Delusions of Grandeur and Their Benefits (and Hazards)

Technical Summary: Delusions of Grandeur and Their Benefits (and Hazards)

Problem and Motivation
This paper investigates the economic ramifications of entrepreneurial overconfidence within a population-wide tournament. The central question is how agents' beliefs about their own search processes—specifically, the "delusions of grandeur" where agents believe their projects have a positive drift despite the true drift being zero—affect aggregate output and the distribution of wealth. The authors frame this through the lens of Adam Smith's observation that "the over-weening conceit which the greater part of men have of their own abilities... is an ancient evil," yet noting that such confidence drives individuals to crowd into professions despite unfavorable odds. The study seeks to determine whether these delusions are merely a source of inefficiency or if they serve a productive function in innovation, and how they impact inequality.

Methodology
The authors construct a continuous-time model of competition among a continuum of ex-ante identical entrepreneurs.

• The Environment: Agents engage in costly experimentation, modeled as a driftless Brownian motion $X_t = \sigma W_t$. Each agent can stop at any time $\tau$ to implement the best option discovered so far, denoted $M_\tau = \sup_{0 \le s \le \tau} X_s$.
• Payoffs: Agents care about both absolute success (the quality of the best discovery) and relative success (their rank in the population). The payoff function is $\phi_K(m) = m + y(K(m))$, where $K$ is the conjectured cumulative distribution function (CDF) of the population's outcomes.
• Belief Asymmetry (The Core Mechanism): While the true process has zero drift, agents are misspecified. Each agent believes their own process has a positive drift $\mu > 0$ ($X_t = \mu t + \sigma W^\mu_t$). Crucially, agents know that others also believe they are special, but they believe others are wrong while their own confidence is warranted.
• Equilibrium Concept: The paper defines a "regular drawdown equilibrium." Agents do not stop based on the absolute level of their current project but on the "drawdown"—the gap between their current value and their personal best ($D_t = M_t - X_t$). An agent stops when this drawdown exceeds a threshold $d(M_t)$. The equilibrium requires that the conjectured rank distribution $K$ matches the realized distribution generated by the optimal stopping rule under the true law.

Key Contributions and Characterization
The paper's primary technical contribution is reducing the equilibrium fixed-point problem (involving both the stopping boundary $d$ and the distribution $K$) into a single, tractable ordinary differential equation (ODE).

1. Dimensionality Reduction: By using the quantile index $t$ (where $K(m) = 1 - e^{-t}$) as a coordinate, the authors derive a scalar ODE for the drawdown boundary $\delta(t)$:
   $$\delta'(t) = \delta(t) - \frac{\mu}{c} \frac{\delta(t) + b(t)}{1 - \exp\left(-\frac{2\mu}{\sigma^2 \delta(t)}\right)}$$
   where $b(t)$ captures the marginal incentive of rank.
2. Existence and Uniqueness: The authors prove that for a given level of optimism $\mu$, there exists a unique regular drawdown equilibrium characterized by a unique bounded positive solution to this ODE.

Key Results
The paper establishes two main comparative statics regarding the level of optimism $\mu$:

1. Output and Persistence (The Benefit):

   • Higher optimism ($\mu$) increases the tolerance for drawdowns. Optimistic agents are willing to endure larger setbacks before quitting because they perceive the future as more promising.
   • This leads to longer experimentation times in the true economy (under the true law $P_0$).
   • Consequently, the distribution of realized discoveries shifts to the right in the sense of first-order stochastic dominance (FOSD). Overconfidence raises the average quality of discoveries and total output.

2. Inequality (The Hazard):

   • The improvement in outcomes is not a parallel shift. The additional persistence generated by optimism is most valuable for agents who are already high in the realized distribution.
   • Agents who survive many early setbacks (the "winners") benefit disproportionately from the willingness to continue, pushing the upper tail of the distribution further out than the rest.
   • Mathematically, the variance of realized outcomes strictly increases with $\mu$. Thus, delusions of grandeur increase inequality.

Empirical Evidence
The authors test two implications of the model using cross-country data:

1. Entrepreneurship and Inequality: There is a positive correlation between the share of entrepreneurs and measures of inequality (Gini coefficient and 90/10 percentile ratio), even after controlling for GDP per capita and population.
2. Beliefs vs. Risk Preferences: The mechanism relies on optimistic beliefs about one's own prospects, not a general taste for risk. The data shows that standard measures of risk tolerance (e.g., equity market participation, lottery sales, uncertainty avoidance) do not explain entrepreneurship or the entrepreneurship-inequality link. In contrast, survey measures of "belief in own abilities" are strong predictors of entrepreneurship.

Significance and Claims
The paper claims to offer a belief-based channel linking entrepreneurship to wealth inequality. Unlike existing literature that focuses on borrowing constraints or ex-ante heterogeneity in ability, this model demonstrates that even with identical agents, the endogenous persistence driven by overconfidence can generate both higher aggregate output and greater inequality.

The authors modestly conclude that while the evidence is correlational, it supports the interpretation that entrepreneurship is tied more closely to self-assessed ability (optimism) than to intrinsic risk appetite. The "delusions of grandeur" are productive in that they sustain search and innovation, but hazardous because the resulting gains are unevenly distributed, exacerbating inequality.