Crowdsourcing Reliable Local Data

The adage “All politics is local” in the United States is largely true. Of the United States’ 90,106 governments, 99.9% are local governments. Despite variations in institutional features, descriptive representation, and policy-making power, political scientists have been slow to take advantage of these variations. One obstacle is that comprehensive data on local politics is often extremely difficult to obtain; as a result, data is unavailable or costly, hard to replicate, and rarely updated. We provide an alternative: crowdsourcing this data. We demonstrate and validate crowdsourcing data on local politics using two different data collection projects. We evaluate different measures of consensus across coders and validate the crowd’s work against elite and professional datasets. In doing so, we show that crowdsourced data is both highly accurate and easy to use. In doing so, we demonstrate that nonexperts can be used to collect, validate, or update local data.