Distance to school

In straight lines, and then on real streets and bus timetables: do children in deprived neighbourhoods live further from a secondary school than their neighbours, where is that unusual, and does absence go with it?

In short

Across England, children in deprived neighbourhoods usually live closer to a secondary school. That holds in cities, towns and the countryside, and in 81% of districts.
Brighton and Hove is the other way round. Children in the most deprived third of its neighbourhoods live 1.57 km from their nearest school, in a straight line, and the rest live 1.18 km. That is 33% further, the 5th largest difference of 293 districts proportionately and the 2nd largest in kilometres. Among the 109 mostly urban local authorities, none has its deprived neighbourhoods further from a school.
Absence goes with distance, more for disadvantaged pupils, in every year tested. Across 150 authorities and six years from 2018/19, disadvantaged pupils’ absence is 3.5 points higher for each extra kilometre between deprived neighbourhoods and a school, against 0.6 for non-disadvantaged pupils and 1.7 for all pupils. It holds on eleven measures of distance and is far weaker in primary schools, as it should be if the journey to secondary school matters.
But most of that slope is who lives in those places, not the journey. Given the ethnic mix of disadvantaged pupils in each authority, the slope falls to 0.6, and with every control to 0.3, with an interval that includes zero. Ethnic mix is a confounder here: places with schools close together tend to have disadvantaged pupils from groups who attend more.
A smaller link remains in three separate tests. In mostly urban authorities outside London, the slope for disadvantaged pupils is still 2.5 points per kilometre given ethnic mix and region, against 0.4 in primary schools. Within one authority, pupils in the most deprived neighbourhoods have 0.45 points more absence per kilometre. And each doubling of how far a school’s nearest children live goes with 0.26 points more absence in every year, given the school’s intake and ethnic make-up, and more where more pupils are on free school meals. For disadvantaged pupils the link is mainly with unauthorised and persistent absence, not illness. These are associations, not proof of cause.
Moving a school would change this. A school back on CoMArt’s old site, or Longhill moved to the top of Elm Grove, would bring the deprived neighbourhoods of Brighton and Hove to about 1.19 to 1.21 km and close most of the gap.
Real journeys tell the same story for Brighton and Hove. By walking and bus, children in its most deprived neighbourhoods are 4.7 minutes further from a school than the rest, and walking alone 9.0 minutes. No comparator place is more than 1.3 minutes further by walking and bus, and its deprived neighbourhoods need the most changes of bus. See journey times.
Changes of bus, school buses and hills show little general link with absence so far. Longhill, where missing the last workable journey means arriving about 17 minutes late, is a striking case, not a pattern.
Most of this page is in straight lines. They ignore hills, bus routes and catchments, and a child’s nearest school need not be the one they attend. Every source and calculation is in the methodology.

What this measures

How far each school’s own pupils travel is not published, so this page starts from where children live and where schools are. This is a summary: the methodology gives every source, vintage and formula.

Schools. Every open state-funded secondary, all-through and middle deemed secondary school in England in the DfE’s register, Get Information About Schools (June 2026): 3,282 schools taking pupils at 13 or younger and keeping them to at least 16. Special schools, pupil referral units and independent schools are left out.
Neighbourhoods. England’s 33,755 Census 2021 Lower Layer Super Output Areas, each placed at the mean of its Output Areas’ population-weighted centroids. Averages are weighted by the 10 to 14 year olds living there.
Deprivation. The share of households deprived in two or more of the Census’s four dimensions (employment, education, health, housing), the same measure as the neighbourhood covariates.
Distance. The straight line to the nearest school. Each place’s neighbourhoods are split into its most deprived third and the rest, so the comparison is always within one place.
Kind of place. The ONS 2021 rural-urban classification. A district or authority is “mostly urban” if at least 80% of its 10 to 14 year olds live in urban neighbourhoods.

Schools in Wales are not in the register used here, so a few neighbourhoods on the border look further from a school than they are.

Across England, deprived neighbourhoods are nearer

Within each kind of place, the most deprived neighbourhoods tend to be among the closest to a school. One likely reason is that they are usually denser and more central.

Figure 1: Mean straight-line distance from home to the nearest state secondary school, by how deprived the neighbourhood is (England’s tenths, 10 = most deprived) and the ONS 2021 rural-urban class of the neighbourhood. Weighted by 10 to 14 year olds.

Taken across England as a whole, children in the most deprived third of neighbourhoods live 1.00 km from a school and the rest 1.75 km. Much of that is town against country, which is why the rest of this page compares neighbourhoods within one place.

Where it runs the other way

In 19% of England’s 293 districts, children in the most deprived third of neighbourhoods live further from a school than the rest. Among the 173 mostly urban districts it is 27%. Brighton and Hove is near the top of both lists.

Figure 2: How much further children in the most deprived third of neighbourhoods live from their nearest school than children in the rest of the same district, for all 293 districts, from largest to smallest. Above 1, the deprived neighbourhoods are further away.

Table 1: The districts where children in deprived neighbourhoods live furthest from a school, relative to the rest of the district, with the rank in kilometres. Straight-line distance to the nearest school, in kilometres.

Rank	District	Most deprived third	Other neighbourhoods	Difference (km)	Rank in km	Kind of place
1	Havering	1.15	0.78	+0.37	5	Mostly urban
2	Three Rivers	1.92	1.37	+0.55	1	Mostly urban
3	Derby	1.24	0.91	+0.33	7	Mostly urban
4	Tameside	1.17	0.88	+0.29	8	Mostly urban
5	Brighton and Hove	1.57	1.18	+0.39	2	Mostly urban
6	Basildon	1.56	1.19	+0.37	6	Mostly urban
7	Hastings	1.67	1.29	+0.38	4	Mostly urban
8	Wandsworth	0.82	0.65	+0.17	15	Mostly urban
9	St Albans	1.29	1.04	+0.25	9	Mostly urban
10	East Hertfordshire	2.10	1.71	+0.39	3	Mixed

The list mixes outer London boroughs, commuter towns and cities such as Derby. Here is Brighton and Hove against other coastal and urban places, and against the rest of Sussex:

Figure 3: Straight-line distance to the nearest school: children in the most deprived third of each place’s neighbourhoods against children in the rest.

Table 2: Places compared. Distance to the nearest school in kilometres for the most deprived third of neighbourhoods and the rest, how much further that is, with the rank of 293 districts in brackets, the share of secondary pupils who are disadvantaged, and absence in state-funded secondary schools, full year 2024/25, for disadvantaged pupils, non-disadvantaged pupils and all pupils (DfE, absence by pupil characteristics). DfE publishes these by authority, not by district, so Hastings has none.

	Distance to school			Pupils disadv.	Absence
Place	Deprived third	Other	Further	Share	Disadv.	Non-disadv.	All
Brighton and Hove	1.57	1.18	+33% (5)	28%	18.0%	8.0%	10.8%
Hastings	1.67	1.29	+30% (7)	–	–	–	–
Blackpool	1.02	0.87	+17% (13)	48%	13.2%	6.1%	9.6%
Southampton	0.87	0.78	+12% (20)	41%	15.7%	6.8%	10.5%
Portsmouth	0.85	0.82	+4% (44)	39%	15.5%	6.4%	10.1%
Plymouth	0.83	0.89	-6% (80)	30%	16.4%	7.4%	10.2%
Newham	0.53	0.58	-9% (93)	51%	8.3%	5.5%	7.0%
Bristol	0.84	0.98	-14% (115)	36%	15.8%	7.2%	10.3%
Tower Hamlets	0.41	0.50	-17% (132)	56%	7.6%	5.1%	6.5%

Brighton and Hove and Hastings on the map

Figure 4: Each dot is a neighbourhood, sized by its 10 to 14 year olds and shaded by the straight-line distance to its nearest secondary school. A ring marks the most deprived third of the place’s neighbourhoods. Diamonds are state secondary schools. Census 2021 and Get Information About Schools.

In Brighton and Hove, the neighbourhoods furthest from a school are in the east of the city, around Whitehawk, between the central schools and Longhill. The city’s school system model finds the same in routed walk and bus times: in its simulator, disadvantaged children travel about 4 minutes further each way than everyone else.

Does it depend on how it is measured?

Table 3: How much further the most deprived third of neighbourhoods is from a school than the rest, on five versions of the measure, with each place’s rank of 293 districts.

Measure	Brighton and Hove	Hastings	Blackpool
Nearest school (headline)	+33% (5)	+30% (7)	+17% (13)
Nearest, measured from Output Areas	+35% (3)	+27% (7)	+17% (12)
Mean of the three nearest	+24% (4)	+6% (36)	+4% (42)
Nearest neither selective nor faith	+23% (12)	-6% (85)	+10% (31)
Nearest; deprived in one or more dimensions	+28% (10)	+23% (13)	+20% (17)

Brighton and Hove’s result holds on every version: its deprived neighbourhoods are 23% to 35% further, in the top 12 of 293 districts. Hastings and Blackpool are less certain. Their deprived neighbourhoods are further from their nearest school, but the difference is much smaller on the mean of the three nearest (+6% and +4%). In Hastings it reverses (-6%) if faith and selective schools are set aside.

Does absence go with distance?

FFT Education Datalab compared pupils within schools, using straight-line distance from their home postcodes. Those living furthest away were absent slightly more often, and the difference was wider for disadvantaged pupils. FFT calls distance “a relatively minor factor” and suggests that travel times may be instructive. Published data allow several further tests. Every source and calculation is set out in the methodology.

A first look: by authority, for all three groups

DfE publishes absence by disadvantage for each local authority, so the distance from deprived neighbourhoods to a school can be set against absence for each group of pupils.

Figure 5: Absence in state-funded secondary schools, 2024/25, against the straight-line distance to a school from the most deprived third of each local authority’s neighbourhoods, for 150 authorities. Lines are simple fits.

Table 4: Points of absence for each extra kilometre between an authority’s most deprived neighbourhoods and a school, 2024/25, for each group of pupils, with 95% intervals. Second and third rows allow for the share of pupils who are disadvantaged and how urban the authority is.

Specification	All pupils	Disadvantaged	Non-disadvantaged
Distance alone (150 authorities)	+1.0 (+0.6 to +1.4)	+3.9 (+3.1 to +4.8)	+0.7 (+0.5 to +0.9)
Given the share disadvantaged and how urban (150 authorities)	+1.8 (+1.2 to +2.4)	+3.8 (+2.5 to +5.1)	+0.4 (+0.1 to +0.8)
And without London (118 authorities)	+0.9 (+0.3 to +1.5)	+2.0 (+0.7 to +3.2)	+0.4 (-0.0 to +0.8)

The slope looks about 9 times steeper for disadvantaged pupils than for non-disadvantaged pupils once share and urbanity are allowed for, and about 5 times steeper without London. London pulls the fit up, because its schools are close together and its disadvantaged pupils attend well.

It is the distance that goes with absence, not the gap within the authority. Added to the same model, how much further the deprived neighbourhoods are than the rest has no clear relationship with absence for any group (disadvantaged t = -0.4, non-disadvantaged t = -1.4, all pupils t = -0.3). Fairness in access and the level of access are separate questions.

One year, one measure and three controls cannot carry much weight. The rest of this section tries to break the result.

How firm is it?

The answer so far

The pattern is consistent. The size of Figure 5 is not to be trusted. Absence goes with distance in every year from 2018/19 and on every measure of distance tried, more for disadvantaged pupils, and far less in primary schools. But most of the slope in Figure 5 comes from who lives where. Given the ethnic mix of each group of pupils it falls by about four-fifths, and with every control no clear slope is left across all authorities. A smaller link remains in mostly urban authorities, within authorities for the most deprived neighbourhoods, and between schools.

Table 5 puts the authority tests side by side, pooling the six years.

Table 5: Points of absence for each extra kilometre between an authority’s most deprived neighbourhoods and a school, for each group of pupils, with 95% intervals. All years pooled: 2018/19 and 2020/21 to 2024/25, with year effects and errors clustered by authority. Every row allows for the share of pupils who are disadvantaged and how urban the authority is, except the authority-effects row, which allows for the share.

Test	All pupils	Disadvantaged	Non-disadvantaged
Secondary, all authorities	+1.70 (+1.11 to +2.28)	+3.54 (+2.36 to +4.73)	+0.58 (+0.26 to +0.90)
... without London	+0.69 (+0.14 to +1.23)	+1.52 (+0.39 to +2.66)	+0.30 (-0.07 to +0.68)
... given region	+0.79 (+0.31 to +1.26)	+1.48 (+0.51 to +2.46)	+0.28 (-0.03 to +0.60)
... given the ethnic mix of that group	+0.43 (-0.03 to +0.88)	+0.64 (-0.25 to +1.53)	+0.21 (-0.13 to +0.56)
... given every control	+0.18 (-0.21 to +0.57)	+0.34 (-0.56 to +1.25)	+0.05 (-0.19 to +0.29)
Mostly urban, without London	+1.48 (+0.59 to +2.36)	+3.61 (+2.05 to +5.18)	+0.53 (-0.12 to +1.19)
... given the ethnic mix of that group	+0.58 (-0.25 to +1.42)	+1.74 (+0.26 to +3.21)	+0.23 (-0.41 to +0.88)
... given ethnic mix and region	+1.17 (+0.55 to +1.78)	+2.53 (+1.40 to +3.66)	+0.75 (+0.23 to +1.26)
... given every control	+0.56 (-0.33 to +1.46)	+1.16 (-0.50 to +2.82)	+0.29 (-0.34 to +0.92)
Less urban authorities	+0.09 (-0.34 to +0.52)	+0.26 (-0.81 to +1.32)	+0.07 (-0.24 to +0.38)
Falsification: primary absence	+0.05 (-0.14 to +0.24)	+0.49 (+0.13 to +0.85)	-0.26 (-0.45 to -0.07)
Secondary minus primary absence	+1.64 (+1.08 to +2.19)	+3.03 (+1.97 to +4.09)	+0.83 (+0.49 to +1.18)
Authority effects: as each school network changed	+1.65 (-0.78 to +4.09)	+3.16 (-0.86 to +7.18)	+1.68 (+0.10 to +3.26)

Across six years

Figure 6: Points of absence for each extra kilometre between an authority’s most deprived neighbourhoods and a school, year by year, using the school network of that September, with 95% intervals and allowing for the share disadvantaged and how urban the authority is. DfE publishes no 2019/20. Full years, state-funded secondary schools.

The slope for disadvantaged pupils is positive in all six years, from 2.3 to 4.5 points per kilometre. It was already there before the pandemic (+2.3 in 2018/19) and has been larger since 2021/22. Authorities whose deprived neighbourhoods were further from a school saw disadvantaged absence rise more between 2018/19 and 2024/25 (+1.6 points per kilometre, t = 2.9), with no clear difference for non-disadvantaged pupils (+0.1, t = 0.7).

Distance is recalculated each year from school opening and closing dates, so earlier years are not measured against today’s schools. The network has grown from 3,226 schools in September 2018 to 3,289 in September 2025.

On every measure of distance, and in every model

To avoid resting on any one choice, the same model was fitted on every combination of 6 years, 11 measures of distance, 6 sets of controls and 5 samples of authorities, for each group of pupils: 1,980 models per group, shown in Figure 7. These models are not independent of one another, so the shares describe how far the answer depends on the choices made. They are not a significance test.

Figure 7: A specification curve of 5,940 models of secondary absence, each dot one model, sorted by size, in points of absence for one standard deviation more distance. Purple dots have a 95% interval wholly above zero, blue dots wholly below, grey dots include zero.

Table 6: The share of those models in which the slope is above zero, and in brackets the share in which its 95% interval is wholly above zero, by the set of controls and the sample of authorities.

Choice	All pupils	Disadvantaged	Non-disadvantaged
None	85% (46%)	91% (79%)	94% (62%)
Share disadvantaged, how urban	94% (68%)	94% (73%)	94% (38%)
And region	86% (55%)	86% (61%)	83% (24%)
And Census place	72% (15%)	69% (18%)	72% (6%)
And pupils	62% (1%)	63% (7%)	54% (0%)
And primary absence	64% (0%)	56% (2%)	58% (0%)
All authorities	85% (39%)	85% (46%)	79% (30%)
Without London	72% (18%)	68% (32%)	81% (19%)
Mostly urban	97% (56%)	100% (66%)	90% (37%)
Mostly urban, without London	94% (35%)	97% (55%)	90% (14%)
Less urban	37% (6%)	33% (0%)	41% (8%)

Across all of these, 77% of the slopes for disadvantaged pupils are above zero, 40% clearly so and 2% clearly below. The choice of measure makes little difference. The choice of controls and sample makes a great deal, as Table 6 shows: with few controls most models are clearly positive, and under the fullest sets few are.

Table 7: Eleven measures of distance from an authority’s deprived neighbourhoods. Points of absence for one standard deviation more distance, all years pooled, allowing for the share disadvantaged and how urban the authority is. A dagger marks a 95% interval that includes zero.

	All authorities			Mostly urban, without London
Measure	All	Disadv.	Non-disadv.	All	Disadv.	Non-disadv.
Nearest school (headline)	+0.61	+1.27	+0.21	+0.53	+1.29	+0.19†
Nearest school, logged	+0.72	+1.53	+0.23	+0.49	+1.23	+0.16†
Mean of the three nearest	+0.54	+0.93	+0.23	+1.07	+2.58	+0.37†
Nearest neither selective nor faith	+0.53	+1.07	+0.19	+0.47	+1.16	+0.13†
Nearest, from Output Areas	+0.62	+1.29	+0.21	+0.55	+1.34	+0.20†
Deprived in one or more dimensions	+0.60	+1.21	+0.22	+0.54	+1.30	+0.19†
Most deprived fifth	+0.49	+1.05	+0.16	+0.35	+0.82	+0.12†
England's most deprived third	+0.51	+1.04	+0.17	+0.47	+1.07	+0.19†
Weighted by household deprivation	+1.10	+2.20	+0.46	+1.11	+2.58	+0.49
Weighted by child poverty (IDACI)	+1.03	+2.06	+0.42	+0.93	+2.09	+0.40
Share of children over 2 km	+0.31	+0.75	+0.12	+0.32	+0.75	+0.14†

Nor does it depend on a handful of places. Table 8 repeats the 2024/25 model with errors that allow for neighbouring authorities being alike, with pupil weights, with a method that discounts outliers, and without the most influential authorities. Dropping any one authority leaves the slope for disadvantaged pupils between 3.6 and 4.2.

Table 8: Checks on the 2024/25 model of Table 4, given share and urbanity, in points of absence per kilometre with 95% intervals.

Check	All pupils	Disadvantaged	Non-disadvantaged
Baseline, 2024/25 (robust errors)	+1.79 (+1.03 to +2.55)	+3.85 (+2.33 to +5.37)	+0.46 (+0.00 to +0.91)
Errors allowing for neighbouring authorities (within 100 km)	+1.79 (+0.78 to +2.80)	+3.85 (+1.67 to +6.03)	+0.46 (+0.03 to +0.88)
Weighted by pupils	+1.23 (+0.41 to +2.05)	+2.62 (+0.97 to +4.27)	+0.18 (-0.29 to +0.65)
Robust regression (down-weights outliers)	+1.88 (+1.25 to +2.52)	+4.24 (+2.87 to +5.61)	+0.38 (+0.02 to +0.73)
Without the five most influential authorities	+2.00 (+1.33 to +2.67)	+5.55 (+4.27 to +6.83)	+0.41 (+0.08 to +0.74)
Without Brighton and Hove	+1.64 (+0.91 to +2.37)	+3.63 (+2.11 to +5.16)	+0.30 (-0.04 to +0.65)

Which neighbourhoods’ distance matters is also informative. Put the deprived neighbourhoods and the others in one model, and for disadvantaged pupils the deprived neighbourhoods’ distance carries about twice the weight (+3.1 against +1.6 per kilometre). For non-disadvantaged pupils the two are about equal (+0.4 and +0.4).

The relationship is not a straight line, either. Compared with authorities under 1 km, disadvantaged absence is 1.8 points higher at 1 to 1.5 km and 2.2 at 1.5 to 2 km, but only 0.7 at 2 to 3 km, where there are just 4 authorities, all large and rural.

What takes the slope away?

Controls overlap, so adding them one by one cannot say which one matters. Figure 8 splits the fall from the first model to the fullest into a part for each block of controls, using a method that does not depend on the order they go in (a Gelbach decomposition; details).

Figure 8: How much of the fall in the slope, from the model with share and urbanity to the model with every control, each block of controls accounts for, in points of absence per kilometre, all years pooled, with 95% bootstrap intervals. The parts add up to the whole fall. Below zero, a block would make the slope larger if taken away.

It is ethnic mix, and very little else. Of the fall from 3.52 to 0.34 points per kilometre for disadvantaged pupils, ethnic mix and language account for 3.01. Region adds 0.69, almost all of it London. Without London region adds nothing (-0.05), and ethnic mix accounts for all of the fall (1.72, against a fall of 1.63, since the other parts are slightly negative overall). Density, health, cars, deprivation and needs each do little once ethnic mix is held, and their parts roughly cancel.

Figure 9 shows it directly, using DfE’s count of pupils by ethnicity and free school meals in each authority, so that each group of pupils is compared on its own ethnic mix.

Figure 9: The slope of absence on distance, all years pooled, with 95% intervals, as controls are added. Ethnic mix is the shares of that group of pupils who are White British, Asian, Black and of mixed heritage, by authority and year. Points of absence per kilometre.

Why would ethnic mix do this? In 2024/25 the White British share of an authority’s disadvantaged pupils correlates with how far its deprived neighbourhoods are from a school (r = 0.69), since diverse places tend to be dense places with schools close together. On its own that share accounts for 63% of the variation in disadvantaged absence between authorities. So where schools are close, disadvantaged pupils also tend to be from groups with higher attendance, and a simple plot such as Figure 5 credits distance with that. This says nothing about why attendance differs between groups: ethnic mix stands in for migration histories, family and community ties and much else, none of which the data can separate.

Confounding or mediation? A mediator lies on the path: distance would change it, and it would change absence. A confounder drives both. The evidence points to confounding.

A longer journey to school does not change the ethnic mix of a place, so ethnic mix cannot sensibly be a mediator of distance. It is a confounder, or composition.
The obvious candidates for mediation contribute little. Households with no car and population density barely move the slope once ethnic mix is held (Figure 8). That does not rule out mediation, because authority-wide averages are a poor measure of how any one family travels.

The test of composition is whether the link holds within one ethnic group. DfE publishes absence by ethnic group for each authority, though not split by disadvantage.

Table 9: Within each ethnic group: points of secondary absence per kilometre between an authority’s deprived neighbourhoods and a school, and from where that group lives (the mean over all neighbourhoods weighted by the group’s Census residents), all years pooled, allowing for the share disadvantaged and how urban the authority is, with 95% intervals. All pupils in each ethnic group; DfE does not split these by disadvantage. Authorities with at least 100 pupils in the group.

Ethnic group	Distance	All authorities	Without London	Mostly urban, without London
White British	From deprived neighbourhoods	+0.84 (+0.26 to +1.42)	+0.14 (-0.45 to +0.72)	-0.02 (-1.23 to +1.18)
White British	From where the group lives	+1.12 (+0.38 to +1.86)	+0.41 (-0.18 to +1.00)	+0.63 (-0.46 to +1.71)
Asian	From deprived neighbourhoods	-0.38 (-0.98 to +0.21)	-0.42 (-1.17 to +0.33)	-0.33 (-1.50 to +0.85)
Asian	From where the group lives	-0.55 (-1.05 to -0.04)	-0.66 (-1.26 to -0.06)	-1.09 (-2.29 to +0.11)
Black	From deprived neighbourhoods	-1.72 (-2.56 to -0.89)	-0.50 (-1.42 to +0.42)	-0.52 (-1.94 to +0.90)
Black	From where the group lives	-1.26 (-1.91 to -0.60)	-0.26 (-0.88 to +0.35)	-0.42 (-1.79 to +0.96)
Mixed heritage	From deprived neighbourhoods	+0.23 (-0.42 to +0.87)	-0.44 (-1.31 to +0.43)	-0.63 (-2.24 to +0.98)
Mixed heritage	From where the group lives	+0.08 (-0.53 to +0.69)	-0.66 (-1.29 to -0.03)	-0.93 (-2.51 to +0.64)

The slope mostly goes. For White British pupils it is +0.84 per kilometre, and about zero outside London (+0.14) and in mostly urban authorities outside London (-0.02). For Asian and Black pupils it is below zero. Directly standardising says the same for all pupils: of a slope of 1.70, 1.01 is what each authority’s ethnic mix would give at national rates, and 0.69 is left (t = 3.3), or 0.28 without London (t = 1.1).

This test cannot settle things for disadvantaged pupils, because most pupils in each ethnic group are not disadvantaged, and non-disadvantaged pupils show little slope anyway. A clear link among disadvantaged White British pupils could be hidden inside a flat line for all White British pupils. DfE publishes nothing that would show it.

Is anything left once the controls are in?

Table 10: What is left in the fullest model: share, urbanity, the group’s ethnic mix, region, seven Census measures and six measures of pupils. Points of absence per kilometre, all years pooled, with 95% intervals. Explained is the share of the variation in distance that those controls account for.

Sample	Authorities	All pupils	Disadvantaged	Non-disadvantaged	Explained
All authorities	150	+0.18 (-0.21 to +0.57)	+0.34 (-0.56 to +1.25)	+0.05 (-0.19 to +0.29)	76%
Without London	118	+0.02 (-0.35 to +0.38)	-0.12 (-1.00 to +0.75)	+0.04 (-0.20 to +0.28)	69%
Mostly urban	109	+0.74 (-0.03 to +1.50)	+1.92 (+0.49 to +3.36)	+0.28 (-0.22 to +0.78)	76%
Mostly urban, without London	77	+0.56 (-0.33 to +1.46)	+1.16 (-0.50 to +2.82)	+0.29 (-0.34 to +0.92)	65%
Less urban	41	-0.16 (-0.53 to +0.20)	-0.54 (-1.37 to +0.30)	-0.00 (-0.24 to +0.23)	72%

Across all authorities, no clear slope is left for any group: +0.34 for disadvantaged pupils, with an interval from -0.6 to +1.2. Adding distance to that model leaves its fit unchanged (R² 0.881 to 0.881). But that interval does not mean zero. It still allows a little over one point per kilometre, and the controls account for 76% of the variation in distance, so the model has little left to work with, and 28 terms for 150 authorities.

The link survives further in towns and cities. In mostly urban authorities it is +1.92 with every control (t = 2.6). Outside London it is +2.53 given ethnic mix and region (t = 4.4), and +1.16 with every control, when its interval includes zero. In less urban authorities there is nothing at any stage.

Other designs are not subject to this limit, and they are in the sections that follow. Within authorities the link stays in the most deprived neighbourhoods given Census character, including their White British residents (below). At school level it survives the school’s own ethnic make-up in every year (below).

A falsification test: primary schools

If the link is about getting to a secondary school, the distance to a secondary school should say little about absence in primary schools. If it is about the place, it should predict both. The same children and families are involved, so this is a sharp test.

Figure 10: The distance from deprived neighbourhoods to a secondary school against absence in secondary schools, in primary schools, and the step up between them in the same authority and year, all years pooled, with 95% intervals. Points of absence per kilometre, allowing for the share disadvantaged and how urban the authority is.

It passes on these controls. For disadvantaged pupils the slope is +3.5 in secondary schools and +0.5 in primary, and it is +3.0 on the step up from primary to secondary within the same authority and year, which sets aside anything about a place that affects both phases. For non-disadvantaged pupils the primary slope is -0.3, below zero. The step up is steeper for disadvantaged pupils than for non-disadvantaged pupils by 2.2 points per kilometre (t = 5.3), or 0.7 without London (t = 2.0).

The small primary slope for disadvantaged pupils is not zero, which suggests some of the secondary slope is the place. Distance to a primary school itself goes with primary absence for disadvantaged pupils (+1.8 per kilometre, t = 2.1), though primary schools are so close that a kilometre is a long way. The test is less clear once ethnic mix is held, because the difference in attendance between ethnic groups is itself wider in secondary school than in primary. Across all authorities the step up for disadvantaged pupils falls to +0.42 (t = 1.0). In mostly urban authorities outside London the test still passes: given ethnic mix and region, +2.5 in secondary, +0.4 in primary, and +2.0 on the step up (t = 3.3).

Within the same authority

Everything so far compares one authority with another, so anything about a whole place can be mistaken for distance. A second DfE data set offers a stronger test. For 2023/24 and 2024/25 it gives absence by where pupils live, not where they go to school, in each authority, by the national tenth of the Income Deprivation Affecting Children Index (IDACI) of their home neighbourhood. Those tenths can be matched to the distance from the same neighbourhoods, giving 2,898 groups of pupils in 150 authorities. The number of pupils in each group lines up closely with the Census count of children (r = 0.99). DfE does not split this file by disadvantage, so it describes all pupils living in each kind of neighbourhood.

Brighton and Hove stands out in it (Figure 11). Across England, children in the most deprived neighbourhoods live nearest a school. In Brighton and Hove they live furthest away, and absence and persistent absence climb much more steeply with deprivation than in England.

Figure 11: Pupils in state-funded secondary schools, 2024/25, by the national IDACI tenth of the neighbourhood where they live (1 = most deprived): straight-line distance to a school, overall absence, and the share who are persistently absent. All pupils; DfE does not publish this by disadvantage. DfE, absence by pupil residence; IoD 2019; Census 2021; Get Information About Schools.

The children in the city’s most deprived tenth of neighbourhoods live 1.93 km from a school, against 1.01 km for that tenth across England. They miss 17.8% of sessions against 11.0%, and 49% are persistently absent against 32%.

Table 11: Points of absence per kilometre for secondary pupils living in each kind of neighbourhood, 2023/24 and 2024/25, weighted by pupils, with 95% intervals and errors clustered by authority. All pupils in those neighbourhoods. Between authorities compares the same IDACI tenth across authorities; within authorities compares neighbourhoods in the same authority, against the national pattern for each tenth. Census character is how urban and dense the neighbourhoods are, and their graduates, White British residents, households with no car, social renting and bad health.

	Neighbourhoods by IDACI tenth
Comparison	Most deprived	Middle	Least deprived
Between authorities	+2.52 (+1.67 to +3.37)	+0.69 (+0.50 to +0.87)	+0.54 (+0.40 to +0.67)
Between, given how urban and dense	+1.57 (+0.61 to +2.53)	+0.12 (-0.30 to +0.55)	+0.39 (+0.09 to +0.69)
Within authorities	+0.21 (-0.09 to +0.51)	-0.06 (-0.18 to +0.06)	+0.10 (-0.01 to +0.21)
Within, given Census character	+0.45 (+0.20 to +0.70)	+0.02 (-0.12 to +0.17)	+0.08 (-0.07 to +0.24)

Three things follow.

Most of the slope between authorities is not there within them. For the most deprived three tenths it is +2.5 per kilometre between authorities, and +0.21 within them, or +0.45 given Census character. Part of that fall will be noise, because the distance for a tenth of an authority’s neighbourhoods is a rough average of places that can be far apart. Even so, the safer reading is that most of the slope between authorities is about the authorities, which fits what ethnic mix does above. Holding Census character raises the slope, because within an authority the further neighbourhoods tend to be less urban and less dense (r = -0.51 and -0.46), and differ little in their White British share (r = 0.10). So the clearer estimate depends on that adjustment: without it the interval includes zero.
What remains is where it should be. Within authorities, given Census character, the link is clear only for pupils in the most deprived three tenths. It is steeper there than in the least deprived three tenths by 0.59 points per kilometre (t = 3.1). It is similar in each year (+0.48 and +0.41), without London (+0.48) and without Brighton and Hove (+0.41).
It is much weaker in primary schools. In the same neighbourhoods the primary slope is +0.14, with an interval that includes zero, against +0.45 in secondary. The persistent absence result is +1.6 points of pupils per kilometre in secondary (t = 3.7) and +0.5 in primary (t = 1.6).

Figure 12: Points of absence per kilometre for pupils living in each national IDACI tenth of neighbourhood, comparing authorities within each tenth, given Census character, 2023/24 and 2024/25, with 95% intervals. All pupils in those neighbourhoods.

When the school network changes

The strongest test available would follow the same place as a school opens or closes nearby. Table 12 does this for authorities and for schools. Very few authorities’ distance moved much: only 17 of 150 changed by 50 metres or more between 2018/19 and 2024/25. Brighton and Hove did not change at all. So these estimates are imprecise.

Table 12: When distance changes in the same place. Authority rows use authority and year effects and allow for the share disadvantaged, in points of absence per kilometre. School rows use school and year effects and allow for intake, in points of overall absence for each doubling of how far a school’s nearest children live, with the second row adding effects for each authority in each year. DfE publishes school absence for all pupils only. 95% intervals; errors clustered by authority or by school.

Comparison	Estimate
Authorities: secondary, disadvantaged	+3.16 (-0.86 to +7.18)
Authorities: secondary, non-disadvantaged	+1.68 (+0.10 to +3.26)
Authorities: secondary, all pupils	+1.65 (-0.78 to +4.09)
Authorities: primary, disadvantaged	+0.18 (-1.32 to +1.68)
Authorities: primary, non-disadvantaged	+0.56 (-0.07 to +1.18)
Authorities: primary, all pupils	+0.25 (-0.30 to +0.80)
Schools: all pupils	+0.57 (+0.06 to +1.09)
Schools: all pupils, against others in the authority that year	+0.51 (-0.03 to +1.04)

All point in the expected direction, and at the secondary level the authority estimate for disadvantaged pupils (+3.2) is similar in size to the slope between authorities. But its interval is wide and includes zero. At school level, 156 of 3,121 schools saw their nearest children’s distance change by 10% or more. These are consistent with distance mattering, but do not establish it. New schools also open where populations are growing, so these changes are not random.

What kind of absence?

Table 13: Points per kilometre by kind of absence, all years pooled, allowing for the share disadvantaged and how urban the authority is, with 95% intervals. Persistent absence is the share of pupils missing 10% or more of sessions; the others are shares of sessions.

Group	Kind	Mean	All authorities	Without London
All pupils	Overall	7.7	+1.70 (+1.11 to +2.28)	+0.69 (+0.14 to +1.23)
All pupils	Illness	3.5	+0.28 (-0.02 to +0.59)	+0.25 (-0.11 to +0.61)
All pupils	Unauthorised	2.8	+1.32 (+0.78 to +1.86)	+0.41 (-0.03 to +0.85)
All pupils	Persistent absence	22.1	+5.16 (+3.30 to +7.02)	+1.91 (+0.36 to +3.47)
Disadvantaged	Overall	12.0	+3.54 (+2.36 to +4.73)	+1.52 (+0.39 to +2.66)
Disadvantaged	Illness	4.2	+0.33 (-0.09 to +0.75)	+0.31 (-0.15 to +0.78)
Disadvantaged	Unauthorised	5.5	+2.81 (+1.79 to +3.82)	+1.06 (+0.13 to +1.99)
Disadvantaged	Persistent absence	37.3	+10.45 (+7.18 to +13.72)	+4.25 (+1.53 to +6.98)
Non-disadvantaged	Overall	6.0	+0.58 (+0.26 to +0.90)	+0.30 (-0.07 to +0.68)
Non-disadvantaged	Illness	3.3	+0.23 (-0.03 to +0.50)	+0.23 (-0.09 to +0.54)
Non-disadvantaged	Unauthorised	1.7	+0.44 (+0.16 to +0.73)	+0.10 (-0.16 to +0.37)
Non-disadvantaged	Persistent absence	15.5	+1.67 (+0.53 to +2.80)	+0.67 (-0.60 to +1.94)

For disadvantaged pupils the link is mainly with unauthorised absence (+2.8 of the +3.5 points) and hardly with illness (+0.3). It is also large for persistent absence (+10.5 points of pupils per kilometre, against a mean of 37). The within-authority test agrees: for pupils in the most deprived three tenths of neighbourhoods, +0.32 of the +0.45 is unauthorised. In mostly urban authorities outside London, given ethnic mix and region, it is +1.5 unauthorised and +0.6 illness, and +6.4 points of pupils persistently absent. That is what missed journeys might look like, more than ill health, though the data cannot say why pupils were absent. It matters here because Brighton and Hove’s overall gap with England is mostly coded as illness, while its disadvantaged pupils are high on both.

Does it matter more in towns and cities?

The roadmap asked whether distance matters more in towns and cities, where a longer journey may pass more places to stop, and less in the countryside, where long journeys are normal and often by school transport. Between authorities the answer looks like yes, and strikingly so.

Table 14: Mostly urban authorities outside London (at least 80% of 10 to 14 year olds in urban neighbourhoods), and less urban authorities, all years pooled. Points of absence per kilometre with 95% intervals. All rows allow for the share disadvantaged and how urban the authority is.

Sample	All pupils	Disadvantaged	Non-disadvantaged
Mostly urban, without London	+1.48 (+0.59 to +2.36)	+3.61 (+2.05 to +5.18)	+0.53 (-0.12 to +1.19)
And without Brighton and Hove	+1.31 (+0.34 to +2.28)	+3.50 (+1.70 to +5.30)	+0.32 (-0.32 to +0.96)
And without the five most influential	+1.33 (+0.59 to +2.07)	+4.22 (+3.00 to +5.44)	+0.48 (+0.06 to +0.90)
Mostly urban, without London, given full ethnic mix of this group	+0.58 (-0.25 to +1.42)	+1.74 (+0.26 to +3.21)	+0.23 (-0.41 to +0.88)
Mostly urban, without London, given ethnic mix and region	+1.17 (+0.55 to +1.78)	+2.53 (+1.40 to +3.66)	+0.75 (+0.23 to +1.26)
Mostly urban, without London, given ethnic mix, region and density	+1.27 (+0.71 to +1.83)	+2.80 (+1.77 to +3.83)	+0.81 (+0.32 to +1.30)
Mostly urban, without London, given every control	+0.56 (-0.33 to +1.46)	+1.16 (-0.50 to +2.82)	+0.29 (-0.34 to +0.92)
Less urban authorities	+0.09 (-0.34 to +0.52)	+0.26 (-0.81 to +1.32)	+0.07 (-0.24 to +0.38)
Less urban authorities, given region	-0.08 (-0.41 to +0.25)	-0.03 (-0.94 to +0.88)	-0.07 (-0.30 to +0.15)

In mostly urban authorities outside London, disadvantaged absence rises 3.6 points per kilometre, in every year, and 2.5 given ethnic mix and region. It holds without Brighton and Hove (3.5) and without the five most influential authorities (4.2). In less urban authorities the slope is +0.3, with an interval around zero, in every year. Secondary minus primary tells the same story (+3.2 against -0.2).

Two other tests do not agree, so this is suggestive rather than settled.

Within authorities the slope for the most deprived neighbourhoods is about the same in less urban authorities (+0.52) and in mostly urban ones (+0.48).
By school the slope is similar across kinds of place (Table 17), though the rural interval is wide.

One reading is that an average distance for a whole rural county says little about any one child’s journey, so the authority test may simply not see rural effects. Journey times, and data on who gets school transport, are needed to settle it.

By school, for all pupils

DfE does not publish absence by disadvantage school by school, so this test is for all pupils, and the closest it can come is to ask whether the link is stronger in schools with more pupils on free school meals. For each school, its “nearest children” are those in the Output Areas for which it is the nearest school, and the measure is how far they live from it, typically 1.14 km (a tenth of schools under 0.61 km and a tenth over 2.88 km). It is a stand-in: pupils often go elsewhere. Selective schools, and schools that are nearest for fewer than 100 children, are left out. Models compare schools within the same local authority.

Table 15: How a school’s absence, autumn and spring 2025/26, goes with how far its nearest children live: the change for each doubling of the distance, with 95% intervals, given its pupils’ disadvantage, English as an additional language, prior attainment, special needs, mobility, sex, neighbourhood education, age range and kind of place, within local authority. The last column is how much more each doubling goes with for each extra 10 points of free school meals at the school.

Kind of absence	Mean	Each doubling of distance	Extra, per 10 points of FSM
Overall absence	8.3	+0.25 (+0.13 to +0.37)	+0.08 (-0.00 to +0.16)
Illness	3.7	+0.13 (+0.07 to +0.19)	+0.03 (-0.01 to +0.07)
Unauthorised	3.1	+0.06 (-0.03 to +0.15)	+0.01 (-0.06 to +0.07)
Persistent absence (% of pupils)	22.5	+0.75 (+0.42 to +1.08)	+0.28 (+0.07 to +0.49)

Each doubling goes with 0.25 points more absence, against a mean of 8.3, which is small, as FFT found. It shows up in illness (+0.13) and in persistent absence (+0.75 points of pupils), more clearly than in unauthorised absence (+0.06). It is also stronger in schools with more disadvantaged pupils, which fits FFT’s result and the authority test above.

Figure 13 repeats this for each year, with each year’s school network.

Figure 13: All pupils, by school, each year: overall absence, autumn and spring, for each doubling of how far a school’s nearest children live, and how much more that is for each extra 10 points of free school meals at the school, with 95% intervals, given intake, kind of place and authority. DfE publishes no split by disadvantage at school level. DfE published no 2019/20, and 2020/21 is left out, as on How school absence changes.

The link appears in all 6 years (from +0.13 to +0.32 for each doubling) and the extra for schools with more free school meals is above zero in every year, though not always clearly. The link is about twice as large since the pandemic as in 2018/19, when absence itself was much lower.

The models above leave out ethnic make-up. Since it matters so much between authorities, Table 16 adds each school’s White British share. It takes about a quarter off the slope for overall absence, and the slope stays clear. The extra for schools with more free school meals does not change.

Table 16: All pupils, by school, six years pooled. Each doubling of how far a school’s nearest children live, against other schools in the same authority and year, with and without the school’s White British share (January 2026, held across years), and the extra for each 10 points of free school meals, with that share in the model. 95% intervals; errors clustered by authority.

Kind of absence	Without ethnic make-up	With ethnic make-up	Extra, per 10 points of FSM
Overall absence	+0.34 (+0.24 to +0.45)	+0.26 (+0.16 to +0.36)	+0.14 (+0.07 to +0.22)
Illness	+0.15 (+0.10 to +0.20)	+0.12 (+0.07 to +0.17)	+0.03 (-0.01 to +0.06)
Unauthorised	+0.10 (+0.03 to +0.18)	+0.05 (-0.02 to +0.13)	+0.07 (+0.00 to +0.13)
Persistent absence (% of pupils)	+1.23 (+0.92 to +1.54)	+0.99 (+0.68 to +1.31)	+0.51 (+0.28 to +0.74)

By kind of place. Table 17 splits schools by the same three kinds of place. The authority test is in the section above.

Table 17: Absence for each doubling of how far a school’s nearest children live, by kind of place, with 95% intervals: overall absence in autumn and spring 2025/26 as in the table above, and absence above what intake predicts, 2021/22 to 2024/25.

Kind of place	Schools	Overall absence, 2025/26	Above intake, 2021/22 to 2024/25
Urban, near a major town or city	2,334	+0.28 (+0.15 to +0.41)	+0.25 (+0.12 to +0.38)
Urban, further from one	186	+0.21 (-0.14 to +0.55)	-0.15 (-0.52 to +0.22)
Rural	373	+0.21 (-0.06 to +0.49)	-0.08 (-0.35 to +0.19)

The estimates are similar in size across kinds of place, but only urban schools near a major town or city, where most schools are, give a clear answer. There are fewer schools elsewhere and the intervals are wide, so this cannot show whether distance matters less in rural areas.

And results?

Table 18: Attainment 8 for each doubling of how far a school’s nearest children live, on the Value Added v2.1 specification for each group of pupils, with a school term, fitted on 2021/22 to 2024/25, with and without absence in the model. Percentage of Attainment 8, with 95% intervals.

Group	Without absence in the model	With absence in the model
All pupils	-0.5% (-1.0 to -0.1)	-0.2% (-0.6 to +0.2)
Disadvantaged	-1.1% (-1.7 to -0.4)	-0.5% (-1.0 to +0.1)
Non-disadvantaged	-0.7% (-1.1 to -0.3)	-0.4% (-0.7 to -0.0)

Schools whose nearest children live further away have lower results than their intake predicts, by more for disadvantaged pupils (-1.1% for each doubling) than for non-disadvantaged pupils (-0.7%) or all pupils (-0.5%). Allowing for absence takes away about half of that for disadvantaged pupils. These are small differences, and part of what they measure may be the kind of place rather than the journey.

What it adds up to

Taking the tests together:

Firm: the pattern. Disadvantaged absence goes with distance in every year from 2018/19 and on every measure of distance, and it is much weaker in primary schools.
Firm: most of the slope in Figure 5 is not distance. Ethnic mix accounts for nearly all of the fall across authorities, and is a confounder rather than a step on the path from distance to absence.
Likely: a smaller link of its own. It appears in mostly urban authorities given ethnic mix and region, within authorities for pupils in the most deprived neighbourhoods, and between schools given their intake and ethnic make-up. It shows up in unauthorised and persistent absence. Three different designs, with different weaknesses, point the same way.
Not established: that distance causes it. Network changes point the right way, but the estimates are imprecise and those changes are not random.

Brighton and Hove’s disadvantaged pupils are close to the median authority for ethnic mix (64% White British, against 66%), so ethnic mix does not explain its high absence. Given mix, region, share and urbanity, the city is 2.5 points above expectation, 6th of 150 authorities, and given distance too it is 2.1 above, still 6th. So on this model distance explains only about 0.4 points of why the city stands out.

Its deprived neighbourhoods are 0.59 km further from a school than those of the median mostly urban authority outside London. On the urban estimates, from 1.2 with every control to 2.8 given ethnic mix, region and density, that would go with about 0.7 to 1.7 points of disadvantaged absence, out of the 4.8 by which the city is above England. On the within-authority estimate, for all pupils in deprived neighbourhoods, it would be about 0.2. So distance is a credible part of the city’s problem, and not the main explanation for it. Journey times on the real networks may show more, because Brighton’s hills and bus routes are what straight lines miss most.

Journey times on the real networks

Straight lines miss hills, the sea, street layouts and bus services. So for Brighton and Hove and the eight comparator places, every Output Area was routed to every state secondary school within 15 km with r5r, on OpenStreetMap streets and the Department for Transport’s bus timetables for a school-term Tuesday, 17 November 2026. The method follows the city’s school system model, with a window of departures in place of a single time, and every source and setting is in the methodology. This covers the journey to the quickest school from each neighbourhood, whichever school a child actually attends, and it is for all children: DfE publishes nothing on journeys by disadvantage.

Figure 14: Minutes to the quickest state secondary school, for children in the most deprived third of each place’s neighbourhoods and the rest, walking and by bus, and walking alone, on 17 November 2026. Weighted by 10 to 14 year olds. Without allowing for hills.

Table 19: Minutes to the quickest state secondary school from the most deprived third of neighbourhoods and the rest, the difference, and the share of children over 30 minutes from a school by walking and bus. Dep. is the most deprived third; Rest is the other neighbourhoods. In order of the difference by walking and bus. Without allowing for hills.

	Walking and bus (min)			Walking alone (min)			Over 30 min
Place	Dep.	Rest	Gap	Dep.	Rest	Gap	Dep.	Rest
Brighton and Hove	27.6	22.9	+4.7	36.9	27.9	+9.0	35%	21%
Southampton	20.1	18.8	+1.3	21.0	19.9	+1.1	13%	6%
Hastings	31.3	30.8	+0.6	40.7	36.9	+3.8	55%	44%
Blackpool	19.4	18.9	+0.5	23.2	20.3	+2.9	0%	4%
Plymouth	19.6	19.7	-0.2	21.4	21.5	-0.1	10%	11%
Newham	12.1	12.6	-0.5	12.8	13.6	-0.8	0%	0%
Tower Hamlets	10.2	11.8	-1.6	10.3	12.1	-1.8	0%	0%
Bristol	19.9	21.5	-1.6	20.4	22.9	-2.5	11%	17%
Portsmouth	15.6	17.3	-1.8	19.2	19.3	-0.1	0%	3%

Brighton and Hove stands out on real journeys, as it does in straight lines. Children in its most deprived third of neighbourhoods are 27.6 minutes from a school by walking and bus, and the rest 22.9: 4.7 minutes further. No other place is more than 1.3 minutes further, and in 5 of the eight the deprived neighbourhoods are nearer. Walking alone, the difference is 9.0 minutes, more than twice the next largest (3.8). 10% of children in the city’s deprived neighbourhoods are more than 45 minutes from a school by walking and bus, against 2% of the rest.

Buses narrow the gap but do not close it. The difference falls by about half when buses are available in Brighton and Hove, from 9.0 to 4.7 minutes. In straight lines, measured here from Output Areas, the deprived neighbourhoods are 36% further. In time it is 32% further on foot and 20% with buses. That assumes every family can use the bus, and the fare is not counted.

Hastings looks different on this measure. Journeys there are the longest of the nine for everyone (31 minutes), but buses almost remove the difference between neighbourhoods (+0.6 minutes, against +3.8 on foot). In Hastings, 55% of children in deprived neighbourhoods and 44% of the rest are over 30 minutes from a school, so getting to school is hard across the town, whichever neighbourhood a child lives in.

Figure 15: Minutes to the quickest state secondary school for children in each national IDACI tenth of neighbourhood (1 = most deprived), in Brighton and Hove and the comparator places, where the place has at least 100 children in that tenth. Without allowing for hills.

The children in the city’s most deprived tenth of neighbourhoods are 33 minutes from a school by walking and bus, against 22 minutes in its least deprived tenth.

What the journeys are made of

A long journey is one thing and a journey with a change of bus is another. Getting off one bus to wait for the next is a point where a child can stay off, and one more connection to miss. So the routes were examined for the number of buses, leaving at 07:45, 08:00 and 08:15. These are still journeys to the quickest school, not to the school a child attends, so they understate the changes many children make.

Table 20: How children would reach their quickest school, leaving at 07:45, 08:00 or 08:15, by the most deprived third of neighbourhoods (Dep.) and the rest (Rest). The share walking, the share needing two or more buses, and the average number of buses on each journey. Late if missed is how many minutes after 08:45 a child would arrive if they missed the last journey that gets there in time.

	Walking		2+ buses		Buses each		Late (min)
Place	Dep.	Rest	Dep.	Rest	Dep.	Rest	Dep.	Rest
Brighton and Hove	14%	30%	18%	9%	1.04	0.79	4.1	2.8
Plymouth	42%	49%	4%	5%	0.61	0.56	1.2	1.8
Hastings	23%	29%	3%	8%	0.79	0.79	6.9	8.1
Southampton	52%	59%	3%	1%	0.51	0.42	1.5	1.6
Portsmouth	49%	39%	2%	1%	0.53	0.62	2.8	1.8
Blackpool	28%	46%	2%	2%	0.74	0.56	2.8	1.8
Bristol	57%	51%	1%	1%	0.44	0.50	1.5	1.6
Tower Hamlets	78%	66%	1%	1%	0.23	0.34	0.6	0.8
Newham	63%	60%	0%	1%	0.37	0.41	1.2	1.3

Brighton and Hove’s deprived neighbourhoods need the most changes. 18% of children in its most deprived third of neighbourhoods would need two or more buses to reach even the quickest school, against 9% of the rest. In the other places shown, the highest figure for deprived neighbourhoods is 4%. 14% of those children can walk fastest, against 30% of the rest.

Table 21: Where children in Brighton and Hove would change buses on the way to their quickest school: the busiest stops, by the 10 to 14 year olds whose journey changes there, averaged over the three departures. The nearest stop name is shown. Stops facing in different directions are listed separately.

Stop	Children changing here	From the most deprived third
White Horses	440	254
Open Market	218	103
Sackville Road	193	86
White Horses	153	0
Bottom of The Avenue	74	74
Station Road	72	0
Mithras House	69	69
Swanborough Drive	65	65

This does not show up in absence, so far. Within authorities and given each school’s intake, schools whose nearest children need more buses do not have clearly higher absence (Table 22). That is 118 schools in 13 authorities, and uses journeys to the quickest school as a stand-in for real ones, so it is a weak test. Before distance is allowed for, the share needing two or more buses is clearer (t = 2.6), but that largely goes with distance. None of this rules out an effect on the children who do change buses.

Table 22: Schools in the routed places: absence for one standard deviation more of each feature of the journeys of the school’s nearest children, within authority, given straight-line distance, free school meals, White British share, special needs and prior attainment, with t values. All pupils; DfE publishes no split by school. A guide only, with so few schools and authorities.

Feature of the journey	Overall absence	Unauthorised	Persistent absence (% of pupils)
Share of children needing a bus	+0.13 (t = 0.5)	+0.21 (t = 1.2)	+0.13 (t = 0.2)
Share needing two or more buses	+0.16 (t = 1.6)	+0.10 (t = 0.8)	+0.56 (t = 1.4)
Buses per journey	+0.19 (t = 0.7)	+0.24 (t = 1.3)	+0.33 (t = 0.4)
Minutes late if the last bus is missed	+0.40 (t = 1.6)	+0.29 (t = 1.7)	+0.80 (t = 1.5)
Minutes on the journey	+0.61 (t = 1.7)	+0.30 (t = 1.1)	+2.32 (t = 2.6)
Metres climbed to school	+0.57 (t = 2.0)	+0.31 (t = 1.2)	+2.37 (t = 2.8)
Metres climbed home	+0.72 (t = 2.8)	+0.31 (t = 1.6)	+2.31 (t = 3.0)

School buses and missing them

Many places run dedicated school buses, with a few trips in the morning and a few back in the afternoon. In Brighton and Hove, for example, the 72 and 72A run from Whitehawk to Longhill’s grounds, three trips each morning and three back. Miss one and the alternatives may be poor. These routes can be picked out of the timetables by their pattern, for all of England: at most 8 trips a day, all within school travel hours, ending or starting by a state secondary school and no nearer a college. That finds 2,682 route and school pairs serving 1,000 schools. Timetables do not say who a bus is for, so this will include a few commuter journeys and miss school contracts that are not in them.

For each school bus in the routed places, Output Areas within 400 m of a stop on its morning trips were routed to its school at every minute from 07:00, with every bus and walking available. That gives the last time a child can leave and still arrive by the bell, and how late they are if they miss it. Bells are not in the data, so the bell is set 15 minutes after the last school bus arrives.

Table 23: Brighton and Hove schools with school buses: children living near a stop on one of them. Trips is morning school bus trips. Bell is inferred, 15 minutes after the last school bus. Uses it is the share whose last workable journey uses a school bus. Late is minutes late if that journey is missed, by the next best, also shown with a bell 10 minutes later. Walking is minutes at 3.6 km/h. Absence is all pupils, autumn and spring 2025/26.

School	Trips	Bell	Uses it	Late	Late, later bell	Walking	Absence
Longhill High	5	08:19	31%	16.5	6.9	64	13.9%
Portslade Aldridge	2	08:33	21%	4.7	5.6	65	9.6%
Dorothy Stringer	4	08:22	57%	4.1	1.6	63	10.5%
Blatchington Mill	1	08:10	37%	3.9	6.2	42	10.7%
King's	3	08:15	77%	3.6	0.8	48	6.4%
Cardinal Newman Catholic	5	08:39	27%	3.6	2.6	52	8.8%
Patcham High	2	08:35	35%	3.5	3.7	64	9.0%

Longhill stands apart in the city. A child living near a stop who misses the last workable journey there arrives about 17 minutes late, against 3 to 5 minutes for the other city schools, and walking would take about 64 minutes. It also has the city’s highest absence. The result depends on the bell: at 10 minutes later the penalty is 7 minutes. The inferred 08:19 is close to the 08:20 in the school’s 2023 consultation, which should be confirmed for today. That document also notes that the school has one of the earliest starts in the area, and that a later start would go ahead only if the authority could provide the school buses, so the school day itself is set around them.

Table 24: Plymouth, and nearby schools its children can reach by school bus, on the same basis as the Brighton and Hove table. Inferred bells are less reliable where a route arrives late in the morning, as at Millbay.

School	Trips	Bell	Uses it	Late	Late, later bell	Walking	Absence
Notre Dame RC	3	08:55	49%	8.4	10.4	82	10.3%
Tor Bridge High	2	08:20	80%	6.9	15.0	81	9.9%
Coombe Dean	4	08:40	64%	6.8	2.5	80	7.8%
Devonport High School for Boys	3	08:50	43%	3.3	11.3	58	3.7%
Millbay	1	09:23	5%	3.2	2.6	51	10.8%
Callington Community College	1	08:30	0%	2.7	23.0	120+	9.5%

Longhill is a striking case, not a pattern. Across the 21 schools in the routed places with school bus figures and published absence, the correlation between the penalty and absence is 0.03, and for unauthorised absence -0.05. Several schools with larger penalties than Longhill’s have low absence, usually with fewer disadvantaged pupils. Nor is having a school bus linked with absence across England (Table 25). Having a bus is not the same as depending on one, and it often means the school has made sure that children can get there, so a null here does not mean that missing one is harmless. Still, these tests give no general support.

Table 25: Secondary schools in England, absence at schools with a school bus against other schools in the same authority, given distance, intake, ethnic make-up and kind of place, with t values. 3,076 schools, autumn and spring 2025/26, all pupils. High free school meals is the top quarter of schools.

Kind of absence	All schools	Urban	Rural	High free school meals
Overall absence	+0.00 (t = 0.0)	-0.02 (t = -0.3)	+0.07 (t = 0.6)	+0.12 (t = 0.5)
Illness	+0.06 (t = 1.5)	+0.05 (t = 1.1)	+0.06 (t = 0.7)	+0.17 (t = 1.5)
Unauthorised	-0.05 (t = -1.1)	-0.07 (t = -1.4)	+0.06 (t = 0.5)	-0.12 (t = -0.7)
Persistent absence (% of pupils)	+0.09 (t = 0.4)	+0.06 (t = 0.3)	+0.21 (t = 0.5)	+0.65 (t = 1.0)

The climb

Straight lines also miss the hills. r5r’s own elevation setting averages each street in both directions and did not change walking times in our tests, so the climb was measured directly, from OS Terrain 50 along each walking route, separately for the way to school and the way home.

Table 26: Walking to the quickest school: metres climbed on the way there and on the way home, for the most deprived third of neighbourhoods (Dep.) and the rest (Rest), and the walking time to school with Tobler’s hiking function, scaled to 3.6 km/h on the flat.

	Metres up, to school		Metres up, home		Minutes to school
Place	Dep.	Rest	Dep.	Rest	Dep.	Rest	Dep., flat
Hastings	71	74	67	57	47.2	43.7	42.6
Brighton and Hove	61	43	62	34	42.8	32.3	38.4
Plymouth	41	33	32	35	25.7	24.8	22.4
Bristol	19	25	16	24	22.5	25.7	21.7
Southampton	18	18	13	14	22.8	21.8	21.9
Portsmouth	7	5	12	6	20.7	20.6	20.5
Blackpool	10	8	9	7	24.2	21.5	24.0
Tower Hamlets	3	4	3	3	11.1	13.0	11.1
Newham	3	4	2	4	14.2	15.1	14.1

In Brighton and Hove, children in the most deprived third of neighbourhoods climb about 62 metres on the walk home, against 34 for the rest. That is the largest difference of the places here. Those children have both further to go and more to climb, which a straight line cannot show.

A signal in the routed places did not hold across England. In the schools of Table 22, more climb went clearly with more absence, more strongly than anything else tested. With so few schools that needed checking. Climb can be measured for all of England without routing, along the straight line from each of 178,605 Output Areas to its nearest school. In the routed places that matches the climb along the real route closely (r = 0.91).

Table 27: Secondary schools in England, absence for each extra 10 metres climbed on the way home by a school’s nearest children, along a straight line, in the same authority, given distance, intake, ethnic make-up and kind of place, with t values. 3,076 schools. All pupils, autumn and spring 2025/26, except the last row.

Kind of absence	All schools	Urban	Rural	High free school meals	Without Brighton and Hove
Overall absence	+0.018 (t = 1.1)	+0.028 (t = 1.0)	+0.020 (t = 1.0)	-0.057 (t = -0.6)	+0.017 (t = 1.0)
Illness	+0.017 (t = 2.0)	+0.031 (t = 2.1)	+0.014 (t = 1.6)	+0.060 (t = 1.3)	+0.016 (t = 1.9)
Unauthorised	-0.006 (t = -0.5)	-0.017 (t = -0.9)	+0.002 (t = 0.2)	-0.123 (t = -2.3)	-0.007 (t = -0.6)
Persistent absence (% of pupils)	+0.041 (t = 1.0)	+0.063 (t = 0.8)	+0.061 (t = 1.3)	-0.257 (t = -0.9)	+0.036 (t = 0.9)
Above intake, 2021/22 to 2024/25	+0.001 (t = 0.0)	+0.006 (t = 0.2)	+0.018 (t = 0.9)	-0.004 (t = -0.0)	-0.001 (t = -0.1)

Across England the link for overall absence is small and not clear: +0.018 points for each extra 10 metres (t = 1.1), and within authorities by home neighbourhood +0.056 (t = 1.4). There is a small link with illness absence (t = 2.0), clearer for schools near major towns and cities, and none with unauthorised absence. Between authorities the link for disadvantaged pupils is larger, given their ethnic mix (+0.34 per 10 metres, t = 3.0; non-disadvantaged pupils +0.15, all pupils +0.18), and falls by about half given region. So the first result looks largely local, with at most a small general link. Straight lines do understate climb, so a modest effect could still be there.

Does time say more than distance about absence?

This cannot yet be answered. Routing covers only 8 authorities with published absence. Table 28 sets them side by side, for all three groups of pupils, without modelling them.

Table 28: Authorities routed so far: distance and journey time from the most deprived third of neighbourhoods, and absence in state-funded secondary schools, 2024/25. DfE publishes absence by disadvantage for authorities, so Hastings has none.

Place	Straight line (km)	Walking and bus (min)	All pupils	Disadvantaged	Non-disadvantaged
Hastings	1.66	31.3	–	–	–
Brighton and Hove	1.60	27.6	10.8%	18.0%	8.0%
Southampton	0.88	20.1	10.5%	15.7%	6.8%
Bristol	0.85	19.9	10.3%	15.8%	7.2%
Plymouth	0.85	19.6	10.2%	16.4%	7.4%
Blackpool	1.03	19.4	9.6%	13.2%	6.1%
Portsmouth	0.86	15.6	10.1%	15.5%	6.4%
Newham	0.54	12.1	7.0%	8.3%	5.5%
Tower Hamlets	0.42	10.2	6.5%	7.6%	5.1%

DfE’s file by home neighbourhood gives a little more, by comparing tenths within these authorities (144 groups of pupils, all pupils, 2023/24 and 2024/25). Distance and time point the same way and are hard to tell apart: for one standard deviation more, overall absence is 0.56 points higher on straight-line distance, 0.64 on walking time and 0.58 on walking and bus time. Persistent absence is 1.7, 2.0 and 2.0 points of pupils. With so few authorities these are descriptive. A proper test needs routing for many more: the 109 mostly urban authorities would allow the models above to be rerun on time.

How far to trust these times

Against the city’s model. On that model’s settings, times for 165 Brighton and Hove neighbourhoods correlate at 0.93 with the city’s school system model, with a mean of 23.4 minutes against 22.1 and a typical difference of 2.3 minutes. The difference between deprived neighbourhoods and the rest is 4.0 minutes here and 2.9 there. The model uses an older timetable, a set list of schools and an earlier departure for Longhill, and these have not been separated. Its simulator’s 4 minute difference is another measure again: it follows children to the school they are allocated and compares disadvantaged children with others.
Walking speed. r5r’s default is 3.6 km/h, slow for a teenager, and is kept to match the city’s model. At 4.8 km/h Brighton and Hove’s difference with buses is 4.8 minutes, so the result does not depend on it.
Buses only. Rail, the Tube, the DLR and school-only services that are not in the timetable files are missing. That matters most in Newham and Tower Hamlets, where it would shorten journeys, not lengthen them.
One day. Departures from 07:45 to 08:15 on one Tuesday, taking the median. Real journeys depend on when a school starts.
The quickest school is not always the school attended. Children go further than this, more so where choice or catchments send them elsewhere. Who can get a place at the quickest school is not considered.
Boundaries. Each place is a district, with schools up to 15 km beyond it. Hastings, Blackpool and others have few schools nearby, and Plymouth and Bristol’s networks include neighbouring counties.

What if a school were somewhere else?

Straight lines make this quick to test. These keep every neighbourhood and its children where they are and move only the school.

Table 29: Brighton and Hove with a school in a different place. Straight-line distance to the nearest school, and where the difference would rank among 293 districts. The Longhill and CoMArt sites are those in the city’s school system model.

Scenario	Most deprived third (km)	Other neighbourhoods (km)	How much further	Rank
Today	1.57	1.18	+33%	5 of 293
A school back on CoMArt's site	1.19	1.08	+10%	27 of 293
Longhill moved to the top of Elm Grove	1.21	1.16	+5%	37 of 293

Either change would take about 0.36 km off the distance from the city’s deprived neighbourhoods and close most of the gap, which fits the closure of CoMArt in 2005 having left the east without a school nearby. Moving Longhill leaves the average for other neighbourhoods much the same, though those near its present site would be further from a school. This is not a forecast of where children would go or how they would travel: the school system analysis models that, along with the costs.

What this cannot say, and what comes next

Straight lines are not journeys. Hills, the sea, bus routes and bus fares are all missing, and they may matter most in places like Brighton and Hove.
The nearest school is not the school attended. Catchments, admissions and choice decide that. FFT measured from pupils’ own postcodes; here distance describes where schools and children are.
These are comparisons between places and schools, so they show patterns, not what would happen to one child’s attendance if their journey changed. DfE publishes absence by disadvantage only down to authority level, and the file by home neighbourhood does not split by disadvantage.

Next on the roadmap:

Journey times beyond the comparator places: the mostly urban authorities, so that the absence models can be rerun on time rather than distance.
Better school-level measures. A gravity model of who goes where would improve on “nearest children”, until a published measure of how far each school’s pupils travel exists.
Whether time predicts absence better than distance, whether the link between authorities and within them comes closer together, and whether the rural and urban difference holds.
A direct test of intervening opportunities: points of interest from OpenStreetMap within a couple of hundred metres of bus changes and along walking routes. The routes and change points are ready for it.