Observational Unit in Portfolio Data
In portfolio-based estimation, data granularity is not merely a measurement detail but part of the structural mapping.
Read the draft →
I am a Finance PhD candidate at INSEAD.
Before academia, I co-founded two companies: one failed rather miserably, and the other raised more venture capital than customers. Those experiences left me curious about how investors think, how capital gets allocated, and why a company's brilliance or stupidity—including my own—can look like noise next to the mighty hand of markets.
That led me to a period of quiet contemplation at an asset management firm, and eventually to pursuing a PhD.
I now work on asset pricing, portfolio choice, and financial econometrics.
Observational Unit in Portfolio Data: Implications for Asset Demand Estimation
I show that in logit demand systems, dispersion in mandate level primitives does not average out, but enters the consolidated demand object through higher-cumulant curvature. Moving estimation closer to the mandate level recovers downward-sloping estimates without sign restrictions, and price elasticity 2–4 times larger than in consolidated 13F implementations.
Portfolio Granularity and Structural Inference
I show that the statistical power of an observed portfolio does not scale with its raw number of holdings, but with its concentration structure. I study how this property can be used for structural inference with portfolio data.