China VC Paper's Causal Evidence Questioned as 2SLS Coefficient "Balloons 20-30x" OLS Estimate

A recently revised working paper, "Learning by Investing: Entrepreneurial Spillovers from Venture Capital," exploring China's venture capital ecosystem, suggests that high-net-worth (HNW) Limited Partner (LP) investors often "learn by investing" and subsequently launch more and higher-quality ventures. However, analyst Jonathon P Sine has voiced significant concerns regarding the paper's causal evidence, particularly its Instrumental Variable (IV) strategy.

The paper, authored by Josh Lerner, Jinlin Li, and Tong Liu, and revised in December 2025, posits that LPs in successfully launched VC funds go on to start a greater number of improved ventures compared to LPs in funds that failed to launch. These new entrepreneurial endeavors are notably concentrated in high-tech sectors, suggesting a direct learning mechanism from their investment experiences. The researchers utilized comprehensive administrative data on entrepreneurial activities and VC fundraising in China to support their theory.

Despite acknowledging the "astoundingly good" data collection and the intuitive theory, Jonathon P Sine expressed skepticism about the study's methodological rigor. "The causal evidence here seem unreliable / unconvincing," Sine stated in a social media post, adding that the IV strategy, "like a lot of IV designs (cough AR cough), is clever but hard to evaluate (at least to me) and probably needs to be taken with a very large grain of salt."

Sine further highlighted a critical discrepancy: while the paper's "simple descriptive stats themselves do show a shift in type of startup (more high-tech after becoming LPs)... BUT show no real difference in the number of startups founded by LPs in launched vs. failed funds." The significant effect on venture founding only emerges after the IV is applied, leading to the 2SLS coefficient inflating "to 20–30× the OLS estimate. Which seems worrying?" This substantial difference raises questions about the robustness and interpretation of the causal claims, suggesting that the observed entrepreneurial uplift might be an artifact of the statistical method rather than a direct, verifiable impact.