Hi Again,
OK, it no longer seems to be on Jstor but I got it on "Business Source Complete." Scratch the above about "selling 8-months after a strong signal."
Rather, while the existing literature talks about yield spreads predicting recessions one year later, and while they thus TESTED whether yield spreads predict bear markets 1-8 months later, they ultimately found that a statistically significant tendency for yield spreads to predict bear markets only shows up 1-3 months out from a yield spread reading. So they then plug some values into their probit model to figure out what's a strong enough signal to sell one month later -- specifically, they figure out the yield spread level at which (per the probit model) there's a 30%, 40% and 50% chance of a bear market in the following month . . .
"Table 2 provides the ex ante probability estimates
from our probit model of a bear market one
month later.^ For example. Table 2 ijidicates that a
yield curve spread of 0.55 percent implies a 30 percent
chance of a bear market in the following month.
A yield spread of-0.17 percent implies a 40 percent
chance of a bear market, and a spread of -0.83
percent indicates a 50 percent probability of a bear
market in one month.^"
(Note that yield spreads are here defined as the 10-year+ minus the 3-month and are presented in percentage terms.)
And here's what they conclude based on those three analyses:
"Table 3 presents the results of the simulations
for each probability level and the results of a simple
buy-and-hold strategy of continual investment in
the S&P 500 mutual fund. Panel A shows the terminal
values of a $1 investment over the 29-year period
and the monthly and annual compound returns.
Investing $1 in a stock-only buy-and-hold strategy
would have produced $46.71 at year-end 1999 and
an armual compound retum of 14.17 percent. Every
probit market-timing strategy shown in Panel A
outperformed the buy-and-hold shategy. As the
probability screen increased from 30 to 40 to 50
percent, the profitability of the market-timmg strategy
increased. When we used a bear market probability
screen of 50 percent, the probit market-timing
strategy yielded an extra $36.31 in terminal value to
the S&P 500 mutual fimd retum and an additional
2.29 pps in annual retum. Panels B, C, and D show
that we found analogous results for the subperiods.
Considering that a tax-deferred investor (e.g.,
an individual or a pension fund manager) could
realistically replicate this simulation by costlessly
switching between a no-load S&P 500 mutual fund
and a money market fund belonging to the same
fund family (e.g., the Vanguard or Fidelity mutual
funds), the probit market-timing strategy produced
economically significant results."
So is it actionable???? I'd still say "meh." If you can figure exactly how they're measuring yield spreads in percentages, you can do exactly what they found was optimal and sell one month after you get a reading of a 50% chance of a bear market. A more practical way to think about it is that, all else equal, it appears to be optimal to sell stocks about one month after someone posts on here that, "wow, the yield curve's is REALLY inverted."
All best,
Pete