Close Ties in VC Don’t Pay Off

New research challenges conventional wisdom surrounding deep relationships in venture capital.

Bigstock photo

Bigstock photo

Close relationships don’t always pay off in venture capital, according to new research from Hong Kong Polytechnic University and the University of Oxford.

Deeper relationships between venture capital firms can lead to lower exit performance, and as a result, erode ties between VC managers co-investing in deals, said Qianqian Du, an assistant professor at Hong Kong Polytechnic, and Thomas Hellmann, an Oxford business school professor, in a paper this month. They analyzed the largest 50 VC firms in the U.S. to determine how their relationships fared overtime.

“Our findings about the negative effects of relationships also challenge some of the received wisdom that tends to merely think of relationships as assets not liabilities,” Du and Hellmann said in the paper. They analyzed co-investment decisions made by firms such as Lightspeed Venture Partners, Draper Fisher Jurvetson International, and Warburg Pincus from 1985 to 2012.

The researchers looked at how frequently the firms on their top-50 list had the same investment targets and used their findings to determine which firms had “deeper” relationships. They found that VC firms that co-invested more frequently in the past were less likely to invest together in the future.

[II Deep Dive: Luck and Reputation Matter in Venture Capital. Skill Doesn’t.]

A deep relationship may fail for market-related reasons, according to the paper.


Relationships built during “hot markets,” or times when the market is generally doing well, tend to have a negative effect on co-investment behavior, the researchers found. Yet 75 percent of co-investments are made in hot markets, they said in the paper.

“The analogy of this in interpersonal relationships would be that friendships forged in hard times (e.g., “wartime friends”) are stronger than those forged in easy times (e.g., “party friends”),” Du and Hellmann said in the paper.