Projecting transient populations Richard Cooper, Nottinghamshire County Council (Thanks also to Graham Gardner, Nottingham City Council) Background The work of the County and City Councils in Nottinghamshire includes the preparation of a Joint Structure Plan covering the former County Council area of Nottinghamshire (or Plan Area). Preparation of housing figures for the Plan is undertaken jointly, although projection work is done by the County Council, working closely with the City. The Joint Structure Plan was published in Deposit Draft form in November 2003, and an Examination-in-Public commenced in June this year. The housing figures for the Plan Area had already been set in Regional Planning Guidance for the East Midlands (RPG8), approved by the Secretary of State in January 2002. Work for this level of planning was done by the County Council again in conjunction with the City and other authorities across the Region, for the Regional LGA. The Joint Structure Plan had two basic questions to answer Should the level of housing (for the Plan Area) be changed from the 49,000 set in RPG8 because of changes in circumstances since being set? What should be the distribution below Plan Area level. i.e to districts? In response to the first question the County & City Councils decided that there should be no change to the overall level of housing, principally because National Planning Policy Guidance indicates that housing figures should not be revisited unless there is strong justification. While circumstances had changed and moved on, there had been no new household projections produced by ODPM, and more up-to-date information gave ambiguous indications as to whether the figures should be higher or lower. In distributing the housing provision below Plan Area level, demographic trends, house construction and economic prospects indicated a distribution to the 3 Sub- Areas, while below that levels of supply also directed what district totals should be. In Nottingham City s case, being surrounded by suburban districts (or land-locked ) supply was the principal factor, especially bearing in mind the Government s (and RPG s) exhortation to concentrate development in major urban areas. Thus Nottingham had a housing provision of 18,500, of which part (11,000 dwellings), up to 2011, was already committed in terms of local plan provision. Preparing population projections for Nottingham The preparation of population projections for Nottingham (and other districts) was led, therefore, by a known level of housing. In this regard preliminary projections using the Chelmer model s dwelling-led option were helpful. However, these early projections were inadequate for more detailed population projections because, (a) a dwelling led projection (using 5-year periods) produces widely fluctuating migration \\SERVER2\F_USERS\SHEPHERA\Projecting_transient_Cooper.doc 1
levels and (b) more information on the constituents of the projection became available as time went on. More importantly a reliable population projection was needed which the City Council and it s partners could use for planning services in Social Services, education, etc.. This would represent the best forecast of how the population would change, especially in terms of age structure, over the middle and long-term. Certain things are difficult, or impossible to model. In this instance, while we knew how many houses were likely to be built, the types of people and households to occupy them would most likely change from the patterns seen in the data in the model. The age of people, especially migrants, their socio-economic characteristics, likelihood to create or fragment households, even capacity to obtain housing, might change from the situation that pertained for the data used in the model. In particular, assumptions are inherent in the modelling process that the likelihood of people to migrate, in terms of age and gender will not change, and similarly the likelihood of people to create households, given their age, gender and relationship. What is the Transient population? If we look at the age /sex profile of Nottingham compared to other districts and the Plan Area we can see a distinctly higher population in the 20-29 year old population. Nottingham does have a younger population than the rest of the County, but the difference is much more than the attraction of young employed people into the City. In fact the large bulge seen in the Census data is principally students. In 2000, the two Nottingham universities had 29,500 full-time students. It is likely that approximately 24,000 of these had term-time addresses in the City, a net gain of about 21,000 university students. This population moves into and out of the City in a regular pattern, maintaining broadly the same age profile year-on-year (i.e. not ageing ) such as students. Such a population also tends to have different characteristics to the population as a whole, which may (or may not) be reflected in the data used in the model, for example, in fertility and mortality. This is the transient population. Fertility is affected significantly, as the transient population tends to be in the most fertile age groups. The introduction of a different transient population means that local fertility rates need changing to suit in order to generate realistic births. Correction factors are calculated automatically if birth data are put into the Chelmer model. For our purposes the transient population is principally students, but not entirely. It includes pupils, prisoners, and possibly the long-term ill, depending on whether they are in their district/area of residence. It does not include those in retirement homes; those people behave, as long-term residents of an area as large as Nottingham, even though their stay in a home may be short-term. The transient population is not simply institutional population, although ifor the 1991-based projections we identified those larger institutions, prisons, nurses/doctors and long-term residents in hospitals, where the population should be treated as transient. \\SERVER2\F_USERS\SHEPHERA\Projecting_transient_Cooper.doc 2
(Size of area and timescale has effect on what the transient population could be seen as. The smaller an area and the longer the timescale, the greater proportion of people are likely to be transient. A comparison of, say, Nottingham with England & Wales would illustrate this. There is a transient element in the England & Wales population, but my understanding is that (a) it s insignificant, (b) migration flows are assumed to deal with the short-term movements. Interestingly, Other potential transient populations are raising interest in regional housing policy, for example migrant workers, whether intra- or international, agricultural or international executives.) The impact of transient populations in the model If any sub-population that remains static because it is transient is not accounted for then it will be aged in the same way as the indigenous population. This is exacerbated in Nottingham s case (as with any area with a significant group) with the 20-24 population peak showing up 20 years older in 20 years time (see chart) in the same way as the small peak seen below in 30-34 year olds is older. Consequently the model needs to reflect the reality of that sub-population s behaviour. In real terms the population moves in and out on a regular basis, but historically (in 1981 & 1991 Census data) those movements have not been collected. Alternatively, the population can be removed from any ageing and migration, reflecting its constant status 1. This is possible in the Chelmer model by identifying a separate transient population. 35000 Comparison of 2021 populations using/ not using a transient population: Nottingham City 30000 25000 20000 15000 10000 5000 0 0-- 4 5-- 9 10-- 14 15-- 19 20-- 24 25-- 29 30-- 34 35-- 39 40-- 44 45-- 49 50-- 54 55-- 59 60-- 64 65-- 69 70-- 74 NCC transient pop 75-- 79 80-- 84 Zero transient pop. 85 + 2001 Population Using a zero transient population would give a resulting population for a set no of dwellings that is 8,000 (3%) fewer. In terms of population growth, that would be 30% 1 This issue was first raised by the County Council in the 1986-based Structure Plan, when the (County Council in-house) projections were first adjusted, in that case by amending the migration structure. \\SERVER2\F_USERS\SHEPHERA\Projecting_transient_Cooper.doc 3
lower. Conversely no transient population would produce 2,500 (in 18,000 about 13%) more dwellings for the same migration level (or population growth). More importantly the age structure of the population would be quite unrealistic. Why is this transient population a problem (to us?!) Up to now this is fairly conventional and well known to demographers who undertake projections. But why is this situation a problem, if we can obtain a working transient population, for the area? A decision was taken earlier in the planning for these projections that the Patient Register data available from ONS should be used to profile migration. Briefly this was done because it was more up-to-date, believed to be more complete than the Census (especially in terms of students), and we could obtain 3 years of data, rather than the single year (1990-1) of the Census. 2 The result gives a very different migration profile for Nottingham (see chart), with a much higher net in-migrant peak in the 15-19 age group. (Gross flows chart at end of paper). Comparison of 1997-2000 Patient Register and 1991 Census migration structure (M&F) 4000 3500 3000 2500 2000 1500 1000 ONS patient register 97-00 (3 years) 1991 Census 500 0-500 -1000 Total net migration: 1991 - -1,900 2000 - -2,200-1500 0-4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75+ The problem related how the migration flows of the transient population are captured in migration flows. How does this relate to what can or cannot be measured in the transient population? The 1991 Census migration did not include students, therefore using a transient population was appropriate for the 1991-based projections, where a transient population was created using student totals. (NB A study by Ludi Simpson did identify more out-flows than in-flows of student populations in the 1991 Census, which it might have been possible to account for). 2 At the seminar it was mentioned that ONS had recently published some research work looking at potential undercount of young male migrants in their patient register estimates of internal migration; these are available on the ONS web site: http://www.statistics.gov.uk/statbase/product.asp?vlnk=11252 \\SERVER2\F_USERS\SHEPHERA\Projecting_transient_Cooper.doc 4
The Chelmer model transient population takes account of all that population in a manner that can be applied to all districts consistently (see box). This gives a transient population of about 27,500, with a profile including older institutional age groups. This would be used in the no students Census migration. While the 1991 Census excluded (theoretically) all students, the ONS Patient Register data should include all but some students. The decision to use ONS PR had implications for the transient population; hence we needed to adjust the transient population to reflect what the ONS PR did in migration. Could this be done with the information available? Establishing what the transient population should be We compared the results of the Chelmer definition with what was known about students in Nottingham. Could it be improved? There were two possibilities - - by estimating students in another way; there is information available to us not available to Anglia. - by ignoring parts of the institutional population. How the CPHM definition is obtained: The calculation is : Standard institutional (non-domestic) (from S01 - includes staff & staff family) Plus - Students in student only and in other households, but not in parental home (S063) Minus - Students who are studying elsewhere, but whose parental home is in the study area (S012) Where are students measured? In the base population students are included in the estimates/2001 Census through term-time residence. Within this population students have different characteristics, which affect their transience and how they should be modelled. In particular, it is clear from 1991 and some 2001 results that not all students have been identified in the Census, especially in migration data. Study of the Census results (full-time students over 18) and information from the universities indicates the size of this population in the City. They are remarkably consistent 24,424 and 24,327 respectively. However, this is entirely coincidental, as the definitions are not comparable and university sources are only apportioned by known postcodes. Both totals include students living with parents. Other definitions can apply, for example, from the Census students over 18 not living with their parents amount to 23,099. The figure including those under 18, at boarding school, for example, is 25,967. On the other hand, economically inactive students, who are more likely to be full-time students in higher education, amount to 18,620 (Census). What transient population was actually used? \\SERVER2\F_USERS\SHEPHERA\Projecting_transient_Cooper.doc 5
Through discussions of the projections with Nottingham City we noted that the projections we now had were giving results that did not reflect the expected structure. In particular, if the student population was to be maintained (as was expected), then the remaining population once those students were removed in future years did not reflect the population ageing through from known younger age groups and births. (NB Was this higher or lower? figures in charts below indicate too high 15-19 yrs fig. in 2021). Therefore we used an iterative estimation process to create a best fit scenario ; in other words trial & error. The purpose was to try to model what the population in the relevant age groups (19-29) should be if the transient element of it remained constant. For this we studied birth trends from 1981 to determine what the indigenous trend indicated. Of course, changes in migration at younger age groups would affect this, and once the 15-19 age group was reached the movement of students themselves overwhelmed the information on resident population. We attempted to calculate what the likely student migration levels were. Using a population of 24,000 for non-home students, Annual in-out moves would be around 6-7,000 (assuming a 3/4 year course).??? more text How much better is the projection? Essentially we used a pragmatic approach, which attempted to adjust for known aspects of concern in the projection. It may have ignored others, for example, the effect upon older age groups. (Note: Anglia university indicate that this solution will not necessarily apply to other areas, and the default settings of the model are those that can be applied consistently for all areas. Considerable work has been done and is continuing into constructing a consistent transient population.) The ONS 1996-based population projection created standard migration profiles that relate to the whole population, however adjustments had to be introduced to compensate for student populations. In Nottingham City s case, while being a trendbased projection and derived from a pre-census mid-1996 base (and therefore inherently unrealistic), the age structure corresponds to what might have been expected (see chart). Was it worth updating the migration data if such an adjustment had to be made? What other options might there be? DO not ignore. Migrate all students in & then out this was done in mid 1980s for Notts structure plan. Could be done, but again there is a difficulty in assessing how many are involved and whether double counting is taking place. It will be possible to check figures we have from the 2001 Census to identify migrants will this be complete? Movements could be obtained from the universities to create a modelled migration profile then non-students could be added. It is likely that the Census includes some but not all student movements. \\SERVER2\F_USERS\SHEPHERA\Projecting_transient_Cooper.doc 6
Use a revised transient population to include students and revising migration to exclude students. Students may be removed from the 2001 Census migration flows and a transient population could be calculated which includes students and other categories of transient population, as we did in the 1991-based projections. \\SERVER2\F_USERS\SHEPHERA\Projecting_transient_Cooper.doc 7
Nottingham UA 2001 Census results Nottinghamshire rest of Plan Area 2001 Census results \\SERVER2\F_USERS\SHEPHERA\Projecting_transient_Cooper.doc 8
Comparison of Census,CPHM and University sources of student / transient figures Bassetlaw Nottingham Plan Area CPHM transient population: 2,922 27,552 41,986 CPHM Transient (15-29) 599 22,553 25,401 CPHM adjustment from Census inst. Pop 20 18,565 22,622 Economically inactive Students (TT02) 617 18,620 25,424 Students 18+ away from home (ST012) 910 1,220 7,816 2002 Students at universities 63 24,237 29,726 Students over 18 not living with parents (TT02) 657 23,099 30,632 Difference between incoming & outgoing students -253 21,879 22,816 (University students out of County:) 2,300 (Bassetlaw has been used as an illustration of a Nottinghamshire district.) NB Transient population assumed for Nottingham in earlier NCC runs was 14,540 1 All students not living with parents. (TT02) 2 All Students 19 & over (inc those living with parents) (TT02) 3 All Students 19 & over economically inactive (TT02) 4 Students over 18 who would be at home, located by home address (i.e. absent) ST012) 5 students described as in (4) - all ages (i.e. inc pupils) (ST012) 6 Students registered at both universities, split into district by postcode (and apportioned where PC not present). For Nottingham The CPHM methodology gives a total of 27,552 transient population for Nottingham, and the adjustment accounts for 18,565. The 15-29 age group gives 22,553 and 16,143 (adjustment) The 15-44 group gives 24,372 and 17,476 (adjustment) The adjustment for 45+ is 272 The institutional population is 6,407 (15-29), 6,896 (15-44) It appears reasonable to assume that the total student population in Nottingham City is 25,000, and that 29 or 44 could be reasonable upper limits to any adjustment to transient or migration population. In other districts it will make little difference if the CPHM transient population is used. \\SERVER2\F_USERS\SHEPHERA\Projecting_transient_Cooper.doc 9
Comparison of gross migration: Patient Register & 1991 Census 7000 6000 PR out Census out Census in 5000 4000 3000 2000 1000 PR in PR in PR out Census in Census out Gross each w ay flows: Census: 28-30,000 ONS PR: 17-19,000 0 0-4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75+ ONS 96-based (trend) and JSP DD projections - Nottingham 35000 30000 25000 20000 15000 10000 NCC DD proj 2001 2021 NCC DD proj Actual 2001 5000 0 0-4 10-14 20-24 30-34 40-44 50-54 60-64 70-74 80-84 \\SERVER2\F_USERS\SHEPHERA\Projecting_transient_Cooper.doc 10
ONS 96-based (trend) and JSP DD projections - Nottingham 35000 30000 25000 20000 15000 10000 2001 2021 NCC DD proj 5000 0 0-4 10-14 20-24 30-34 40-44 50-54 60-64 70-74 80-84 \\SERVER2\F_USERS\SHEPHERA\Projecting_transient_Cooper.doc 11