Damodaran’s Implied ERP and U.S. Stock Returns (2008–2026)
Description
This dataset combines Damodaran’s implied U.S. equity risk premium (ERP), subsequent stock returns and financial indicators to support research on equity-premium predictability. It contains 214 monthly forecast observations from September 2008 to June 2026 and 71 quarterly forecast observations from October 2008 to April 2026, with realized-return coverage ending on 30 June 2026. The data include S&P 500 total returns, compounded risk-free returns and excess returns at three-, six- and twelve-month horizons, together with twelve-month Fama–French market excess returns. Financial predictors include VIX, the Treasury term spread, the corporate credit spread, inflation, economic policy uncertainty, Shiller CAPE, earnings yield and dividend yield. Stock and risk-free returns are compounded separately before excess returns are calculated. Forecast origins and observed return-window dates are recorded explicitly, and incomplete outcomes are left blank. The twelve-month estimation samples contain 68 quarterly and 203 monthly observations, with July 2025 as the last forecast origin having a complete twelve-month outcome. The Excel workbook contains four worksheets: Read me, Quarterly data, Monthly data, and Variables. Sources include Damodaran’s data archive, the S&P 500 Total Return Index distributed through Yahoo Finance, the Kenneth French Data Library, FRED and Multpl. Variable definitions, units, source links and timing conventions are documented within the workbook. Controls use specified publication lags applied to historical series; they are not archived real-time vintages.
Files
Steps to reproduce
1. Open ERP_FRL_Mendeley_Dataset.xlsx and consult the Read me and Variables worksheets for definitions, units, sources and timing conventions. The following steps reproduce the analysis from the processed observations supplied in the workbook. 2. Use Quarterly data for the main analysis and Monthly data for monthly robustness checks. Treat forecast_date as the prediction date and retain observations with a complete outcome at the selected horizon. The twelve-month samples contain 68 quarterly and 203 monthly observations. 3. Use spx_excess_3m, spx_excess_6m and spx_excess_12m as the S&P 500 excess-return outcomes. Each equals the corresponding sp500tr_return variable minus rf_return, with stock and risk-free returns compounded separately. Use ff_excess_12m for the alternative market-return specification. Rates and returns are stored as decimals; blank outcomes must remain missing. 4. Estimate OLS regressions with an intercept and erp as the predictor. For the control specification, add vix_rt, term_rt, credit_rt, inflation_rt and epu_rt. For the valuation specification, add cape_rt, earnings_yield_rt and dividend_yield_rt. Compare augmented models with ERP-only regressions estimated on the same complete observations. 5. Calculate Bartlett-weighted Newey–West standard errors without a finite-sample covariance correction, using two-sided normal p-values. Use 0, 1 and 3 HAC lags for quarterly three-, six- and twelve-month regressions, respectively, and 2, 5 and 11 lags for the corresponding monthly regressions. 6. For quarterly twelve-month out-of-sample evaluation, generate expanding-window forecasts at the 36 common origins from October 2016 through July 2025. At each origin, train only on observations whose return_end_12m precedes that origin. Apply the same rule to the historical-mean benchmark. Compare mean squared prediction errors and calculate out-of-sample R² as 1 minus model MSPE divided by benchmark MSPE. For nested-model comparisons, use the Clark–West adjusted loss differential, testing its positive mean with three HAC lags and a one-sided normal p-value. 7. For crisis sensitivity, distinguish exclusions based on forecast origins from exclusions based on twelve-month earned-return windows. Use December 2007–June 2009 for the GFC and February–April 2020 for COVID. Window overlap begins in the forecast month and excludes the preceding base-close date; retaining complete quarterly windows that overlap neither crisis gives 60 observations.
Institutions
- Ca' Foscari University of VeniceVeneto, Venice