Whether increases in school inputs lead to better education outcomes is essential for
education policy-making. While empirical studies that aim to estimate this causal relationship
abound, a consensus is lacking. One confounding factor that has mostly been ignored in the
literature is the behavioral response of households to changes in public education policies by
varying their own education inputs. This paper studies how increases in government spending
on public schools affect household education spending in urban China. We use detailed
information about household spending on public school tuition, textbooks, and private
tutoring from the 2002-2006 Urban Household Survey data and focus on households with
children in compulsory education (Grades 1-9). We provide evidence that municipal public
education spending in China is exogenous to household preferences, and this allows us to
identify a causal relationship. In a model controlling for city fixed effects, year-province
fixed effects, local per capita GDP and growth rate of number of basic education students,
and a wide range of household characteristics, we find that increases in public education
spending lead to significant decreases in household spending on public school tuition and
private tutoring but no change in spending on textbooks. In addition, while the reduction in
household spending on tuition is quite homogeneous across income quintiles, the reduction in
household spending on private tutoring comes primarily from the lowest income households,
suggesting diverse educational demand. Moreover, the impact on private tutoring spending
differs for metropolitan areas and smaller cities.