| dc.description.abstract | Knowledge produced in a firm is a source of competitive advantage, as well as a currency which can be exchanged both inside the firm and with other firms. Patents are a key mechanism by which firms protect the new knowledge they produce: intellectual property rights enable a startup to enter a market, a sole inventor to create a firm in the absence of capital or customers, and a small multinational subsidiary to increase its significance in a large network of subsidiaries. This three-chapter dissertation analyzes how firms use knowledge they produce, specifically how multinational subsidiaries inventing technologies interact with their multinational headquarters and their local partners; how cutting-edge methods derived from machine learning and natural language processing can enable study of otherwise intractable problems in codifying and transferring knowledge; and how startups use patents strategically, with a focus on implications of intellectual property policy. This dissertation stands at the intersection of the fields of entrepreneurship, innovation, and machine learning.
Chapter 1 introduces a model for the relationship between the multinational firm’s headquarters, its subsidiary, and the host country of the subsidiary. The model, loosely based on the gravity trade model and featuring a measure of knowledge distance introduced here, yields an answer to a longstanding topic in the multinational literature, namely whether a multinational subsidiary in a foreign country gravitates towards its host or continues the strategy of its headquarters. The findings include a relative shift in the influence of the headquarters and host country over the subsidiary as the subsidiary grows to closer to the host, as well as the result that a highly specialized skill temporary migration visa can increase influence of the headquarters over the subsidiary when utilized. The results are relevant for both multinational managers and governments hosting multinationals.
Chapter 2 surveys key machine learning methods applied to management research, and dives especially into natural language processing applications. Applications include analyses of the patent corpus, topic modeling, and sentiment analysis. The perspectives in this chapter are relevant to the study of knowledge and broadly firm strategy, as tools from machine learning can create new measures of knowledge, transfers, and firm strategy; or improve existing ones.
The third chapter analyzes a policy shock to startup firms as a window to studying the value of reducing uncertainty in the patent examination process. Startups especially benefit from granted IP rights, as often their IP is the basis for venture funding and market entry. As the duration of the examination process is uncertain, firms treated with accelerated patenting yield significantly improved outcomes. Methodologically, the paper also adds a matching algorithm based on natural language processing to standard econometric techniques. | |
