Dividend yield, does it matter by ramhood15

VIEWS: 36 PAGES: 21

									Dividend yield, does it matter?

     Based on Fast Forward case.
          What is risk premia?
•   P(t)=Div(t+1)/(r-g)=Div(t) (1+g)/(r-g)
•   P(t+1)=Div(t+2)/(r-g)=P(t)*(1+g)
•   Thus, Exp. Returns = DivY *(1+g) +g
•   Risk premia = DivY *(1+g) +g –Rf
•   But g Rf => Risk premia  DivY *(1+ Rf)
Another argument: valuation ratios
  shold be within some bound
• Conventional efficient market theory: prices
  are random walk.
  – Thus, neither D/P nor P/E ratios should not
    have any forecasting power.
• But if we will accept that for whatever
  reason DivY should be within some bound,
  then either numerator or denominator
  should move in a way that makes market
  variables forecastable. Then what moves?
                        Dividend Yield

0.12




 0.1

                                                Long-term
0.08                                            Mean DivY=4.65%

0.06




0.04




0.02




  0
  1870   1890   1910   1930              1950   1970   1990       2010
 Till next mean
   crossing...
  (a bit of cheating...)

• Poor job for Div growth
  forecasting, R2=0.25%
• Good job for price growth
  forecasting R2>30%
• Thus, it is
  DENOMINATOR that
  brings back DivY within
  ”decent” range
One year horizon
• Div growth is fairly
  predictable, R2=13%
• Price changes are
  almost not predictable,
  R2 is about 1%
Ten years Horizon

• R2=1% for DivG and
  16% for PriceG.
• Note that DivG results
  are not really
  explainable within eff.
  Mkt theory at all.
• Based on that, within
  next 10 yrs we should
  expect 55% drop in S&P
50
                P/MA10(E)
45




40



35




30




25




20




15



10




 5




 0
 1880   1900   1920   1940   1960   1980   2000
Forecasting from
 P/smoothed E
      ratio
• R2 for price growth
  regression is high
  (40%)
• Superior to DivY
• Forecast is really
  bad..
• We cannot forecast productivity growth....
Anything new happened within the
  last 30 years? Share buybacks
• Share repurchases have tax advantages
  w.r.t. paying simple dividends.
• Part of earnings that can be used for
  dividend payout is now smaller and DivY is
  underestimated w.r.t. similar number 50
  years ago
• What difference does it make?
  Cole,
Helwege
& Laster,
 FAJ 96
Assuming both
  buybacks and
  new issues are
  done at market
  prices,
  significant
  adjustment for
  DivY is
  necessary
   Cole, Helwege
     & Laster,
    FAJ 96 (2)
• Adjustment of 0.8% in
  96.
• Problem: most of
  options are issued at
  below mkt price. Liang
  and Sharpe: for 144
  S&P500 firms in 97
  adjustment is 1.39%, in
  98 0.75%
Hard assets and financial claims
            diverge
     Intangible investmenst (1)
• ”New economy” involves substantial
  investments in intangibles.
• Accounting procedures do count activity to
  promote web site as expenses but ”...they
  are really investments”. Hall (2000) called it
  e-capital.
• McGrattan &Prescott : understatement in
  earnings are about 26%
        Intangible investmenst
• McGrattan &Prescott :
  – understatement in earnings are about 26%
  – Fits only last couple of years
• Bond & Cummins: if intangibles are
  counted as R&D and marketing, then for
  459 industrial firms 82-98 still there is no
  explanation of overvaluation.
        Demographic changes
• Affluent society is formed
• Baby boomers are educated, have money to
  invest and need to save for retirement
  (uncertainties related to Social Securities,
  etc. )
• Thus, one-time shift in risk premium.
                 Inflation
• Responce to inflation is not always rational.
  Modigliani & Cohn 79: People discount
  dividends not at real, but at nominal rate.
  Thus, when inflation is high, stock market is
  undervalued, and when it is low, stock mkt
  is overvalued.
• Now CPI is low...
International Evidence: Mixed
    What else Dividends can tell us?
•   Let us consider 2 variables: Prices and dividends.
•   Gordon/Shapiro says, that
•   P(t)=E(k D(t+k) *(1+r)-k)P*(t)
•   P*(t)=P(t)+e(t), assume E(e(t)| P(t))=0
•   =>cov(P,P*)=cov(P,P)+cov(P, e)=var(P)
•   -1<Cov(P,P*)/(StdPStdP*)<1
•   => Std(P)/Std(P*)<1=> std(P)<Std(P*)
•   Volatility of forecast (prices) should be smaller
    than volatility of payouts (dividends). But the
    relationship is exactly opposite!!!!

								
To top