Private equity managers' report language tracks final returns in a holding's first three years

The authors' account of why text leads valuations rests on incentives. According to them, raising a valuation early does little for the GP up front: management fees are usually set on committed capital, and carried interest is paid after an exit ^[11]. Some LPs base spending or distribution decisions on net asset value, and those decisions are hard to undo ^[12]. Changes made in public view could also hand information to competitors or weaken the GP's hand in later financing talks ^[13]. A narrative report may let a GP share its assessment without all the consequences of changing a reported mark ^[14].

The effect is sizable against the base. A one-standard-deviation rise in sentiment is associated with roughly 3 to 4 points of higher annualized final return in the first three years, controlling for the interim return ^[4], on a sample mean final return of about 7 percent ^[5]. That is about 43 to 57 percent of the mean ^[19]. The valuation channel moves the other way. The interim-return coefficient rises from about 0.16 early in a holding to 0.62 in years three and four and 0.78 later ^[7], almost five times the early weight ^[20]. The authors read this as private assessments appearing first in text and increasingly in valuations as uncertainty resolves ^[9].

The obvious objection is that a sentiment score only measures cheerfulness. The authors report that both favorable and unfavorable words convey early information about ultimate performance ^[8]. A second test points the same way: among companies observed at least twice, early sentiment has significant predictive power and later-life measures do not ^[10]. The incentive to inflate exists. A GP seeking another fund could use narratives to exaggerate prospects, conceal setbacks or keep investors uncertain ^[15]. Against that, the authors note that systematic misstatements risk reputational damage when outcomes become visible ^[16], and that accurate early discussion may signal skill even when some investments fail ^[17].

The evidence has limits. Every report came to one investor, StepStone, from US funds in 2015 to 2020 ^[2], and sentiment is a dictionary count of positive minus negative words relative to report length ^[3]. The post summarizing the paper does not say whether the pattern holds in reports sent to other LPs.

For an LP, the sequencing is plain. In a holding's first years the mark carries the smaller coefficient ^[7], so the prose is where added information sits, and in our view a reviewer who skips it is leaving out something the study finds adds to the mark ^[4]. By years three and four the mark's coefficient has risen to 0.62 ^[7]. For a GP, the wording in an early update is a record. Later marks and the final return will show whether it was accurate, and the reputational cost the authors describe applies when those outcomes become visible ^[16].