Research

This page brings forward earlier academic research from my PhD-era economics site. For more recent applied work, the portfolio also includes posts, publications, and technical briefings elsewhere on the site.

Working Papers

AI According to HANK: Redistribution, Monetary Transmission, and Stabilization

This paper studies the AI transition as a redistributional technology shock in a heterogeneous-agent New Keynesian model with liquid and illiquid assets. A rise in the share of automatable tasks raises productivity, but it also lowers the labor share and shifts current income toward households with lower propensities to spend. In the model, that redistribution can generate a contraction in demand and a decline in the natural rate even as potential output rises. The key difference from tractable TANK benchmarks is that policy transmission depends on the joint distribution of labor-income exposure, liquid wealth, debt, and dividend ownership. I build a full-distribution HJB/KFE benchmark with a sticky-price transition block and use it to study monetary easing, labor-income support, and targeted transfers during the AI transition. The quantitative results show that TANK remains informative about the sign of the redistribution channel, but materially understates the role of balance-sheet heterogeneity for output-gap dynamics, inflation pressure, and welfare incidence. Policies that support current labor income or target low-liquid-wealth households dominate passive adjustment, but their relative ranking is state dependent and shifts with rebalancing frictions and the distribution of capital income. These findings imply that macroeconomic policy for AI cannot be designed from average productivity effects alone; it must be designed around the household distribution that receives, spends, and finances the gains from automation.

Boom or Slump? The AI Transition, Redistribution, and Stabilization in a Heterogeneous-Agent Economy

This paper studies how AI-driven automation can simultaneously raise potential output and weaken demand in a tractable New Keynesian model with household heterogeneity. AI is modeled as an increase in the share of production tasks that can be performed by capital and is embedded in a TANK environment with constrained high-MPC households and unconstrained asset holders. Because constrained households remain more exposed to labor income than to capital income, automation redistributes current cash flow away from the households with the highest propensity to spend. The result is a negative output gap even when actual output eventually rises with capital accumulation and higher productivity. Relative to worker-capitalist environments, debt-service exposure makes monetary policy more powerful, while sticky wages and a lower-bound constraint limit stabilization through interest rates alone. The paper’s main contribution is to provide a transparent mechanism linking AI adoption, redistribution, and stabilization policy in a small heterogeneous-agent framework. Quantitative illustrations, interpreted as reduced-form benchmark exercises rather than as fully disciplined policy evaluation, suggest that labor-cost support can outperform untargeted transfers in the benchmark calibration because it supports high-MPC demand while easing inflationary pressure. The analysis is intended as a tractable step toward richer quantitative work on AI and heterogeneous-agent stabilization.

Firm Dynamics with Firm-specific Intangible Capital

This paper studies simple models of firm investment under uncertainty, which often lack an internal mechanism that propagates shocks. It introduces a time-to-build feature in firm-specific intangible investment to generate internal propagation. Because this intangible capital is firm-specific, there is no market for trading it across firms, so firms must produce it internally over time using factor inputs. Firms with a higher intangible share exhibit a more hump-shaped impulse response and a more delayed peak even without explicitly adding adjustment-cost terms. In this setting, internal intangible capital accumulation becomes a real friction, a micro-foundation for adjustment costs, and a Q-theory-style mechanism for firm investment. The model also introduces a time-varying wedge between factor prices and marginal products, with the degree of firm-specificity determining the size of the wedges.

Price-setting under Bounded Rationality

I present a theoretical model in which price-setting firms choose what to pay attention to and how much attention to allocate, subject to cognitive limits, in a sparsity-based bounded-rationality setting in the spirit of Gabaix (2014). Unlike rational inattention models such as Sims (2003), where firms reset prices every period with a fixed amount of attention, this framework allows prices to remain unchanged for a time even when adjustment is physically costless, and it allows attention to vary across conditions. Inaction arises when both aggregate and idiosyncratic conditions move little enough that firms allocate zero attention to them. Firms allocate more attention to conditions that are more variable. The model qualitatively matches micro price data in which prices respond quickly and strongly to idiosyncratic shocks but only slowly and modestly to nominal shocks. It also implies that nominal shocks can have strong and persistent real effects when aggregate uncertainty is sufficiently low.

Research Areas

  • Firm dynamics
  • Macroeconomics
  • Asset pricing
  • Natural language processing in economics and finance