Understanding Communications by Private Fund Managers

A large literature on delegated asset management has analyzed the role of information disclosure, especially in situation where managers have significant discretion about reporting on investment choices, performance, and risk. Managers might have incentives to disclose or withhold information strategically, either obscuring or overstating the performance of transactions. This problem is especially prevalent in private capital markets where the underlying assets typically lack continuous market prices and often are difficult for outsiders to scrutinize.

In Communicating Performance: Textual and Financial Disclosures in Private Equity, Josh Lerner, Fiona Paine, and Antoinette Schoar study how general partners (GPs) in private equity funds provide information to their limited partners (LPs) about the performance of their portfolio investments. The paper focuses on two primary channels through which managers can communicate information about the performance of portfolio firms—financial reporting on the valuation of the portfolio firms and more qualitative textual reports.

Conceptually, why might the information conveyed via these two channels differ? At the early stages of a company investment the valuation of a portfolio company might not reflect all the embedded option values and risks due to market power or information asymmetries. In addition, early valuation increases may bring modest immediate benefits because management fees usually depend on committed capital and carried interest arrives after an exit. Inflated values can have costly consequences: some LPs make decisions on spending or distributions using net asset value, and reversing those decisions is difficult. Publicly visible changes might also reveal information to competitors or weaken the GP’s position in later financing negotiations. Narrative reports may let GPs share their assessments without all the consequences of changing a reported mark. GP seeking to raise another fund could use these narratives to exaggerate prospects, conceal setbacks, or keep investors uncertain about its ability. Yet systematic misstatements risk reputational damage when outcomes become visible. Accurate early discussion may therefore signal a manager’s skill at assessing companies, even when some investments fail.

The authors analyze year-end GP reports received by StepStone, a major private equity investor, covering 7,271 company-year observations held by 536 US funds during 2015–2020. They combine company discussions and valuations with deal and exit data. A finance-specific dictionary, adjusted for private equity language, measures sentiment as positive minus negative words relative to report length, then standardizes it. A topic model classifies sentiment-bearing sentences into areas such as cash flow, business model, capital structure, and macroeconomic conditions.

The central result relates to the evolution of where predictive information appears. During the first three years of the investment, the sentiment conveyed in the text about the company has a great deal of information. A one-standard-deviation increase in reported sentiment is associated with roughly three to four percentage points higher annualized final return (controlling for the interim return). That is substantial beside the sample’s mean final return of about seven percent. But sentiment’s association with ultimate returns weakens after years three and four and disappears by year five.

In contrast, the predictive power of interim returns increases over time: the regression coefficient rises from about 0.16 early in an investment’s life to 0.62 in years three and four and 0.78 later. Both favorable and unfavorable words convey early information about ultimate performance, so the result is not merely a measure of optimism. The authors interpret the pattern as private assessments appearing first in text and increasingly in valuations as uncertainty resolves.

Several tests demonstrate the robustness of the results. Among companies observed at least twice, early sentiment has signficant predictive power, but sentiment measures from later in the investment’s life do not. This results suggests the results are not driven by a differing mixture of investments over time. Controlling for whether a valuation changed recently leaves the early sentiment result largely intact. Limiting the influence extreme returns does not eliminate the patterns: the results are not driven by a few outliers. Follow-on financing provides an external price signal: in years three and four, sentiment becomes less informative around or after such rounds. The reduction is not statistically clear in the earliest years, when uncertainty may remain high even after financing.

These findings support the idea that text and valuations can complement each other as information becomes measurable over time.

Content matters as well as tone. Early reports focusing on cash flow and business models are associated with better final performance, while a capital-structure focus is associated with worse performance. Positive or negative language about cash flow, business models, and macroeconomic conditions is particularly predictive early on; leadership sentiment carries little information. The result suggests that a report’s subject can reveal what risks or opportunities a GP sees. A macroeconomic comment may be informative because it describes how a broad shock affects a particular company, beyond what the interim mark captures.

The broader lesson is that disclosure quality cannot be judged from valuations alone. Conservative valuation marks may coexist with narratives whose tone conveys useful, forward-looking information. For limited partners, the paper makes a case for analyzing narrative and financial reports together, while leaving open the extent to which these patterns extend to other asset classes.

Lerner, Josh and Paine, Fiona and Schoar, Antoinette, Communicating Performance: Textual and Financial Disclosures in Private Equity (September 27, 2026). Available at SSRN: https://ssrn.com/abstract=7532840.