n.V1-3.01| THE STARTING STATE
01| A school is judged on what comes out of it. Test scores, pass rates, graduation percentages, inspection grades — the whole accountability apparatus measures the exit. Almost nothing measures the entrance. Not the neighbourhood, which everyone accepts as background scenery, but the actual condition of the child sitting in the chair at nine in the morning. Did she sleep. Did she eat. Did she arrive settled.
02| The proposition is not controversial in principle: a person who arrives unrested, underfed and under stress will not learn at the level they are capable of. It applies to everyone. What makes it consequential is that these conditions are not randomly distributed. They cluster, and they cluster by income. That turns a universal fact about human learning into a specific mechanism for unequal outcomes. The same lesson, delivered identically to the whole class, produces different results because the students were not in the same state when it landed.
03| Sleep does more than keep a student awake. Material studied and then slept on is retained substantially better than material studied and then not — consolidation is part of the learning, not a precondition for attending it. The applied evidence here is unusually clean. Districts that shifted school start times later saw attendance and grades improve, with the gains concentrated among the students who had been sleeping least. That is close to an experiment on the preparation phase alone, curriculum and teaching held constant. Sleep in a poor household is disrupted by crowding, noise, family shift work, unstable housing and unsafe streets. None of it is a choice the student makes.
04| Food costs attention and behaviour in the short run, and cognitive development in the long run — iron deficiency most clearly, and it tracks income closely. The evidence on school feeding is more mixed than its advocates present. Effects on attendance and enrolment are consistent and substantial. Effects on learning outcomes are smaller and less reliable. The honest reading is that food removes a barrier to being present and attentive rather than improving learning directly. A feeding programme should be judged on what it actually delivers, which is a child in the room.
05| Stress does two things. It occupies capacity that would otherwise be available for the task in front of the person, and over years it appears to affect the development of the systems that manage attention and self-regulation. The first is demonstrable under controlled conditions: inducing financial worry measurably degrades cognitive performance in the same individuals, at magnitudes comparable to a night without sleep. The second comes from longitudinal work linking the duration of childhood poverty to working-memory performance in adulthood.
06| Two conditions belong on this list and are almost never on it. Housing instability: frequent moves are associated with lower achievement independent of income, and each move costs continuity of instruction, of relationships, and of the routines that make sleep and study possible. And crowding and noise: chronic noise exposure is associated with impaired reading acquisition, and crowding removes the physical possibility of quiet study. Both are direct functions of housing cost. Neither appears in any measure of student effort.
07| A classroom is a system, not a collection of individuals, and the preparation phase does not stay with the child it belongs to. Carrell and Hoekstra found that adding a small number of children from households with domestic violence to a classroom lowered achievement and raised misbehaviour for every student in it, at magnitudes large enough to matter for policy. The direction of that effect is documented. Its shape is not. A threshold claim — that a room tips once some share of students arrives in this state — and a gradient claim — that each additional affected student costs something — fit the same evidence and imply opposite interventions. A threshold justifies targeting a few. A gradient justifies improving conditions for all. Nobody has settled which it is, and it is the most useful thing anyone could settle here.
08| What follows is a problem for measurement before it is a problem for policy. If the preconditions vary this much and matter this much, then scoring a school on output alone scores its intake as much as its institution. The strong version of the objection — that ignoring the preparation phase makes the institution complicit in the result it reports — is a claim about responsibility, and it should be argued as one rather than smuggled in. The weaker version is enough: output-only measurement produces systematically wrong conclusions about which schools and which methods work. It rewards intake and it penalises effort.
09| What is not known is the size. Nobody has produced the number that would matter most — the share of the attainment gap attributable to the state the child arrives in rather than to what happens after. Until that number exists, this establishes that a mechanism is present, not that it is the dominant one. We think it is large. We cannot prove it, and we would rather say so than pretend otherwise.
