Relatively little research has explored whether schools differ in their effectiveness for different group of pupils (e.g. by ethnicity, poverty or gender), for different curriculum subjects (e.g. English, mathematics or science) or over time (different cohorts). This paper uses multilevel modelling to analyse the national test results at age 7 and age 11 of over 6,000 pupils attending 57 mainstream primary schools over three successive years in a socially and ethnically diverse inner London borough. The pupil groups with the poorest progress were White British pupils on Free School Meals (FSM) and Black Caribbean pupils, both those entitled and those not entitled to FSM. Differences between schools in average pupil progress were large, but there was no evidence of differential school effectiveness in relation to FSM, ethnicity or gender, all pupil groupings benefitted from attending the more effective schools and to a broadly similar extent. More effective schools "raised the bar" but did not "close the gap" suggesting that differences between schools in "quality" plays little role in equity gaps. While school residuals for different pupil groups were extremely highly correlated, school residuals were only moderately correlated across subjects (English, mathematics and science) and over time, with particularly poor stability for English.