RainbowStats • Markets, Accounting and Public Policy

Ibbotson Was Right—but Retained Earnings Explain Why

Long-run equity returns are real. Yet part of their extraordinary compounding reflects a legal and political choice: corporations may keep shareholder earnings and reinvest them as internal capital.

Replication: Run the complete analysis in RainbowStats.

In July, I asked what happens when we look beneath the familiar story of stocks, money and compounding. That earlier RainbowStats essay, “S&P 500, M2, and Retained Corporate Capital,” focused attention on a source of equity wealth that is easy to overlook: corporations routinely retain part of the shareholders’ earnings and place those funds back into the business.

The familiar Ibbotson Stocks, Bonds, Bills and Inflation presentation is right to count the resulting appreciation as part of shareholder return. A retained dollar does not cease to belong economically to the shareholder merely because it never passes through the shareholder’s bank account. It is capital reinvested inside the corporation.

If the dollar is distributedThe shareholder receives a dividend and may reinvest it by purchasing more shares. A total-return index counts both the distribution and the reinvestment.
If the dollar is retainedThe corporation reinvests the shareholder’s dollar internally. If that investment adds value, the result appears in the market price.

Both paths can increase shareholder wealth. Excluding the second path from ordinary equity returns would therefore be wrong. But we can still ask a more demanding question: How much value did corporations create after we charge them for the earnings they retained?

Two valid questions, not one corrected chart

The blue line below is an aggregate equity return above a modeled 10-year Treasury return. The orange line makes one additional deduction: undistributed corporate profits, treated as new shareholder capital committed to the corporate sector.

(Change in equity value − net issuance + dividends − retained earnings) ──────────────────────────────────────────────────────────────────── beginning equity market value

The first calculation measures what equity holders collectively received through dividends and market appreciation after adjusting for share issuance and repurchases. The second resembles an economic-profit calculation: it asks what remained after charging corporations for internally supplied equity capital.

Equity performance above Treasury bondsRolling 25-quarter annualized conventional aggregate equity return and return after charging retained earnings, both above an estimated ten-year Treasury return. -15-10-505101520251960197019801990200020102020 Conventional equity return above TreasuryNet of retained earnings Quarter Annualized percentage points
Figure 1. Rolling 25-quarter annualized aggregate equity performance above an approximate 10-year Treasury total return. The adjusted line subtracts retained earnings as internally contributed shareholder capital. Source: RainbowStats calculations using Federal Reserve Z.1, BEA and FRED data.

The distance between the lines is not an “Ibbotson error.” It is the mechanical effect of charging retained corporate capital against the conventional return under our accounting counterfactual. The size of that accounting charge has varied considerably over time.

The retained-earnings accounting chargeThe annualized accounting difference between the conventional and retained-earnings-adjusted equity calculations over rolling 25-quarter windows. 02468101960197019801990200020102020 Annualized retained-earnings contribution Quarter Annualized percentage points
Figure 2. Annualized accounting difference between the conventional and retained-earnings-adjusted calculations. Because the underlying returns are compounded, this is not a simple contemporaneous retained-earnings yield and should not be interpreted as a causal estimate.
Retained earnings do not invalidate long-run stock returns. They reveal one of the institutional mechanisms that made those returns possible.

How the FRED reconstruction works

The calculation uses the Federal Reserve’s estimate of aggregate corporate-equity liabilities as the market-value measure. Transactions in corporate-equity liabilities represent net issuance; subtracting them prevents newly issued shares from masquerading as investment performance and treats net repurchases as a distribution to shareholders.

BEA after-tax corporate profits are divided between distributed and undistributed profits. We estimate dividends as after-tax profits minus undistributed profits. Because the flow series are reported at seasonally adjusted annual rates, the script divides them by four before comparing them with the quarterly market-value level. RainbowStats normalizes the published millions-versus-billions scales during arithmetic.

The Treasury comparison is necessarily approximate. It uses the 10-year yield with a duration-and-convexity estimate rather than the return on a continuously maintained investable bond portfolio. The purpose is comparison, not replication of the proprietary Ibbotson series.

The political rules have changed before

Corporate retention is not merely an accounting custom. Its attractiveness depends on law. The United States demonstrated this dramatically with the Undistributed Profits Tax of 1936. The new surtax applied graduated rates reaching 27 percent to earnings corporations did not distribute. It was intended in part to push profits out as taxable dividends and sharply increased the price of internal corporate finance.

The experiment was short-lived. The Revenue Act of 1938 essentially removed the broad undistributed-profits surtax, although the accumulated-earnings tax continued to address profits retained for the purpose of avoiding shareholder-level tax. Research on the 1936–1938 episode finds that corporate payout behavior responded to the tax and that managers also searched for other ways to preserve internal funds.

A restrained warning. The present treatment of retained earnings is not a permanent law of nature. Political priorities change, especially when governments face fiscal pressure or perceive that the corporate form is being used to avoid shareholder-level taxation. We should not predict a replay of 1936–1938. We should recognize that Congress has used this instrument before—and that a future political environment could again alter the relative price of retaining, distributing or repurchasing corporate equity.

Such a change would not retroactively make the historical Ibbotson result wrong. It could, however, change the mechanism that generates future equity returns. A tax that discourages retention might increase dividends while reducing internally financed investment. A policy that favors retention might do the reverse. Either way, payout policy, capital formation and market valuation would adjust together.

The conclusion

The great long-run advantage of equities over bonds is not a statistical illusion. Shareholders supplied capital not only when they purchased newly issued shares, but also whenever corporations retained earnings that could otherwise have been distributed. When those retained funds produced future profits, the gains properly appeared in shareholder returns.

Our adjusted series asks corporations to clear a higher bar. It separates the return received by shareholders from the value created after charging management for the shareholders’ money left inside the firm. That is not a replacement for Ibbotson. It is a second lens—and one that makes the institutional foundations of long-run compounding easier to see.

The central distinction: Ibbotson measures shareholder wealth. The retained-earnings adjustment measures value creation after internally financed shareholder capital. Both are legitimate, provided we do not confuse the questions.
RainbowStats replication script
vixcls
M=BOGZ1FL883164105Q
E=BOGZ1FA883164105Q
P=W273RC1Q027SBEA
U=W274RC1Q027SBEA
Y=TO_QUARTERLY(GS10)/100
Data=SAME_DATE_RANGE(LIST(M,E,P,U,Y))
M=EX(Data,series:0)
E=EX(Data,series:1)
P=EX(Data,series:2)
U=EX(Data,series:3)
Y=EX(Data,series:4)
M0=LAG(M,1)
Dividends=(P-U)/4
Retained=U/4
Issuance=E/4
StockReturn=(M-M0-Issuance+Dividends)/M0
StockReturnNetRetention=StockReturn-Retained/M0
Yield0=LAG(Y,1)
YieldChange=Y-Yield0
BondReturn=Yield0/4-8*YieldChange+35*YieldChange*YieldChange
StockLog=EX(LOG(1+StockReturn),cumulative)
AdjustedLog=EX(LOG(1+StockReturnNetRetention),cumulative)
BondLog=EX(LOG(1+BondReturn),cumulative)
StockAnnual=100*(EXP((StockLog-LAG(StockLog,25))*4/25)-1)
AdjustedAnnual=100*(EXP((AdjustedLog-LAG(AdjustedLog,25))*4/25)-1)
BondAnnual=100*(EXP((BondLog-LAG(BondLog,25))*4/25)-1)
Reported=SET_NAME(StockAnnual-BondAnnual,"Aggregate Equity Return above 10-Year Treasury")
Adjusted=SET_NAME(AdjustedAnnual-BondAnnual,"Equity Value Creation Net of Retained Earnings")
Chart=LINE_CHART(SAME_DATE_RANGE(LIST(Reported,Adjusted)))
Chart=SET_TITLE_SUBTITLE(Chart,"Equity Performance above Treasury Bonds","Rolling 25 quarters; conventional return and value creation after charging retained earnings")
Chart=PUBLICATION_CHART(Chart)

Sources and further reading

Methodological note: This is an aggregate historical reconstruction, not an investable index and not a causal estimate of the return to retained earnings. Market-value changes can include revaluations and other adjustments. The constant-duration Treasury approximation differs from a realized bond-index return. FRED series are subject to revision.