Accounting for data uncertainty: Biases in web-scraped Chinese aid data

Most foreign aid research uses data donors report to the OECD’s Development Co-operation Directorate (OECD DAC). In the last two decades, China has become a major donor but neither reports to the OECD DAC nor publishes its own figures. Accounting for Chinese aid is important since China follows a different strategy than other large donors and so could undermine attempts to use aid to leverage change in regimes with poor governance or economic policy. A widely used dataset on Chinese development finance assembled by AidData, a research lab at the College of William and Mary, employs an army of undergrads to scour the web for information on individual Chinese aid projects from media and other sources. This leads to the question: Does this novel approach introduce systematic bias, if, for example, web-scraping by predominantly English-speaking undergraduates works better for countries with more press freedom, more postings in English, or more internet access? In cases with known underreporting, I explore what statistical methods can incorporate this information to yield more reliable results.