A brief guide to choosing your models in Witan

With Witan, you can create population projections by feeding development data into a model, to get some output you can reuse for your own analysis. However, which model should you use? Below, Ben Corr, the GLA’s head demographer, describes when you might use one model over another, and explains some of the other options available to you:

All projections use a common ‘ward model’ to distribute borough-level population between wards.  The ward model is based on cohort component methodology.  Annual migration estimates are not available at this geographic level, so proxy flows are generated based on the housing trajectory and census data.  The impact of development in a particular ward is dependent upon the characteristics of migrants to and from the ward and the historic ratio of adults per dwelling.

Trend-based Ward Population Projection Model

This model uses projections of the overall borough-level population from the GLA’s trend-based cohort-component model.  The overall population is independent of the assumed housing trajectory.  Ward level projections are constrained to match the overall borough-level population, but the distribution of population between the wards is determined by the housing data input by the user.

This model is most useful in areas where recent population change has occurred largely independently of changes in available housing stock.  In areas where recent population growth outstrips planned housing development, the model results will imply increasing household size.

Housing linked Ward Population Projection Model

Two variants of model are available to produce population projections linked to the user’s chosen housing trajectory.  They differ in how borough-level populations are generated, both use the ward model to distribute these between wards.

DCLG variant

This model adjusts domestic migration assumptions until it arrives at a borough-level population that fits the available dwelling stock.  The model’s name derives from the use of relationships from DCLG’s household projections to convert a candidate population into a number of households.  The model estimates the capacity for households based on the housing trajectory input by the user and the historic relationship between households and dwellings.

This model is useful in areas where housing development is expected to be the predominant driver in future population change.  Projections produced by this model will usually imply falling household size.  This results from an increasing proportion of older people in the population and an assumption that older people tend to form smaller households.

Capped Household Size variant

This model attempts to account for both recent trends in population change and future changes in housing stock.

It operates on the following principles: if housing capacity is increasing faster than recent population growth, then housing drives growth as it does in the DCLG-based model;

if recent trends imply growth above available capacity, population growth is limited by available housing stock though average household sizes do not fall as they would in the DCLG-based model.  Average household sizes are prevented from rising above recent levels.

This is the GLA’s general use model and provides more intuitive results across a wider range of scenarios than either the trend-based model or DCLG variant of the housing linked model.

Which fertility assumption?

The models are initiated with fertility rates for each borough and ward based on recent birth estimates.  These are then projected forward using trends taken from the 2014-based National Population projections for England.  The Low, Standard, and High options correspond to the trends from the Low, Principal, and High fertility variants of the national projections.

• Standard – fertility relative to 2014 rises by 4% over the next decade before levelling off
• Low – fertility relative to 2014 falls by 6.5% over the next decade before levelling off.
• High – fertility relative to 2014 rises by 13% over the next decade and then rises at a slower rate to reach a level 15% higher by 2040.

Further questions

If you have further questions when using Witan to generate projections, you can always send an email to witan@mastodonc.com, or use the in-app help system to speak to the support team  – just click the blue speech bubble in the bottom right of any screen when logged in.

 

 

 

 

Bulletin number 7

Hello, this is Sunny Townsend and Chris Adams at Mastodon C, with a few updates on the Witan project.

Generating projections for the SRP

Witan is now ready to generate population projections for the GLA to run their School Roll Projections (SRP) model.

To help guide you we have put together a few pointers:

• If you are using Witan for the first time or need some reminders, there is lots of help in the user guide and the bulletin archives (password ‘london’).
• If you have questions about which inputs or model variant to choose, try reading this bulletin.
• If you have further questions about which model to choose then please contact the GLA demography team (demography@london.gov.uk).
• For questions around validation or referencing, try this bulletin for answers. The inputs are now updated to those that will be used in the GLA’s 2015 round.

Once you are happy with your projections, download the csv called ‘Ward-population-projection-split-gender’ and include it when submitting your roll data.

You can always reach us at witan@mastodonc.com – if we don’t have the answer right there, we’re in daily contact with the GLA, so we’ll let you know if we need to check something and usually get back to you the same day.

If you forget your password, you can now automatically reset

If you’ve been using Witan, and forgot your password, you can reset your account manually instead of needing to contact Witan directly.

Just hit ‘forgotten password’ on the front page when visiting witan-alpha.mastodonc.com, and you’ll be able to reset your password to get back to making projections.

A question we’d like some help with – capped household size

One question we’ve had come up is whether users can tweak the capped household size for their borough. We don’t offer this as something that can be changed at present, but it’s something we’re considering in future – is this something you can imagine using? If so, we’d like to speak to you.

Send an email to witan@mastodonc.com, expressing interest, and we’ll follow up later to speak directly with you to get a better idea of how to make this useful to you.

Until next time,
Chris & Sunny

Bulletin number 6

Hello, this is Sunny Townsend at Mastodon C, with a few updates on the Witan project.

 

Validated models: you can now use the numbers from Witan

The GLA have validated the models and and can confirm that the outputs are consistent with those produced by the GLA in previous rounds.

Even so, we encourage users to make comparisons of newly generated projections against other datasets such as previously supplied BPO projections, published SHLAA based projections and ONS data.

In addition, in situations where wards contain large scale development, we would encourage further comparison with alternative approaches such as the GLA population yield calculator.

If  you have questions about using the Witan projections, you can always reach us at witan@mastodonc.com – if we don’t have the answer right there, we’re in daily contact with the GLA, so we’ll let you know if we need to check something and usually get back to you the same day.

 

Referencing your projections

With access to Witan, you can generate you own projections at any time of year.

We need to make you aware though, that projections will not fit neatly into the GLA’s round system as the input data will be updated over the course of the year.

The inputs will be updated to those that will be used in the GLA’s 2015 round in the next 2 weeks, in preparation for the school roll projections.

We’re currently working on guidance for how users might most appropriately reference projections produced in Witan and are keen to get your input on this.

 

Your projections now have by-gender breakdowns

You are now able to download projections results split by gender. This is provided as an additional output when you generate projections.

The gender-split projections are currently provided in a format that the GLA would like them for the school roll projections.

Is this format useful? Are there other outputs that you would value?

 

New features for Witan – what is valuable to you?

In the last bulletin we mentioned we had a few requests for by-gender breakdown for population projections, and it should be available to you when you next sign into Witan.

We’re hoping to continue building features users want, so have a couple more feature requests which we’d like to share to get an idea of which are the most useful to build:

• demographic projection breakdown by ethnicity
• population yield calculations from large scale developments

If any of these were available, would they help you carry out a task you already do? If so, how? Or would they make something possible that you currently can’t do?

We can’t guarantee we’ll build every feature, but if we see people requesting one feature repeatedly much more than another, it gives us a good idea of how to make Witan useful to as many people as possible, and we’re more likely to build that one first.

 

Using Witan to provide data for School Roll Projections

We’re sending out the next bulletin in mid March, in which you will receive instructions for how to use Witan to generate the data the GLA need to run your school roll projections.

Til then, we’ll be here, so if you have questions about using Witan, or know someone in your borough who would find it useful, you can reach us at witan@mastodonc.com.

Have a good week-end!
Sunny