Modelling Scenarios using the Witan SEND Model

Published in

Mastodon C

8 min readSep 5, 2018

When developing the Witan SEND model we talked to local authorities (LAs) about what was most important to them when doing projections for their SEND population. One of the things we kept hearing was the need to be able to model “alternative scenarios”.

Being able to project the future SEND population and the cost to serve that population is really valuable to LAs, but budgets and capacity are limited, and if a projection indicates that future demand will exceed those budgets and capacity, then the Witan SEND model can show you where the gaps are and hence begin to shape potential alternative scenarios.

Working with LAs, we are able to define and test their scenario ideas, which can then lead to policy change or other critical decisions being made (for example an extension to a school which caters to a specific need).

In the Witan SEND model a person is described by their ‘state’, which is the combination of their primary ‘need’ and the ‘setting’ in which they receive education and support. Each setting is available for one or more year group. At the beginning of each academic year a person will move from one setting year group to another setting year group. This move is described as a ‘transition’. People who join the total SEND population we call ‘joiners’, and those leaving, ‘leavers’. People moving from one setting year group to another setting year group we call ‘movers’.

A simple illustration showing only two settings, A and B, and the transitions between them, is shown above.

By default the Witan SEND model uses historic data about transitions between states to model future transitions that may occur. This is our ‘baseline projection’. However, policy or service provision changes may mean that historic transitions are not representative of expected future transitions — we describe projections based on different expected patterns of transitions as ‘scenarios’.

A common scenario LAs wish to test is what would happen if people with new EHCPs were directed to alternative settings. For example individuals with social, emotional and mental health needs (SEMH) in Year 7 may typically be placed in setting A. It may turn out that setting A is not the optimal setting for SEMH, either because it is not cost effective, because it doesn’t lead to the best outcomes for pupils, or perhaps because the authority simply doesn’t have enough capacity for those pupils. We can then test what the future population and cost profile would look like if we were to divert a certain proportion of new individuals previously joining setting A at Year 7, instead to join setting B.

The above example is a relatively straightforward scenario for us to simulate (see 2b). Below you will find a list of the types of scenarios we can currently simulate. These can also be combined in all sorts of ways to test a wide variety of more complex possible scenarios — some further examples of combining scenarios are outlined later.

1. “Ignore historic data before a specific calendar year for an age group”

By defining a calendar year and a starting national curriculum year (NCY) you can set the model to ignore historic data for that age group, from a specific time
Example parameters may be “2016” and “NCY 11”
This would mean that transitions data for NCY 11 or older are not taken into account for years before 2016, and hence not included when calculating transition rates. Transition rates would instead be based only on data for 2017 and 2018
This may be used if a recent policy change misrepresents the current trends in the data

2a. “Modify a setting’s joiner transition rate”

By defining a setting to modify and a value to modify it by, you can directly change the current rate of joiners to a specific setting
Example parameters may be “Special Independent” and “1.2”
This would modify the joiner rate to “Special Independent” setting to increase it by 20%
This may be used if you wish to test an historically unobserved change in the rate people join a specific setting

2b. “Modify a setting’s joiner transition rate and transfer those joiners to another setting”

As above, however take that proportion now not expected to join a setting and instead increase joiners to another setting
This changes two settings joiner rates proportionately
Example parameters may be “Special Independent”, “Mainstream” and “0.5”
This would halve the joiner rate to “Special Independent” and instead allow that proportion of individuals to join “Mainstream”
This may be used to test scenarios whereby new EHCPs for are directed towards a different setting from that which is historically seen

3. “Modify a setting’s leaver transition rates”

By defining a setting to modify and a value to modify it by, you can directly change the current rate of leavers from a specific setting
Example parameters may be “Mainstream” and “0.3”
This would modify the leaver rate to “Mainstream” setting by a third
This may be used if you wish to test an historically unobserved change in the rate people leave a specific setting, for example students staying on longer in further education since the upper age limit changed to 25

4a. “Modify a setting’s transition rates to other settings”

By defining a setting to modify and a value to modify it by, you can directly change the current rate of movers to a specific setting
Example parameters may be “Further Education” and “0.1”
This would modify the mover rate to “Further Education” setting by 10%
This may be used if you wish to test an historically unobserved change in the rate people move to a specific setting

4b. “Modify a setting’s transition rates to other settings and transfer those movers to another setting”

As above, however take that proportion now not expected to move and instead increase their movement to another setting
This changes two settings mover rates proportionately
Example parameters may be “Special Independent”, “Mainstream” and “0.5”
This would halve the mover rate to “Special Independent” and instead allow that proportion of individuals to move to “Mainstream”
This may be used to test scenarios whereby current transitions betweens settings are directed towards a different setting from that which is historically seen

5a. “Modify a setting’s transition rates from other settings”

By defining a setting to modify and a value to modify it by, you can directly change the current rate of movers from a specific setting
Example parameters may be “Special School” and “0.2”
This would modify the mover rate from “Special School” setting to others by 20%
This may be used if you wish to test an historically unobserved change in the rate people move from a specific setting

5b. “Modify a setting’s transition rates from other settings and transfer those movers to another setting”

As above, however take that proportion now not expected to move and instead increase their movement from another setting
This changes two settings mover rates proportionately
Example parameters may be “Special Independent”, “Mainstream” and “0.5”
This would halve the mover rate from “Special Independent” and instead allow that proportion of individuals to move from “Mainstream”
This may be used to test scenarios whereby new EHCPs for are directed towards a different setting from that which is historically seen

6. “Modify transitions from a specific future calendar year”

This scenario is used in conjunction with one or more of scenarios 2–5
By defining a year to begin using the alternative transition rates, you can maintain the current historic trends until a specific year
An example parameter would be “2020”
This would only apply the new transition rates (changes to setting(s) and a type of transition (i.e. joiners)) from 2020, if this is when you expect to make that policy change

Examples of combining scenarios:

“Modify multiple settings joiner transition rates” — modify joiners rates to a number of settings by the same value (i.e half them all)
“Modify both the joiner and transitions to a settings rate” — modify the rate individuals join and move to a setting from other settings by the same value
“Modify multiple settings joiner rates and transfer them all to a single setting” — modify joiner rates of a number of setting and transfer those expected individuals to another setting
“Modify multiple settings transitions rates and transfer them to multiple specific setting — modify joiner, leaver and mover rates to and from multiple settings, anad transfer them elsewhere all at the same new rate

Evaluating an alternative scenario

Finally with our alternative scenario and baseline run (the default model behaviour) we are able to compare the two possible futures to evaluate whether the proposed scenario or policy change is likely to be effective in achieving your goals.

Blue indicates the default scenario, whilst in yellow is the alternative scenario, showing how the total population may change

This can be repeated any number of times to model differences between alternative scenarios. For example if you wanted to test multiple different levels of modification on any rate — to explore the overall range of possible impacts, modify any of the rates above by e.g. 10%, 20%, 30%, 40%, and so on, to understand at what level you would achieve budget or capacity goals, and let you assess the realism of those goals.

Glossary

Need — a description of an individual’s specific need or disability

Setting — the educational placement an individual receives to accommodate their need(s)

Transition — a movement between states, inferred from the rate of historic transitions between states

State — a combination of need and setting to describe an individual’s current situation

Joiner — an individual who has joined SEND, with a described need and setting

Leaver — an individual who has left SEND

Mover — an individual who “stays” in SEND. Every calendar year their state will change or stay the same

SEND — Special Educational Needs and Disabilities

LA — Local Authority

EHCP — Education, Health and Care Plan

SEMH — Social, Emotional and Mental Health needs