Purpose: The purpose of this study is to further develop an emerging thread of quantitative research that grounds investment in school infrastructure in a unified theoretical framework of investment in human, social, and physical capital. Design/methodology/approach: To answer the research question, what is the impact of investment in human, social, and physical capital on student achievement, the author uses canonical analysis, a multivariate statistical approach that allows for multiple independent and dependent variables. The null hypothesis is selected given the limited body of research on this question, and the state is selected as the unit of analysis. Level of student poverty is added as a control variable given an extensive body of research that supports its negative impact on achievement. Descriptive statistics are generated as well as a Pearson product moment correlation matrix to diagnose and address potential issues of multicollinearity and simultaneity. Three national databases are used: United States Census Bureau, US Department of Education's National Assessment of Educational Progress test score data, and the US Department of Education's Common Core of Data. Years analyzed are 2003, 2005, and 2007. Findings: The results of the canonical analysis are robust, statistically significant, and consistent over time. Investment in human, social, and physical capital accounts for between 55.8 and 77.2 percent of the variation in student achievement in fourth and eighth grade Reading and Mathematics. Investment in human capital is consistently the largest influence on student achievement followed by social and physical capital. The null hypothesis is rejected. Originality/value: This study advances the use of theory in explaining the impact of investment in school infrastructure on student achievement, a feature which distinguishes it from much previous research in this domain. The use of a theory also addresses a major weakness of traditional, atheoretical education production function research. In addition, utilization of canonical analysis rather than multiple regression to operationalize the theoretical model and analyze the data represents an advance in research design and statistical analysis for this type of research question. (Contains 5 tables and 1 note.)