ECN 141 Economic and Financial Forecasting

University of California, Davis Prof. Shu Shen

FN Sample II

Short Answer Questions on Portfolio Choice:

1. True or False. An AR(1) model discounts the effect of last period’s shock on this period’s
outcome. The discounting factor is the model’s slope coefficient, which is smaller than one in
absolute value with a stationary and weakly dependent time series process.

2. Which dots below represent: 1) the minimum variance portfolio; 2) a portfolio that mixes 80%
of risk-free asset and 20% of risky asset 2; 3) a portfolio that mixes 80% of risky asset 1 and
20% of risky asset 2. [Answer A, B, C, D, E, or F for each part of the question.]

3. On the above figure, roughly draw the line representing optimal portfolios for mixing risky
asset 1, risky asset 2, and the risk-free asset. Suppose your maximum tolerance of risk is 0.10.
On the above figure, roughly mark your optimal portfolio.
R Output Questions

Time Series: Series medlistprice stored in R data frame HomeSale is the monthly median listing
price of zip code 94598 in Walnut Creek, CA. Series t counts the number of time period. For the
first period Jan. 2010, t=1. For the last period Jan. 2019, t=109. So the total number of time
period T is 109. Series month =1, 2, …, 12 states the month of the data.

4. Read the following R codes and output. Explain what the codes are doing and what you learnt
from the R output.
results<-lm(medlistprice~t+factor(month),data=HomeSale)
HomeSale$medlistpriceadj<-residuals(results)
auto.arima(HomeSale$medlistpriceadj)

Series: HomeSale$medlistpriceadj
ARIMA(1,0,0) with zero mean

Coefficients:
ar1
0.8157
s.e. 0.0558

sigma^2 estimated as 671730685: log likelihood=-1262.44

5. Explain why the 95% forecast interval for the real value of medlistpriceadj in period T+1
would always be wider than the 95% confidence interval for conditional mean of medlistpriceadj
in period T+1.

1
Yield curve: The following R output table and figure give out the estimation result of the Nelson
Siegel yield curve for U.S. Treasury bonds in Nov. 2006. (Unit is % on the y-axis of the graph.)

6. What does the shape of the yield curve tell us about public’s expectation of future interest
rates in Nov. 2006?

7. Find (roughly) the yield spread between 1-year and a 2-year T-bills from the above graph.

8. Suppose the annual yield of the 1-year T-bill you found in the last question is A%, and the
annual yield of the 2-year T-bill you found in the last question is B%, Suppose the liquidity
premium of a two-year bond over a one-year bond is 0.1%. What is the expected yield of a one-
year bond to be issued a year from Nov. 2006 (timing of the figure).

2
9. The Nelson Siegel model gives the following results for yield curve estimation. Interpret the
beta_0 and beta_1 estimates.
beta_0 beta_1 beta_2 lambda
2006-11-01 4.84046 0.3990924 -1.456601 0.3912666

Oil Price and CAPM: This question looks at daily performance of an investment fund “USO”
traded in Nasdaq that seeks to have its daily return tracking the oil price in U.S., or specifically
“the daily changes in percentage terms of the spot price of light, sweet crude oil delivered to
Cushing, Oklahoma”. We collected daily closing price of the fund (USO), daily closing price of
SP 500 (SP500), daily closing price of the US-EURO exchange rate (USEUR, USERU=0.8
means 1 US dollar is worth 0.8 EURO), and the daily interest rate of the 3-month T-bill
(DGS3MO) from the beginning of 2010 to the end of 2018. Those four series are stored in the R
data frame called data.

First, results from the “auto.arima” command suggests that all four time series are random walks.

10. Consider a general random walk model: 𝑦! = 𝛽" + 𝑦!#$ + 𝑢! , t = 0,1, … , T. Suppose that
𝑦" = 0, show that 𝑦! = 𝑡𝛽" + 𝑢! + 𝑢!#$ + ⋯ + 𝑢$ .

11. Are past shocks discounted in the above model of 𝑦! ? What does 𝛽" in the above model
represent?

3
Next, we define a time series for the daily return of USO. We call the new series r.USO. The
following two graphs give out the autocorrelation function of the daily return series, or r.USO,
and the squared daily return series, or r.USO^2.

12. What does the left graph suggest about the serial correlation of the daily return series of
USO? Is the daily return series weakly dependent? What about the right graph?

The following codes model the r.USO series using an AR(0) model with GARCH(1,1) errors.
library(fGarch)
garchFit(formula = ~ garch(1, 1), data = data$r.USO[-1],trace=F)

Coefficient(s):
mu omega alpha1 beta1
9.7188e-07 2.4438e-06 5.7328e-02 9.3924e-01

Error Analysis:
Estimate Std. Error t value Pr(>|t|)
mu 9.719e-07 3.740e-04 0.003
omega 2.444e-06 1.009e-06 2.422
alpha1 5.733e-02 8.696e-03 6.593
beta1 9.392e-01 9.114e-03 103.050

13. Interpret the shaded estimate for mu. Is it statistically significant?

4
14. Based on the above AR(0)+GARCH(1,1) model, we find that the 1-period ahead forecast for
the standard deviation of r.USO is 0.032. Calculate the 1-step ahead point forecast of r.USO as
well as its 95% forecast interval. [You can just use 2 as the critical value.]

Finally, we use the CAPM model to study the daily return of USO. We create the excess return
of USO and SP500 using the following lines of R codes.
data$r.USO<-(data$USO-Lag(data$USO,1))/Lag(data$USO,1)
data$r.SP500<-(data$SP500-Lag(data$SP500,1))/Lag(data$SP500,1)
data$r.f<-data$DGS3MO*c(NA,diff(as.numeric(time(data$DGS3MO))))/100/365
data$exr.SP500<-data$r.SP500-data$r.f
data$exr.USO<-data$r.USO-data$r.f

15. If someone holds a portfolio that invests 30% in USO and 70% of the 3-month T-bill. What
is the return of his portfolio on Dec. 11, 2018 based on the following numbers? What is the
excess return of his portfolio Dec. 11, 2018?
USO DGS3MO r.USO r.f
2018-12-11 10.95 2.41 0.01860 0.00007

The following R output table reports the CAPM estimation results for USO.
summary(lm(exr.USO~exr.SP500, data=data))

Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) -0.0009134 0.0004208 -2.171 0.0301 *
exr.SP500 0.8799800 0.0413129 21.300 <2e-16 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.01868 on 1972 degrees of freedom

16. Interpret the slope coefficient as well as the R-squared of the above regression model.

5
17. What is the alpha of USO? Is USO over-priced or underpriced? Explain.

18. If you invested in a portfolio that mixed the market portfolio SP500 and the risk-free asset 3-
month T-bill by an 87.998% Vs. 12.002% ratio over the same time period (2010-2018). Would
your portfolio have a higher excess return than USO? Would your portfolio have a lower risk
than USO? Explain.