Trading Rule and Forecasts of Dividends Presented by Dooruj - - PowerPoint PPT Presentation

Trading Rule and Forecasts of Dividends

Presented by Dooruj Rambaccussing Supervised by Ian Bulkley James Davidson

Introduction

• Objective : ‘Can Econometric Forecasts of

Dividends help in building up a trading rule for enhancing wealth ?’ Innovation of this paper :

• Forecasting of Dividends (Brown et. al 2000)
• Use of Fama and French (2002) Earnings and

Dividend growth rate as the discount rate.

Background Literature

• Trading Rule: (Bulkley and Tonks 1989,

1991), (Bulkley and Taylor 1996)

• Models for forecasting(Timmermann 2008)
• Forecast comparison (Hansen 2005)

Trading Rules

• Moving Average Oscillator
• Trading Break-Out
• Comparison of REPV

with Actual Price

REPV of Equity

• Model 1:
• Model 2
• Innovation: Et [Dt+1 ] is proxied by one step

ahead forecasts

• 4 Empirical Valuations of r were used.

P t 

1 r E tDt1

P t 

1 rg E tDt1

Trading Rule

• Assumption of 2 assets (Equity and

Bonds)

• Compare REPV (P*) with Actual Price (P)
• P*> P (Go long on Equity Index)
• P*< P (Go long on Bonds)

Forecasting Accuracy

• Hansen Superior Predictive Accuracy Test

(Global)

• Recursive Mean Squared Error Plots

(Local)

Background for Forecasting Dt+1

• S&P 500
• Sample: Jan 1871 to Aug 2008
• Initial Estimate January 1871 to Dec 1900
• Use Rolling and Recursive Windows
• No formal misspecification test applied in

sample –Objective is forecasting misspecification

• 40 models were estimated (Window type,

Functional form)

4 Best Models

• Model 1:
• Model 2
• Model 3 (Generic Form)

9 forms of this model is considered based on type of Trend and Seasonality. Best Model is selected using AIC.

lnDt1    i1

p

i lnDt1i  t1

Dt1    1t  1  Dt  t1

F t1   F t (1- Dt

Forecast Models (Continued)

• Model 4

Dt1    1Dt  2P t  3E t  4Rt

f  t1

Results on Forecasting Accuracy

• Hansen’s test
• MAE:

Rolling Exponential Smoothing (Model 3)

• MSE:

Recursive Exponential Smoothing (Model 3)

• Unconditional MSE :
• AR (p) (Model 1)

Recursive Mean Squared Error Plot

0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 01/1904 01/1916 01/1928 01/1940 01/1952 01/1964 01/1976 01/1988 01/2000 Time Plot Model 1 Rec Model 1 Rol Model 2 Rec Model 2 Rol Model 3 Rec Model 3 Rol Model 4 re Model 4 Rol

Benchmarks

Annually compounded Return on S&P 500 is 5.96% 2008. $100 invested in January 1901 is worth $ 51 919 in 2008 Experiment: Invest $100 January 1901 and shift assets as dictated by trading rule

Accumulated Wealth

500000 1e+006 1.5e+006 2e+006 2.5e+006 3e+006 3.5e+006 4e+006 4.5e+006 01/1904 01/1916 01/1928 01/1940 01/1952 01/1964 01/1976 01/1988 01/2000 Time Plot Model 1 Rec Model 1 Rol Model 2 rec Model 2 rol Model 3 Rec Model 3 Rol Model 4 Rec Model 4 Rol

Accumulated Wealth under Buy and Hold

10000 20000 30000 40000 50000 60000 70000 80000 90000 01/1904 01/1916 01/1928 01/1940 01/1952 01/1964 01/1976 01/1988 01/2000 Buy and Hold

Major Findings and Conclusion

• Out of 160 accumulated wealth values, the

trading rule beats the accumulated wealth if invested on the stock market 111 times.

• Fama and French (2002) discount rate model of

Earnings Growth does not beat benchmark.

• Strength of Rule : identifies timing when to shift

Assets from Equity to Bonds and vice-versa. (and not necessarily the number of times going long on equity)

• REPV of Stock is sensitive to the type of

discount rate being used.