How to Estimate the Residual Income Persistence Factor

The residual income persistence factor, typically denoted omega (ω), measures how much of a company’s current abnormal earnings will persist into future periods. Estimating it is critical to the residual income model because it directly affects the terminal value. Analysts use three main approaches: historical regression of abnormal earnings on lagged values, industry or sector benchmarks, and forward-looking adjustments based on competitive position and earnings quality.

What the Persistence Factor Represents

The residual income model values a firm as:

Equity value = Book value + Present value of abnormal earnings

Abnormal earnings (or residual income) = Net income − (Cost of equity × Beginning book value)

When abnormal earnings are high, the question arises: how long will they last? If a company earns 15% returns on equity in year 1 and the cost of equity is 10%, the residual income is 5%. In year 2, will abnormal earnings still be 5%, or will competition erode them?

The persistence factor ω answers this by modeling how much of year-1 abnormal earnings flow into year-2 residual income, how much of year-2 flow into year-3, and so on. It captures the sustainability of competitive advantage.

Method 1: Historical Regression

The most direct approach is to regress a company’s abnormal earnings on lagged values:

RI_t = α + ω × RI_{t-1} + ε_t

Where RI is residual income in period t, ω is the persistence coefficient, and ε is the error term.

Steps:

1. Calculate abnormal earnings (residual income) for the company over 10–15 years of historical data.
2. Run an ordinary least-squares (OLS) regression with RI_t as the dependent variable and RI_{t-1} as the independent variable.
3. The slope coefficient is your estimate of ω.

Example: If you regress a bank’s residual income for 2010–2024 and the slope is 0.62, that means 62% of abnormal earnings in one year persist to the next on average.

Strengths:

• Purely empirical; no judgment calls
• Captures actual competitive dynamics over a full business cycle

Weaknesses:

• Requires 10+ years of clean data; many younger firms don’t have it
• Assumes past persistence equals future persistence (ignores industry shifts, technological disruption, regulatory change)
• Outlier years (crises, one-time charges) distort the regression
• Small sample sizes inflate noise

For cyclical industries, the regression will capture mean reversion during downturns and expansions, making ω appear lower than the “true” structural persistence.

Method 2: Industry and Sector Benchmarks

Academic research and practitioner surveys have established typical persistence factors by industry:

This approach saves time and avoids data limitations. If a company is a utility, starting with ω = 0.70 is reasonable unless you have specific evidence otherwise.

Adjustments: Even within a sector, adjustments may apply:

• Market leader in concentrated market (e.g., a dominant cloud provider): Add 0.05–0.10 to the benchmark (e.g., 0.50 → 0.60).
• Fragmented market with weak competitive advantage: Subtract 0.05–0.10.
• Negative abnormal earnings (company earning less than cost of equity): Consider using a lower ω or even negative persistence, since losses tend to trigger restructuring.

Method 3: Earnings Quality and Forward-Looking Adjustment

A hybrid approach starts with an industry benchmark, then adjusts based on qualitative factors:

Factors that suggest higher persistence (raise ω):

• Switching costs (customers rarely change vendors)
• Network effects (value rises with user base; hard to dislodge)
• Intangible assets (brand, patents, data assets)
• High gross profit margin relative to peers (pricing power)
• Recurring revenue model (subscriptions, long-term contracts)
• High return on equity with stable margins

Factors that suggest lower persistence (lower ω):

• Commoditized products; pricing pressure
• High customer churn or contract concentration
• Regulatory risk (future changes could cap earnings)
• Rapid technological change or disruption risk
• High debt-to-equity ratio (financial risk reduces safety of abnormal earnings)
• Declining market share or erosion of competitive position

Example: A biotech firm with a large pipelines of patent-protected drugs might start with a pharma benchmark of ω = 0.50. But if one drug faces imminent patent expiry and the pipeline is weak, you might lower it to ω = 0.40. Conversely, if the firm has 10 late-stage candidates and first-mover advantage, you might raise it to ω = 0.55.

Practical Estimation Workflow

1. Gather historical residual income for 10 years if available; calculate the regression coefficient.
2. Cross-check against industry benchmarks. Does your regression result align? If it’s far outside the typical range (e.g., ω = 0.85 for a tech company), investigate outliers or consider whether the company is exceptional.
3. Assess competitive position using the qualitative factors above.
4. Settle on a point estimate. Most practitioners use a single ω (e.g., 0.55), though a sensitivity analysis with a range (0.45 to 0.65) is valuable.
5. Test the terminal value. Plug your ω into the residual income model and check whether the implied equity value is reasonable relative to peers and a simple price-to-earnings-ratio benchmark.

Sensitivity of Valuation to ω

The persistence factor has material impact on value. Consider a company with:

• Book value: $10 billion
• Year 1 abnormal earnings: $500 million
• Cost of equity: 9%
• Growth rate: 2%

Small changes in ω swing valuation by billions. This is why sensitivity analysis and triangulation with other methods (peer multiples, DCF) are essential.

Common Pitfalls

• Using ω = 0 (abnormal earnings disappear immediately) is too conservative and rarely empirically supported.
• Using ω = 1 (perpetual abnormal earnings) ignores competition and is appropriate only for companies with durable, legally protected competitive advantages (e.g., a utility with a 40-year concession).
• Ignoring cyclicality. If you estimate ω during a cyclical peak, you’ll overestimate persistence; use multi-cycle data.
• Overfitting the regression. A single company’s 15-year regression can be noisy; compare to peer groups.

See also