Thesis title: Essays on Growth Accelerations, Knowledge Transmission, and the Hidden Costs of Artificial Intelligence
This thesis studies economic growth and the effects of technological change on the welfare of economic agents. Specifically, it analyzes the empirical determinants of growth accelerations, the role of productive public spending and international knowledge transmission in convergence processes, and the welfare implications of adopting artificial intelligence (AI) in the presence of externalities in product markets.
Chapter 1 examines accelerations in GDP per capita growth using a large panel dataset of countries since the postwar period. After identifying acceleration episodes and rigorously addressing the time-series properties of the data from an econometric perspective, the chapter shows that physical capital accumulation significantly predicts accelerations, but robustly only in economies that are already relatively capital-abundant.
Chapter 2 develops a multi-country AK model with international knowledge transmission in a leader–follower setting. The core mechanism is an absorptive-capacity threshold, measured as capital per capita relative to the leader: below the threshold, productive public spending supports long-run growth and spillover effects are limited; above the threshold, knowledge spillovers become operative and drive convergence, while public spending no longer shows long-run effects on growth rates. The empirical evidence, based on Dynamic Common Correlated Effects estimators, identifies a statistically significant threshold and shows that spillovers weaken as the technological gap narrows.
Chapter 3 proposes a theoretical model that formalizes a new mechanism documented so far mainly in experimental settings, but potentially relevant at the macroeconomic level as well. Moving to the technological frontier, it studies a trade-off: AI can raise productivity, but it can also make products more similar. In a general-equilibrium model with monopolistic competition and heterogeneous firms, AI operates through an “ecosystem” channel (lower fixed costs, more entry, more variety) and a “monoculture” channel (lower perceived distinctiveness, a lower welfare value of variety). The thesis characterizes when monoculture costs can dominate and discusses the optimal corrective policy.