Abstract Background Many studies have reported an inferred causal association of income poverty with physical health among children; but making causal inference is challenging due to multiple potential sources of systematic error. We quantified the short-run effect of changes in household poverty status on children’s health (asthma and ear infections) and service use (visits to the doctor and parent-reported hospital admissions), using a national longitudinal study of Australian children, with particular attention to potential residual confounding and selection bias due to study attrition. Methods We use four modelling approaches differing in their capacity to reduce residual confounding (generalized linear, random effects (RE), hybrid and fixed effects (FE) regression modelling) to model the effect of income poverty (<60% of median income) on health for 10 090 children surveyed every 2nd year since 2004. For each method, we simulate the potential impact of selection bias arising due to attrition related to children’s health status. Results Of the 10 090 children included, 20% were in families in poverty at survey baseline (2004). Across subsequent years, ∼25% experienced intermittent and <2% persistent poverty. No substantial associations between poverty and child physical health and service use were observed in the FE models least prone to residual confounding [odds ratio (OR) 0.94, 95% confidence interval (CI) 0.81–1.10 for wheeze], in contrast to RE models that were positive (consistent with previous studies). Selection bias causing null findings was unlikely. Conclusions While poverty has deleterious causal effects on children’s socio-behavioural and educational outcomes, we find little evidence of a short-run causal effect of poverty on asthma, ear infections and health service use in Australia.