ECONOMICS · PUBLIC POLICY · MICROSIMULATION · AI

Research

WORKING PAPER · 2026

Who Bears the AI Shock? The Distributional and Fiscal Incidence in the UK

What an AI shock costs the UK depends on who bears it. Passing occupation-level AI exposure through PolicyEngine UK on Family Resources Survey 2024–25 data, I find displacing 7% of employees costs the Exchequer £18.2bn a year and raises poverty by 1.75 points. Holding that shock fixed but changing who is displaced swings the cost from £14.9bn to £32.6bn while poverty barely moves — and delivering the same loss through wage cuts rather than job losses almost erases the poverty rise.

WORKING PAPER · 2026

Optimal Allocation of Public R&D under Structural Change

with Seyed M. Hosseini (Harvard University)

Paper Slides

How should public R&D be allocated across sectors as economies shift from manufacturing to services? Using micro-level firm data (1950–2020) and a shift-share IV, we find a 1% rise in public-R&D spillovers raises value added per worker by ~1.55% in manufacturing but only ~0.36% in services. Embedding these estimates in a multi-sector growth model, structural change has roughly halved aggregate productivity growth — and optimal policy should steer public R&D toward the growing sector, weighted by how well it absorbs spillovers.

CONFERENCE PAPER · IMA BRUSSELS & ECOMOD LUXEMBOURG 2026

A Firm-Level Microsimulation for VAT Policy Analysis

Paper Code

An open, fully reproducible firm-level microsimulation for costing reforms to the UK VAT registration threshold — released in full, including its synthetic data generator, as open-source code. Building synthetic firm records from ONS business-structure counts calibrated to HMRC VAT statistics, it reproduces HMRC's official costing of the £85,000→£90,000 reform, documents firm bunching at the notch (bunching ratio ≈ 0.06, implied turnover elasticity ≈ 0.45), and characterises the notch's £21,250-wide dominated region.

CONFERENCE PAPER · ECOMOD 2026 · SUBMITTED TO JOSS

policyengine: A Microsimulation Tool for Tax-Benefit Policy Analysis

with Max Ghenis, Nikhil Woodruff, Pavel Makarchuk, Anthony Volk (PolicyEngine)

Code

An open-source Python package for tax-benefit microsimulation. It encodes UK and US tax and benefit rules as code, letting researchers and analysts estimate how reforms reshape household incomes, poverty, and inequality — with reproducible, transparent, and auditable computations.