Discussion of John Y. Campbell, Stefano Giglio, and Christopher Polk

Discussion of
“Hard Times”
John Y. Campbell, Stefano
Giglio, and Christopher Polk
ASSA Meetings, Chicago IL
January 2012
Jonathan A. Parker
Kellogg School of Management, Northwestern University
The idea: use historical relation between assets returns
and subsequent cash flow and discount rate changes to
infer whether the market expects a given market
downturn to be followed by higher future discount
rates or lower future cash flows
The method
The data
The findings
Thoughts on the big question: interpretation
1. The method
Or what the Dickens is going on?
Step 1: Campbell Shiller (1988): assume the
Dividend-Price ratio is stationary
Any deviations in ratio must lead to future
changes in dividends or prices (returns)
Implies that an unexpected return leads to a
change in expected future dividends or returns
Shocks to stream of cash flows (CF) and discount
rates (DR)
• Can calculate NCF and NDR from any
forecasting model, use a VAR
• Applies to any asset with stationary dividendprice ratio (is this true of the market?)
Optional: impose some restrictions on returns implied by
optimization of
• a representative agent
• with KPEZW utility
• who faces only risks spanned by the two shocks (e.g. no
uninsurable labor income, no private equity, etc.)
• and all returns are all jointly lognormal
• and homoskedastic
While all are violations of reality, they are common modeling
assumptions and some models with these features can fit may
asset pricing facts with heteroskedasticity
– Time-variation in risk is large for the focal episodes, 2008 in
– We are referred to Campbell, Giglio, Polk, and Turley (2011) for
time-varying volatilities
Great Expectations of returns:
Expected returns determines by exposure to cash flow
shocks and discount rate shocks
• Decomposition of expected return as due to two
different exposures
– The more exposed to good cash flow news, the higher
expected return and the lower price
– The more exposed to good discount rate news (low rates in
future) the higher expected return and the lower price
• Beta’s can be calculated, used in regression with
average returns (actually use as moment restrictions)
• Restriction: Cross-section has expected returns  times
more sensitive to CF than DR
2. Data
1. Excess log return on the CRSP value-weighted index
2. Log ratio of the S&P 500's price to the S&P 500's ten-year
moving average of earnings
3. Yield difference between the log yield on the ten-year US
constant-maturity bond and the log yield on the threemonth US treasury
4. Difference in the log book-to-market ratios of small value
and small growth stocks
5. Yield spread in percentage points between the log yield
on Moody's BAA and AAA bonds
For expected returns/cross-section: 6 portfolios measuring
size (ME) and value (BE/ME) premia
3. Findings
Time-Smoothed shocks to CF (left) and DR (right) unrestricted
model (top) and restricted model (bottom)
Note: shocks negatively correlated (-0.070 unrest. -0.577 rest.)
Cash-Flow shocks
- Positive in recovery from G.D.
- Negative in 1980’s and 1990’s
- Positive in 2000’s
Suggest one-sided smoother
Discount rate shocks
- Negative on market in G.D.
- Positive in end of G.D. and WWII
- Negative in early 1950’s & 1970’s
- Positive in 1990’s internet boom
It was the best of times, it was the
worst of times . . .
Shaded areas: NBER recessions
Dark lines: market peaks with two year-windows
Hard times: cash flow news followed by disc rate (G.D. and 1937)
Pure sentiment: discount rate news (end of WW II and 1961)
Other: Sentiment followed by cash flow/ NBER recession
4. Interpretation/thoughts
Not primitive shocks – and not claimed to be
- Why are times with high future discount rates not hard times?
- They are times when output today has become scarce relative to the
- If because of increased future cash flow, these are good times, if
because of increased risk or anxiety, these are bad times
- What about credit in the “credit episode”?
- Institutional view: lowered constraints, low market risk aversion
- Are there enough similar episodes to identify market expectations?
- Is it reasonable to use the restricted model?
- Is the paper proscriptive for a long-horizon, attentive, CRRA
- The decomposition is useful for refining models
- My interest: identify structural shocks that map into each type of
reduced form shock and generate facts for models
- E.g. does a monetary policy shock hit mostly cash flow or discount
5. Conclusion
Old literature: are all business cycles alike?
Answer: lots of similar comovement, but some
This paper starts to build similar facts for stock
market cycles, crashes in particular
Main finding: different market declines in US
history have had quite different
implications for future cash flows that are
visible in contemporaneous asset prices

similar documents