Overview
Accurately estimating enterprise value (EV) is a critical task in corporate finance, driving decision-making in mergers and acquisitions (M&A), investment evaluations, and financial reporting. Traditional methods such as discounted cash flow (DCF) and market multiples are commonly used but often suffer from subjectivity and limited applicability across different industries and company types.
We introduce a regression-based model for enterprise value estimation using a comprehensive dataset of all publicly listed U.S. companies from Cap IQ. Our regression approach addresses several limitations of traditional valuation methods by providing:
- Objectivity: A data-driven methodology that removes subjective biases.
- Transparency: Clear interpretation of the impact of each financial metric on enterprise value.
- Scalability: Applicability across diverse sectors and company sizes.
Our model has been validated using out-of-sample testing and diagnostic checks to ensure accuracy and robustness. Such testing includes both public and private companies.
This regression-based approach offers a valuable alternative to traditional valuation methods, enabling a more systematic and empirical evaluation of enterprise value. It underscores our firm’s commitment to innovation in valuation and model development, helping our clients make more informed financial decisions in an increasingly complex business environment.
The Limitations of Traditional Approaches
Multiples-based valuation has long been the go-to method, but it comes with several drawbacks:
- Subjectivity: Analysts select comparable companies and adjust for perceived differences, which can lead to inconsistent results.
- Data sparsity: For private companies or niche sectors, it is difficult to find truly comparable peers.
- Limited scope: Multiples are usually based on one or two metrics like earnings before interest, taxes, depreciation, and amortization (EBITDA) or revenue, missing the full picture.
- No allowance for firm-specific risk: Factors like credit rating, ownership structure, or balance sheet strength are often overlooked.
Our Alternative: A Quantitative Valuation Engine
Our regression-based approach leverages historical data from U.S. public companies, mapping enterprise value against a comprehensive set of explanatory variables. These include financial performance indicators, capital structure metrics, market-based risk factors, and firm-level attributes such as industry, rating, and ownership type.
This model identifies patterns and relationships in the data — not unlike how machine learning algorithms operate — and applies them to estimate the value of new companies, even if they are private and lack market data. This way, we move from reliance on generic industry averages to customized, data-informed valuation grounded in real market behavior.
The model covers the data, model development, validation processes, and use cases for this regression-based approach. It also explores its applications across various industries and how it complements traditional valuation methods. Ultimately, our goal is to demonstrate how advanced statistical modeling can enhance valuation accuracy and provide actionable insights for investors, analysts, and corporate finance professionals.
Through this work, we aim to showcase our firm’s expertise in model development and validation, underscoring our ability to offer innovative and customized solutions to complex financial challenges.
Key Inputs That Drive Our Model
Some of the major variables that inform our model include:
- LTM EBITDA: A core profitability metric.
- LTM Revenue: To assess the top-line performance.
- Net Debt and Cash: These influence capital structure and enterprise value.
- Credit Ratings: A proxy for firm-level risk and borrowing capacity.
- Industry Category: Controls for sector-specific valuation norms.
- Ownership Type: Differentiates public vs. private valuation behaviors.
- Size Metrics: Like total assets or revenues, which often impact valuation multiples.
- Market indicators: Indicators like treasury rates and high yield spreads are used.
Of these, the final list of variables and their transformations are arrived at by conducting a multistep process to ensure the inclusion of relevant, non-redundant variables while maintaining model robustness. This process included correlation analysis, variance inflation factor (VIF) analysis, stepwise regression analysis, significance testing, residual analysis, and economic interpretability.
Diagnostic Tests
Linear regression models are built on key assumptions that ensure the accuracy and reliability of estimates. To validate the robustness of the enterprise value estimation model, a series of diagnostic tests were conducted to check for violations of these assumptions. The following key assumptions were tested, the diagnostic tests performed, and any corrective actions taken to address violations were considered:
- Residual Analysis for Linearity: The residual plot for the regression model did not exhibit any discernible pattern, indicating that the linearity assumption of the model is satisfied. Therefore, no significant model adjustments are necessary to account for non-linear relationships.
- Independence of Errors (Autocorrelation Check): Durbin Watson statistics within the acceptable range confirmed that the residuals are independently distributed, satisfying another critical assumption of linear regression. This further validated the robustness of our model.
- Homoscedasticity (Constant Variance of Errors): The model uses log transformations of key variables and heteroscedasticity-consistent (HC) standard errors to ensure valid hypothesis testing and confidence intervals.
- Normality of Errors: Despite heavy tails, there is minimal impact on predictive accuracy. Although slight deviations from normality are present, the model’s validation results demonstrate strong predictive capability. The use of HC standard errors and regular residual analysis ensures that coefficient estimates remain reliable, even with non-normal residuals.
- Multicollinearity: All variables resulted in VIF values below 10, hence ruling out significant relationships between independent variables.
Model Results
An adjusted R-squared metric of 81.81% indicates that the model explains a substantial portion of the variance in enterprise value, after adjusting for the number of predictors.
The following plot demonstrates that the predicted median enterprise values closely track the actual median values across time. This suggests that the model effectively captures the underlying trends and variations in enterprise value over the in-sample period.
All coefficients are statistically significant at conventional significance levels (0.05), with t-values indicating strong relationships between the predictors and EV. No multicollinearity issues were detected, as VIF analysis ensured that highly correlated variables were excluded. Overall, the model provides robust insights into the key drivers of enterprise value across sectors and economic conditions.
Model Validation
The model was validated using two robust out-of-sample testing approaches to evaluate its predictive performance and generalizability:
Holdout Test Set Validation:
The original dataset was split into an 80% training set and a 20% holdout test set. The model’s predictions in the test set were compared against actual enterprise values to assess performance. To assess the model’s out-sample performance, a time series plot was generated using the median predicted and actual enterprise values across all companies in the test dataset. The plot indicates strong predictive accuracy, with performance consistent with the in-sample fit.
External Dataset Validation:
The model was further tested on a separate set of 16 public companies not included in the original dataset. This external validation demonstrated the model’s ability to generalize unseen data. Results were similarly strong, indicating that the model effectively captures the underlying relationships in enterprise value drivers across different companies and sectors.
The predicted enterprise values closely aligned with actual values for 14 of the 16 companies, indicating that the model effectively captures the key drivers of enterprise value for most companies.
However, two companies exhibited significant deviations between predicted and actual values. Such deviations can occur for several reasons, like idiosyncratic factors, sector-specific dynamics, extreme financial metrics, extrapolation effects, and so on.
While these two deviations highlight the inherent challenges of financial modeling for highly variable entities, the overall model performance remains strong. Removing these two companies from the analysis resulted in nearly perfect alignment between predicted and actual enterprise values, underscoring the model’s robustness for the majority of companies.
Such deviations are expected in financial modeling, and they highlight potential areas for further investigation or the inclusion of additional variables to capture unique business conditions. Overall, the model’s generalizability and ability to predict across a wide spectrum of companies remain evident from the out-of-sample results.
How We Handle Private Companies
Private firms typically lack price transparency, leading to valuation challenges. Our model adjusts for this by applying a control premium or a discount for lack of marketability (DLOM) — a concept well-known in M&A — to bridge the gap between public and private market expectations.
The base model is trained in public firms to ensure reliability and data integrity. Then, adjustments are layered in to reflect real-world transaction behavior observed in private markets. This ensures the resulting estimates are realistic and defensible.
Using the Black-Scholes-Merton model, we treated DLOMs like the cost of a protective put option — reflecting the price an investor would pay for the right to sell a non-marketable asset. Inputs like holding period, risk-free rate, and volatility were estimated using market data or comparable companies. This helped ensure our valuation reflects real-world investor expectations for less liquid assets.
The figure below shows the out-sample test results from the analysis. Overall, the variance has been reduced, but some outliers persist, possibly due to idiosyncratic factors associated with those specific investments.
Model Outputs: More Than Just a Number
Our approach does not just give you an EV estimate — it provides a detailed decomposition of value drivers. This enables clients to understand:
- What is contributing most to enterprise value.
- How sensitive the valuation is to different metrics (e.g., debt, earnings).
- Where a firm sits relative to market trends in its sector.
These insights can feed into deal pricing, negotiations, fairness opinions, and internal performance bench-marking.
Applications Across the Board
This framework is suitable for application in various scenarios, including:
- M&A advisory: Providing fast and objective valuation support.
- Private equity: Screening targets or estimating net asset value (NAV) in illiquid portfolios.
- Fair value estimation: For financial reporting or strategic planning.
- Investor presentations: Backing growth stories with credible valuations.
Why It Matters Now
As private markets grow and scrutiny over valuation increases, especially post-pandemic, having a repeatable and data-backed valuation method is no longer a luxury — it is a necessity. With regulators, auditors, and investors demanding greater rigor, our model offers a credible and scalable solution that balances economic logic with statistical robustness.
Let Us Discuss How This Can Work for You
Interested in seeing how this quantitative valuation model can enhance your decision-making? Whether you are evaluating an investment, conducting a fairness opinion, or seeking a more transparent method for pricing private firms, our team is here to help.
Sign up to receive all the latest insights from Ankura. Subscribe now
© Copyright 2025. The views expressed herein are those of the author(s) and not necessarily the views of Ankura Consulting Group, LLC., its management, its subsidiaries, its affiliates, or its other professionals. Ankura is not a law firm and cannot provide legal advice.