Working Paper: NBER ID: w4890
Authors: John R. Graham; Campbell R. Harvey
Abstract: We analyze the advice contained in a sample of 237 investment letters over the 1980-1992 period. Each newsletter recommends a mix of equity and cash. We construct portfolios based on these recommendations and find that only a small number of the newsletters appear to have higher average returns than a buy-and-hold portfolio constructed to have the same variance. Knowledge of the asset allocation weights also implies knowledge of the exact conditional betas. As a result, we present direct tests of market timing ability that bypass beta estimation problems. Assuming that different letters cater to investors with different risk aversions, we are able to imply the newsletters' forecasted market returns. The dispersion of the newsletters' forecasts provides a natural measure of disagreement in the market. We find that the degree of disagreement contains information about both market volatility and trading activity.
Keywords: investment newsletters; market timing; asset allocation; volatility
JEL Codes: G11; G14
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| increases in equity weights (G32) | market rises (G10) |
| forecast dispersion (C46) | future realized volatility (G17) |
| forecast dispersion (C46) | future implied volatility (G17) |
| forecast dispersion (C46) | future trading volume (G13) |
| past performance of newsletters (G14) | future performance (L25) |
| dispersion of newsletter forecasts (G17) | market activity (G10) |