower confidence limit of the local authority with the larger proportion of commuters (local authority one) is higher the then upper confidence limit of the local authority with the smaller proportion of commuters (local authority two), then it would be expected that local authority two had a higher proportion of commuters to or from the selected local authority than local authority two at least 95 times out of 100.
For the Labour Force Survey in general, the smallest estimates which are considered reliable are based on a criterion of a maximum of 20 per cent coefficient of variation (the standard error as a percentage of the estimate)1. However, the data used for the Commute-APS products here do include estimates with a greater than 20% coefficient of variation because removing such instances would adversely affect the presentation of the data visualisation. As such, care should be taken when drawing conclusions from some of the smaller commuting flows estimates, the uncertainty around which may be too large to make meaningful conclusions.
Details of the policy governing the release of new data are available by visiting www.statisticsauthority.gov.uk/assessment/code-of-practice/index.html or from the Media Relations Office email: email@example.com