Using Linguistics To Mine Unstructured Data From Fasb Exposure Drafts

Accounting standard setting is a high-stakes, political, and market process influenced by constituents through a public commenting mechanism. Comment letters are widely studied by researchers and the Financial Accounting Standards Board (FASB), typically manually because the letters contain unstructured data. Our study employs a topic modeling method, latent Dirichlet allocation (LDA), to overcome the difficulties posed by the unstructured data. We analyze comment letters on two exposure drafts proposed by the FASB in 2008 and 2010 regarding loss contingencies. Results show that LDA is effective in compiling information from unstructured data. LDA also enables us to identify topics and detect shift in focus of the letters responding to the two exposure drafts. The findings have practical implications for standard setters, regulators, and researchers while also contributing to the digital reporting, data analysis, economic theory of democracy, and coalition and influence theory literatures.