Would I lie to you? On Private Equity Intermediary Performance Reports

Abstract

In Would I lie to you? On Private Equity Intermediary Performance Reports (2025), Tamayo et al. compile a new international dataset of over 20,000 interim performance reports from 1,769 private equity investments, leveraging over 600,000 sentences of qualitative commentary. While existing research has examined interim Net Asset Values (NAVs), the accompanying qualitative disclosures remain underexplored. Using FinBERT to analyze report tone alongside quantitative metrics (sales, EBITDA, multiples), the authors find that narrative tone is a robust predictor of future investment performance. Investments with fully positive tone achieve approximately 0.7x higher multiple uplift (MOIC) until exit versus neutral tone. Unlike interim valuations, positive tone consistently predicts performance across geographies and institutional settings. Machine learning confirms substantial out-of-sample predictive power. Results indicate fund managers strategically employ qualitative information to convey expectations, consistent with double agency dynamics where both GPs and LPs balance transparency with reputational concerns.

Introduction

Private equity managers provide periodic performance reports containing both NAV valuations and qualitative written commentary for each portfolio company. While NAVs are known to be overly smooth and potentially biased, the informational value of accompanying textual disclosures remains largely unexplored. The challenge lies in understanding whether GPs use narrative tone to convey material information that complements or substitutes for quantitative valuations.

Tamayo et al. address this gap by analyzing whether the tone of GP reports contains forward-looking information about investment performance and how this relates to the well-documented smoothing of interim valuations.

Methodology

Dataset:

20,000+ semi-annual reports from European and US buyout investments (2000–2024), containing:

• 600,000+ sentences of qualitative commentary (~13 sentences per NAV figure).
• Quantitative metrics: sales, EBITDA margins, debt, enterprise value.
• Performance measures: MOIC (Multiple on Invested Capital) and MOIC uplift (Exit MOIC minus Interim MOIC).

Tone analysis:

• Primary: FinBERT neural network classifying sentences as negative (-1), neutral (0), or positive (1).
• Alternatives: FinRoBERTa and Loughran-McDonald dictionary for robustness.
• Average tone: 0.25 (moderately positive), more positive in the US (0.34) than in Europe (0.24).

Machine learning:

Lasso regression and Random Forest models trained on qualitative and quantitative data to predict investment outcomes out-of-sample.

Key Findings

• Tone predicts performance: A one-standard-deviation increase in tone yields 0.23x higher MOIC uplift, while the same increase in interim MOIC only adds 0.02x. Tone remains significant throughout the investment life; interim MOIC predicts uplift only in the first two years.
• Short-term validation: 44% of subsequent mark-ups are preceded by positive tone; 39% of mark-downs are preceded by negative tone, confirming tone’s informational content.
• Geographic consistency: Tone predictiveness holds across all geographies, while the interim NAV-performance relationship shows opposing patterns between Europe and the US, suggesting universal use of narrative signaling.
• Fundraising dynamics: During fundraising, the tone-uplift correlation is stronger for established GPs who leverage reputational capital through qualitative signals. Emerging managers emphasize NAVs instead.
• Machine learning success: ML models achieve 0.79 ROC-AUC predicting outperformance one year post-investment. Top-quartile predicted investments achieve 1.93x MOIC uplift versus 0.31x for the bottom quartile.

Implications for Practice

The findings reveal a sophisticated disclosure strategy where GPs complement potentially biased NAVs with more accurate qualitative signals:

• LPs should systematically analyze report tone alongside valuations for better performance prediction.
• Established GPs use narrative as a credible signaling mechanism, especially during fundraising.
• The 0.7x MOIC uplift difference between positive and neutral tone represents substantial economic value.
• A regulatory push for independent valuations may be warranted given GPs’ apparent ability to produce better NAVs.

Conclusion

The paper demonstrates that private equity managers strategically use narrative tone to convey material forward-looking information not reflected in interim valuations. This behavior aligns with double agency theory, where both GPs and LPs navigate competing incentives between transparency and reputation management. While NAVs are smoothed for various strategic reasons, written reports serve as a scalable channel for disseminating soft information to hundreds of LPs.

The research highlights a fundamental inefficiency: despite containing substantial predictive information, qualitative disclosures remain underutilized by investors. Machine learning tools could help LPs systematically extract these signals, potentially improving capital allocation decisions and monitoring of portfolio companies across the $8 trillion private equity industry.