Can Starving Start‐ups Beat Fat Labs? A Bandit Model of Innovation with Endogenous Financing Constraint*

Abstract

Is there any such thing as too much capital when it comes to the financing of innovative projects? We study a principal–agent model in which the principal chooses the scale of the experiment, and the agent privately observes the outcome realizations and can privately choose the novelty of the project. When the agent has private access to a safe but non‐innovative project, the principal starves the agent of funds to incentivize risk‐taking. The principal quickly scales up after early successes, and can tolerate early failures. If the principal is equally informed about the outcome, then the agent is well‐resourced, resembling a large research and development department.

Wealth Inequality and the Exploration of Novel Alternatives

“…First, it relates to many theoretical papers studying the incentives for experimentation using principal-agent models (e.g. Bergemann and Hege, 1998, 2005, Manso, 2011, Hörner and Samuelson, 2013, Bouvard, 2014, Drugov and Macchiavello, 2014, Halac et al, 2016, Gomes et al, 2016, Spiganti, 2020a. I share with Drugov and Macchiavello (2014), Halac et al (2016), and Gomes et al (2016) a focus on adverse selection on the agent's ability to experiment: whereas we all highlight that inefficiencies in the credit market may lead to misallocation in experimentation, our aims are different.…”

“…Indeed, these papers analyse the problem of a principal hiring an agent to work on a project of uncertain quality, whereas this paper focus on a multiple agents economy to focus on the consequences of misallocation at the aggregate level. Similarly to Manso (2011), Drugov andMacchiavello (2014), andSpiganti (2020a), this paper focuses on the tension between exploration and exploitation, and its implications in a principal-agent frameworks. In the spirit of Weitzman (1979), these models study innovation, interpreted as the discovery of superior actions through experimentation and learning.…”

“…The first is the repetition of the conventional work method in any contingency C C C , which is usually referred to as "exploitation" in the bandit problem literature (e.g. Manso, 2011, Spiganti, 2020a. I refer to such an entrepreneur as an "exploiter" and to such a firm as a "conventional firm".…”

“…Since this action plan is traditionally referred to as "exploration" (e.g. Manso, 2011, Spiganti, 2020a, I refer to an entrepreneur who sets up an "explorative firm" to engage in exploration as an "explorer". Hereafter, exploiters are indicated with an index C (mnemonic for conventional), talented explorers are denoted by H, whereas untalented explorers are denoted by L.…”