Benjamin W. Arold
ETH Zurich



(Wednesday, 22nd May 2024)

Title : Measuring the Language of Contracts: An Application of Natural Language Processing to Labor Economic

This workshop introduces natural language processing tools for measuring the content of contracts. We apply our method to extract worker rights in collective bargaining agreements (CBAs) and provide evidence that workers attach a pecuniary value to these rights. Applying text-as-data algorithms to a new collection of 30,000 CBAs from Canada in the period 1986-2015, we parse legal obligations (e.g. “the employer shall provide...”) and legal rights (e.g. “workers shall receive...”) from the contract text. We validate that worker rights clauses provide amenities and give workers control over the work environment. Companies that provide more rights to workers also score highly on a survey indicating pro-worker management practices. In an empirical analysis motivated by a model of efficient collective bargaining over wages and rights, we find that worker rights increase with local labor income tax rates, while union wage premiums decrease, consistent with workers valuing rights. Further, when worker outside options (local sectoral employment rates) increase, CBAs include more worker rights.