Introduction
The ROE-P/TBV regression is the workhorse relative valuation tool in FIG equity research and investment banking. The concept is straightforward: if P/TBV is driven by ROE relative to cost of equity, then plotting ROTCE against P/TBV for a peer group of banks should produce a clear positive relationship. The regression line represents the market's "fair" P/TBV for any given level of profitability. Banks plotting above the line may be overvalued relative to their returns (or the market is pricing in future improvement). Banks below the line may be undervalued (or the market sees risks that depress the multiple beyond what current ROE justifies). FIG analysts at KBW, Morgan Stanley, Goldman Sachs, and every major research platform maintain updated scatter plots that anchor their buy/sell recommendations.
How the Regression Works
The regression is mechanically simple. For a peer group (typically 15-30 banks of similar size), plot forward ROTCE on the x-axis and current P/TBV on the y-axis. Run an ordinary least squares (OLS) regression to find the best-fit line. The slope tells you how much additional P/TBV each percentage point of ROTCE is worth. The intercept (often near 0 or slightly negative) represents the P/TBV a bank would receive at zero ROE.
The relationship is grounded in the justified P/BV formula:
This formula shows that P/TBV is a linear function of ROE when cost of equity and growth are held constant. The regression empirically estimates this relationship across a cross-section of banks, with each bank's idiosyncratic factors (management quality, asset quality, growth trajectory, regulatory position) creating dispersion around the line.
R-squared values typically range from 47% to 68% for US bank peer groups, meaning ROTCE explains roughly half to two-thirds of the variation in P/TBV. The remaining variation captures factors the regression does not include: credit quality differences, deposit franchise value, geographic concentration, revenue mix, and market sentiment. Regional bank peer groups in the US Northeast have shown R-squared values as high as 68%, while broader samples with more diverse business models produce lower R-squared (closer to 47%).
- Regression Residual (Distance from the Line)
The residual is the vertical distance between a bank's actual P/TBV and its regression-implied P/TBV. A positive residual (above the line) means the bank trades at a premium to what its ROTCE would justify. A negative residual (below the line) means the bank trades at a discount. The magnitude of the residual, expressed in standard deviations, indicates how extreme the mis-pricing appears. Banks more than 1.5 standard deviations from the line warrant investigation: either the market knows something the regression does not capture (a pending acquisition, regulatory action, or management change), or the bank is genuinely mispriced. FIG analysts use the residual as a starting point for deeper fundamental analysis, not as a standalone trading signal.
Reading the Scatter Plot
The current US large-cap bank scatter illustrates the regression's explanatory power.
| Bank | ROTCE (Forward) | P/TBV | Position vs. Line |
|---|---|---|---|
| JPMorgan | ~20% | ~2.6x+ | Upper right (premium justified by scale, moat) |
| Morgan Stanley | ~21% | ~2.8x | Upper right (wealth management drives premium) |
| Goldman Sachs | ~15-16% | ~1.6x | Mid-range (markets volatility limits premium) |
| Wells Fargo | ~17-20% | ~2.1x | Near line (asset cap resolution could drive upside) |
| Bank of America | ~15% | ~1.9x | Near line |
| PNC | ~18% exit rate | ~1.6-1.8x | Below line (potentially undervalued for returns) |
| Truist | ~12.7% | ~1.2x | Near or below line (restructuring discount) |
| Citigroup | ~10-11% target | ~1.14x | Lower left (below line, execution skepticism) |
JPMorgan consistently plots at the upper right corner. Its 20% ROTCE combined with a 2.6x+ P/TBV premium appears "on the line" or slightly above, reflecting the market's view that JPMorgan's competitive advantages (dominant deposit franchise, diversified fee revenue, management execution) justify a premium even beyond what the regression predicts. Citigroup at the lower left demonstrates the opposite: even its modest P/TBV of 1.14x may be above the regression-implied value given an ROTCE that only recently crossed above cost of equity.
The regression is a direct empirical application of the excess return model. Banks with ROE above cost of equity generate positive excess returns, which translates to P/TBV above 1.0x. Banks with ROE below cost of equity generate negative excess returns, producing P/TBV below 1.0x. The regression line maps this theoretical relationship onto observable market data, with the slope reflecting the market's aggregate view of the cost of equity and the persistence of excess returns across the peer group.
The ROE-P/TBV regression transforms FIG relative valuation from subjective peer comparison into a systematic, data-driven framework. Together with the justified P/BV ratio (the theoretical foundation), the excess return model (the intrinsic valuation equivalent), and normalized P/E analysis (the earnings-based cross-check), it completes the relative valuation toolkit for any bank equity analysis.
Free Interview PDF Guide