Predicting the Geospatial Economic Impact of New York City Capital Investments

According to the Preliminary Capital Budget Fiscal Year 2016, the City of New York’s total expenditures for capital projects in FY2016 is $8.6 billion. These expenditures will be invested in vital capital projects, such as repaving sidewalks or creating green spaces, and to improve livelihood in the boroughs. These investments are anticipated to be fully funded through the city’s Capital Budget and carried through by the Capstone’s client, the New York City Department of Design and Construction (“DDC”).

The Capstone team examined the impact of a Staten Island capital project on surrounding property sale values. The project goal was to develop a rigorous methodology to quantify neighborhood outcomes of a chosen capital project type. Specifically, the team focused on projects that involved exterior renovation, upgrading, or new construction. The properties involved in measuring project impact were all owner-occupied homes in Staten Island, New York.

The impetus of this analysis is to create a methodology that can be used to study how NYC capital projects can benefit neighborhoods. Large, high profile projects often involve powerful stakeholders that have the capacity to make a very compelling case. However, lower profile capital projects may have a better return on investment. The Capstone project will help policymakers create a robust tool to identify the potential benefits of infrastructure projects on a neighborhood.

Drawing on prominent literature in the field, such as a study conducted by Ellen, Schill, Susin and Schwartz (2001), the analysis relied on a hedonic regression model integrated with a difference-in-difference approach (“DID”). This approach compared the prices of properties in small rings surrounding investment project sites with prices of properties in the same census tract but outside the ring. As a result, the team concluded that the Great Kills Library project increased property sales prices in a predefined surrounding radius of 2,000 feet by 6.5%, on average. When assessing robustness, the team also employed measures of statistical significance and analysis of variance, including potential sources of modeling error, such as multicollinearity, heteroskedasticity, and autocorrelation.