Luka Vulicevic

Luka Vulicevic

PhD Student in Finance

Rady School of Management
University of California San Diego

I am a second-year PhD student at UC San Diego studying finance. My research combines computer science and econometrics to address questions in asset pricing and financial economics. Currently, I am focused on Large Language Models—developing methods to address challenges like lookahead bias and statistical inference—to improve our understanding of financial markets and the role of alternative data. Prior to the PhD, I worked as a quantitative analyst in Toronto focusing on alpha research and asset allocation. I hold an MFE from the University of Toronto.

Research Interests

Asset Pricing Large Language Models Econometrics
Google Scholar CV GitHub Twitter Email

Working Papers

Compute, Complexity, and the Scaling Laws of Return Predictability

with Allan Timmermann 2026

Investors' computing power ("compute") governs how much signal they can extract from high-dimensional data. Drawing on insights from LLM training, we show that forecast performance follows scaling laws—stable power-law functions between accuracy and training compute—that pin down both an irreducible bound on return predictability and the rate at which extra compute closes the gap. In firm-level cross-sectional return prediction, scaling laws explain over 80% of performance variation across models. Treating compute as a primitive yields sharp economic implications: scaling laws quantify predictability limits and the value of new data, they distinguish problems with strong versus weak returns to scale, and they deliver a market-efficiency metric by mapping computational advantage into a certainty equivalent. Measured this way, computational superiority earns sizable rents: a 25% marginal increase in compute would have raised an investor’s Sharpe ratio by about 10% over the last 30 years, indicating substantial returns to computational sophistication.

PDF

Entity Neutering

with Joseph Engelberg, Asaf Manela, & William Mullins 2026

Cutting-edge LLMs are trained on recent data, creating a concern about look-ahead bias. We propose a solution called entity neutering: using LLMs to find and remove all identifying information from text. Our procedure uses an LLM agent that iteratively (i) masks entity-related terms and (ii) paraphrases the text until an independent LLM fails to recognize the target company from the text. In a sample of over 250,000 financial news articles we verify that, after neutering, ChatGPT and other LLMs identify the subject firm at the rate of random chance. Among the unidentified articles, the sentiment extracted from the raw text and the neutered text agree more than 90% of the time and have similar return predictability, with the difference providing an upper bound on look-ahead bias. The evidence here suggests that LLMs are able to effectively neuter text while maintaining semantic content. For look-ahead bias, LLMs can be both the problem and the solution.

PDF Code
Presentations
NBER Summer Institute 2025 Finance Research Revolution Conference Monash Applied Young Economists

Miscellaneous

Quotes I Like
"We have to forgo some mental tortures to deal with it, but I will accept it and see what happens."
— Gerolamo Cardano, on inventing complex numbers in "Ars Magna"

This quote constantly resurfaces in my mind when doing theoretical work. When solving a problem and approaching a breakthrough, I often feel the most lost and confused. I've learned to embrace this feeling of being 'lost' and to keep progressing regardless, just to see what happens.

"I'd compare stock pickers to astrologers, but I don't want to bad-mouth the astrologers."
— Eugene Fama

Although markets are likely not efficient, they are undeniably difficult to beat. Given this challenge, I find it puzzling why industry professionals often exude such confidence in their work. A marginal unit of effort spent on sales yields far greater returns in the industry than in 'alpha' research.

"There is no such thing as Alpha, only the Beta you don't know about."
— John Cochrane (attributed)

This idea has always stood out and compelled me to think deeply about asset pricing. It has, without a doubt, influenced my own interest and research in asset pricing.

"Buybacks are divisive. They divide people who understand finance from those that don't."
— Ken French

I don't care for the topic as much as I care about the delivery. An eloquent shutdown. I can think of many topics which can replace the first word of this quote—can you?

"I saw the plot first, ... and the first thing I thought was: programming error."
— Richard Roll, on making Figure 2 of FFJR (1969)

This is the first thought an empiricist should have when examining their results. Especially when they are as beautiful and clear as FFJR (1969).

"The past appears to be a predictably unpredictable predictor of the future."
— Benjamin
"Your Grandmother Is Wrong."
— Jonathan Berk, "Money Management in Equilibrium" (2019)
Why We Choose to Study Economics

No person can fully grasp how far and how fast economic thought has traveled, but condense, if you will, the 50,000 years of human economic activity into a time span of but a half-century.

Stated in these terms, for the first 40 years, we know very little—except that at the end of them, man had learned to barter one thing for another. Then about 10 years ago, we invented money and discovered we could store value in metal coins. Only 5 years ago did we establish the first banks and learn that trust could be an asset. The joint-stock company—less than 18 months ago. Adam Smith wrote The Wealth of Nations and gave us the invisible hand just last year. The Industrial Revolution—which transformed everything—came last summer. Marx wrote Das Kapital last fall. The Federal Reserve was created last spring. Keynes published his General Theory and revolutionized our understanding just 8 months ago. Bretton Woods and the modern international monetary system—6 months ago. Credit cards arrived 3 months ago. We floated our currencies off the gold standard 10 weeks ago. The first derivatives appeared 5 weeks ago. The internet enabled digital commerce just last month. The 2008 financial crisis happened two weeks ago last Tuesday. Bitcoin was invented 10 days ago. And this morning, as we woke, artificial intelligence began trading securities faster than human thought.

This is a breathtaking pace, a pace that makes us dizzy. And now we stand at a crossroads, asking not whether our economy can grow, but whether our wisdom can match our innovation—whether we can understand these forces—whether we can master them.

We choose to study economics, not because it is easy but because it is hard.

— Made by Claude Sonnet 4.5 when asked to roleplay JFK being asked why study economics during a speech at Rice.

Technology

Server: Apple Mac Studio, M3 Ultra, 500GB Unified Memory

Working with LLMs is hard. It is even harder when you want to work with larger models. This machine allows me to quickly (and cheaply) test new ideas. I would not recommend it for many people but it has been a game changer for my research.

Computer: Apple MacBook Pro, M2, 16-inch

My first Mac and it is my baby. I never thought of using/buying a Mac, but with the recent chip and integration with the rest of the Apple ecosystem: this thing is awesome.

Tablet: Apple iPad Pro M2

I used to be pretty 'old school' when doing math: pen, paper, and a massive binder. Being able to search handwritten documents is a theorist's life hack.

Source Control: Azure DevOps (private) & GitHub (public)

DevOps is complicated with a high learning curve, but the ability to integrate into VMs and blob storage is unbeatable. The project management abilities also help with building big research infrastructure. GitHub is the easiest way to share code publicly.

IDE: PyCharm

It has a bunch of features which makes coding faster. I have encountered several circumstances where PyCharm becomes schizophrenic and requires resetting, but—so far—the functionality is worth more than this nuisance.