Sequential Matching of Entrepreneurs to Accelerators and Venture Capitalists

From edegan.com
Revision as of 15:23, 7 April 2019 by Ed (talk | contribs) (→‎Summary)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search
Academic Paper
Title Sequential Matching of Entrepreneurs to Accelerators and Venture Capitalists
Author Ed Egan, Jeremy Fox
RAs James Chen
Status In development
© edegan.com, 2016


Files and Dbases

The primary dbase is vcdb2 The main SQL script is

E:\McNair\Projects\MatchingEntrepsToVC\DataWork\MatchingEntrepsV2.sql

Summary

The objective of this project is to create a multi-stage matching model and estimate this model using data on entrepreneurs that match to accelerators and (lead) venture capitalists.

See Fox (2008) - An Empirical Repeated Matching Game Applied to Market for a brief write up on a relevant theory paper by Jeremy

Simple Outline of Model

As of now, the goal is to simulate a repeated matching model with dynamically optimizing agents. More specifically, there are two sides for a matching market with transferable utility (generically, call these men and women for now) with a continuum of agents, but a finite number of types. They participate in matches for T periods and receive utility that is a sum of a structural component (determined solely by their type and the type they are matched with) and a individual taste component (with some known distribution).


What distinguishes this model from a static matching model is that the agents have some probability of transitioning between types that is conditional on the match they make in the previous period (e.g., a man of low type might be more likely to change into a man of high type after being matched to a woman of high type). When making these matches, the agents take these transition probabilities into account when evaluating expected future utility. This adds a dynamic element to the model.

Work to do

  • Code up three algorithms to simulate match(primal,IPFP,dual)
  • Compare with R code from NYU

Work done so far

  • Coded up primal and IPFP (may have errors)

Work to do in near term

  • Compare with R code from NYU (using current solvers/optimizers, and with Gurobi)