About
I am an economist at the International Finance Corporation (IFC), World Bank Group, where I work on the impact measurement and monitoring of IFC's financial sector projects. My work focuses on measuring the development impact of investment in emerging markets.
My research sits at the intersection of AI economics, Labor economics, development finance, and credit markets. I study how artificial intelligence is reshaping labor markets, firm productivity, and financial access — with a particular focus on the Global South, where the implications differ sharply from patterns documented in the US and European literature.
I hold a PhD in Economics and have peer-reviewed publications in Economic Inquiry, Oxford Economic Papers, and Global Economy Journal. My regional expertise spans West Africa, Latin America, South and Southeast Asia, and the Middle East and North Africa.
I built GAIA (Global AI Adoption Index) as a contribution to the empirical infrastructure for AI economics research.
Research agenda
AI & Credit Markets
How generative AI exposure affects firm credit conditions, bank lending behavior, and the financial accelerator channel in emerging markets.
Development Impact Measurement
Automating AIMM-style scoring using LLMs. AI as a tool for development finance due diligence and portfolio monitoring.
Algorithmic Divide
Cross-country evidence on differential AI adoption — why low-income countries adopt AI through education rather than production.
Government Credit & Monetary Policy
How fiscally backed credit programs interact with monetary policy transmission — evidence from Brazil's credit registry.
GAIA Dataset
GAIA provides two distinct occupation-level AI exposure scores — GAIA-E (generative-AI era, Eloundou et al., Science 2024) and GAIA-B (pre-generative-AI / supervised-ML era, Brynjolfsson et al., AEA 2018) — plus Anthropic's observed Claude.ai behavioral data and a country-level panel covering 178 countries. The two exposure scores are kept separate because they correlate at only 0.19, capturing distinct technology paradigms.
Occupation Explorer → Country Explorer →Data access by request. Cite as: Aghabarari (2026) DOI: 10.5281/zenodo.20320112
The data, briefly
GAIA-E distribution
923 occupations · index points
Country adoption
usage × work use · size = GAIA-A
Measure correlations
Pearson r across occupations
GAIA-E = gaia_e · E1+E2 = dv_rating_beta (Eloundou 2024) · GAIA-B = sml_score (Brynjolfsson 2018) · Anthropic obs. = AI autonomy. Source: gaia_occupations.csv / gaia_countries.csv.
Publications & working papers
Working papers in AI economics, labor markets, credit markets, and measurement. Each paper has its own page with figures, abstract, and citation.
Methodology
Anthropic Economic Index — observed usage
Downloaded from Hugging Face (Anthropic/EconomicIndex). Task-level Claude.ai usage aggregated by occupation and country. Matched to O*NET task descriptions. SOC-level exposure = employment-weighted sum of task usage percentages.
OpenAI E1+E2 — theoretical exposure
Eloundou et al. (Science 2024) replication data. E1+E2 = share of occupation tasks performable by GPT-4 plus software tools. Main specification in empirical work. SOC 2018 codes, 923 occupations.
Brynjolfsson SML — pre-GPT baseline
Recomputed from raw CrowdFlower ratings (ICPSR 114436). 2,069 direct work activities scored by crowd workers on 23 rubric questions. Aggregated via O*NET Tasks-to-DWAs bridge. Used as falsification test.
GAIA-E — Generative-AI Era Exposure
Min-max normalization of OpenAI E1+E2 (dv_rating_beta) to [0, 1] across all 923 occupations. Captures how exposed an occupation is to large language models with software tools. Source: Eloundou et al. (Science 2024).
GAIA-B — Pre-GenAI Era Exposure (Supervised ML)
Min-max normalization of the SML score to [0, 1] across 605 occupations with non-missing data. Captures suitability for pre-GPT machine learning as of 2018 — a conceptually distinct baseline that correlates only 0.19 with GAIA-E. Source: Brynjolfsson, Mitchell & Rock (AEA 2018).
Contact
Citation
Aghabarari, L. (2026). GAIA — Global AI Adoption Index (Version 1.0). Zenodo. https://doi.org/10.5281/zenodo.20320112