Você está na página 1de 5

Evan Kiely FIN 380 January 2012 After gathering all of the data from Yahoo Finance for

our five stocks (Texas Instruments, Clorox, Kelloggs, Proctor & Gamble, and Tiffany & Co.) we narrowed it down to the adjusted closing price for all five stocks. Calculating the monthly holding period returns for the last five years put us in a position to be able to calculate both the annualized return and standard deviation for each company. A necessary piece of information needed to complete efficient portfolio construction is a covariance matrix for all stocks used in the construction of the portfolio. After using Excel to calculate the covariance matrix we then had to go in and adjust this matrix because Excel uses N in its calculations but since we are using a sample of data and not the entire population we must then adjust for this error by multiplying the existing matrix by 60/59, or 1.01695. Once we had these calculations we then plotted each companys risk (X axis) and return (Y axis) to see, graphically, how they compared to one another. The first step in constructing our efficient frontier was to formulate a range of returns in which to work with and then narrow that range into an even finer micro-range to pinpoint a more precise optimal portfolio return. To determine our range of returns we simply took the highest return of individual stocks and the lowest return because these two values would be the maximum and minimum possible returns for our portfolio assuming that we constructed the portfolio with 100% in either stock. Once we had our entire range figured out, we set up a table that, with the help of Excel Solver, would give us the necessary weights of each individual stock in order to produce our desired returns while remaining within the constraint of providing combinations of stocks that had minimum variances. To figure out which combination in this range was the most efficient we set up a column to return the Sharpe Ratio for each return-standard deviation set. The Sharpe Ratio is a unit-less measure of reward to variability. It is the measure of excess return of the asset divided by its standard deviation. This value tells us what level of efficiency we are getting out of a given combination of stocks. For the optimal portfolio, we need the most efficient combination of stocks so that we can treat the portfolio as one equity and use it as our only stock in combining it with the risk free rate. Graphic A shows the most efficient combination of stocks and a few surrounding values that highlight lower Sharpe values. The optimal portfolio resulting from this process has a return of 14.05713%, a standard deviation of 26.4393%, and a Sharpe measure of 0.493854. Our next task was to combine this optimal portfolio with the riskless asset, in this case that is our T-bill, which has a guaranteed return of 1%. The result of combining the optimal portfolio, which we now treat as a single equity, and the risk free T-bill is the formation of the Capital Market Line. The CML is simply a line that shows the change in given levels of return for taking on higher levels of risk. However, since this line represents the combination of the optimal portfolio and the riskless asset (which adds no risk, hence the term riskless) from 100% riskless asset to 200% optimal portfolio and -100% riskless asset, it is essentially just showing the level of return received for holding a certain percentage of the optimal portfolio. We plotted this line on our existing graph and noticed that our result is exactly what we were looking for. The CML is both a straight line and tangent to our efficient frontier at only one point, where the optimal portfolio makes up 100% of the combined portfolio. The slope of this line is the Sharpe ratio we calculated for the optimal portfolio which proves that the optimal portfolio is in fact the most efficient combination of stocks because the Sharpe Ratio is constant at its maximum no matter what weights you split the T-bill and the optimal portfolio up in. All this can be seen in Graphic B. In the next section of the project we added a few benchmark data sets, the S&P500 Index (^GSPC), two ETFs: SLYG and SLYV, and the 13 week T-bill which served as our approximated risk free rate. We calculated monthly returns for all three index data sets and then took the risk free rate and divided by 1200 to adjust for it already being quoted in whole number yearly percent. The point of adding these data was for comparison purposes. The S&P500 Index gives us a measurement of how well

the market performed so when we look at an individual stocks performance we will get a sense of how much of that stocks performance can be attributed to overall returns of the market. To quantify the amount of an individual stocks return that are explained by the market we ran 5 regression simulations with the help of Excels regression analysis tool. We regressed each firms excess monthly returns (Stock1-Rf) against the market excess return (S&P500-Rf) for 60 holding periods. The equation for the line that Excel came up with is subsequently our equation for that stocks CAPM: E[Ri]=Rf+Bi(Rm-Rf), with the X Variable Coefficient representing that stocks beta compared to the S&P500. Before accepting the slope and intercept as true, we first must evaluate the p-value of each to determine its statistical significance. Graphic C shows the main components from the regressions of the five stocks in our project. We compared the return given by the CAPM equation to that of the market consensus return. We were able to calculate the consensus return by dividing the difference of the year end price target and the most recent price with that of the most recent price. This comparison of returns led us to conclude whether each stock was under or overvalued and by how much. The last task here was to do this same regression analysis for both the ETFs with a slight adjustment. The adjustment for the two ETFs being that instead of regressing the excess return of the S&P500 on the excess return of each ETF, we used Fama-Frenchs data which has separate data sets for Mkt-Rf and ETF-Rf. Using Fama-Frenchs data allows us to compare apples to apples when we complete the more in depth 3 factor regression. Next, we used a 3 factor model and the data provided by Fama-French to run similar regressions to the CAPM except this time we used more than just excess market returns to explain the variation of the two ETFs. Our resulting equation had 3 explanatory terms, the market excess return (Mkt-Rf), small minus big capitalization (SMB), and high minus low book to market ratios (HML). SMB is used to explain the amount of excess return that can be attributed to higher historic returns for smaller firms when compared to those of larger firms. HML looks to explain the amount of historic excess returns for value versus growth firms. Thats not the only difference though. When looking at the amount of explanation provided by the model we need to now focus on the adjusted R^2 because the original R^2 term is now biased because whenever you add more explanatory variables to a model the R^2 increases regardless of relevance. Graphic D shows the regression outputs for CAPM and the 3FM for both of the ETFs. As expected the two equations give different expected returns because of the different explanatory variables. Another interesting point is that our more complex 3 factor regression was able to provide more explanation (about 6% more) when compared to the CAPM regression. The next aspect of the analysis required us to analyze bond data. We first regressed yield to maturity on the three sets of year to maturity variables and which revealed a few interesting points. The first point is that introducing both time^2 and time^3 into the regression allowed the model to explain a higher percentage of the data. Our R^2 value increased from 0.82 in the single variable regression to 0.95 in the three variable regression. Our conclusion is that introducing the time^2 and time^3 variables allowed our model to go beyond a simple linear function into a function with a non-linear relationship. From our third regression we were able to calculate predicted values for each bonds yield to maturity by using the regression equation from the output. These values were then plotted on our existing graph of individual actual YTMs. The resulting line is the estimated yield curve for our given set of bonds. Because some of the bonds actual YTMs lie above or below the yield curve this leads to mispricing opportunities. Graphic E shows the resulting betas and p-values for this three variable regression. The two most extreme mispriced bonds are Cornings 21935NAV7, which is severely undervalued, and Monsantos 61166WAE1, which is severely over-valued. These two mispriced bonds give substantial opportunities for high return because you can short the over-valued bond and use the proceeds to buy up all of the under-valued bonds. These two bonds may be mispriced due to various reasons, liquidity may play into this fact a little bit since there is a low quantity of each bond available and thus finding buyers and sellers for each may prove difficult. Also, the fact that Monsantos bond is a relatively new issue may play into its over-valuation. The expectation of the near futures short term interest rate may also be affecting the pricing on both of these bonds. Since our yield curve is upward sloping we are expecting future rates to increase but these are just expectations and therefore are not observed.

We then constructed two bond portfolios, one using a ladder technique and the other using a barbell technique. This just means that we took bonds with maturation years of 1-5 and split them up into proportional weights of our initial $250,000. For the ladder we used $50,000, or as close to $50,000 as we could get because the bonds are not being sold at par because of the differing yields, and allocated that amount for each bond years 1-5. The barbell placed more weight on years 1 and 5, allocating $87,500 for each and giving the remaining years (2, 3, and 4) each $25,000. With their respective weights allocated to each bond we were then able to calculate the portfolio duration for each strategy. This process was done by taking each individual bonds duration and multiplying that by their respective weights. Graphic F shows the overall distribution of bonds for each strategy and their resulting modified durations. In the case that interest rates drop 1% we would prefer to hold the ladder portfolio because it has less weight in year 5 bonds. Having more bonds in year 5 leaves you more vulnerable to interest rate risk because these prices will drop the most (when interest rates rise by 1%) when compared to the year 1 bonds. Graphic A
Rp 0.15105838 0.14231914 0.1405713 0.13882345 0.1370756 Ann p 0.28596265 0.26794326 0.26439269 0.26086098 0.2573489 TXN 0 0 0 0 0 Weights CLX 0.42332654 0.44378278 0.44518925 0.44659573 0.4480022 K 0 0 0 0 0 PG 0 0.03067096 0.03934198 0.04801301 0.05668404 TIF 0.57667346 0.52554626 0.51546876 0.50539126 0.49531376 Sharpe 0.493275518 0.493832701 0.49385365 0.493839473 0.493787237

Graphic B
Weights W Optimal portfolio 2 1 0

W Tbill -1 0 1

R 0.27114259 0.1405713 0.01

Sig 0.528785381 0.26439269 0

Sharpe 0.49385365 0.49385365 -

30.000% 25.000% 20.000% 15.000% 10.000% 5.000% 0.000% 0.000% 10.000% 20.000% 30.000% 40.000% 50.000% 60.000% Sigma

Return

Indiv. Stocks Efficient Frontier CML

Graphic C
Stock
TXN CLX K PG TIF

Intercept
0.006376 0.004004174 0.003812623 0.003280442 0.017504342

p-value
0.404313 0.46390315 0.5149379 0.537348333 0.109336898

Beta
1.127432 0.404517494 0.444791594 0.451659992 1.783323272

p-value
4.33E-11 0.000150508 0.00010547 1.9429E-05 1.2149E-12

R^2
0.530161 0.221032668 0.230057357 0.27182235 0.583977103

CAPM
0.132743 6.045% 6.448% 6.517% 19.833%

Consensus
0.861803 -0.822% 1.486% 14.539% 12.735%

Under/Over
Under: 72.906% Over by: 6.867% Over: 4.96% Under: 8.022% Over: 7.098%

Graphic D
ETF FF3FM SLYG Adj.R^2=0.976 SLYV Adj.R^2=0.977 Intercept 0.002106 p-value 0.16219 Mkt-Rf 1.106 p-value 8.36E-41 SMB 0.849 p-value 2.194E-17 HML -0.2065 p-value 0.000754 3FM Return 16.203% 13.426% 0.00135 0.351782 0.965501 1.99E-38 0.700856 1.06E-14 0.397904 2.32E-09

ETF CAPM SLYG SLYV

Intercept 0.004783238 0.001757478

p-value 0.087544875 0.554733906

Beta 1.216912146 1.183615453

p-value 4.44503E-33 8.66081E-31

R^2 0.917275297 0.900811999

CAPM 14.169% 13.836%

Graphic E
Intercept -0.00898119 p-value 1.14E-53 Beta 1 0.004797168 p-value 1.6E-146 Beta 2 -0.00013215 p-value 1.36E-28 Beta 3 8.10914E-07 p-value 0.001423

Graphic F
5 Year Ladder PARKER HANNIFIN CORP CONOCOPHILLIPS BOEING CO OCCIDENTAL PETE CORP DEL OCCIDENTAL PETE CORP DEL Total Ladder Investment Ladder Portfolio Duration 5 Year Barbell PARKER HANNIFIN CORP CONOCOPHILLIPS BOEING CO OCCIDENTAL PETE CORP DEL OCCIDENTAL PETE CORP DEL Total Barbell Investment Barbell Duration # of Bonds 47.80343229 45.85809671 45.43554514 46.49086919 48.20113369 Whole # Bonds 47 45 45 46 48 $ $ $ $ $ $ 2.839103323 # of Bonds 83.6560065 22.92904835 22.71777257 23.2454346 84.35198396 Whole # Bonds 83 22 22 23 84 Bonds*Price 86,813.85 23,987.04 24,210.12 24,736.04 87,134.88 246,881.93 Weights 0.351641167 0.097159966 0.098063556 0.100193805 0.352941505 Bonds*Price 49,159.65 49,064.40 49,520.70 49,472.08 49,791.36 247,008.19 Weights 0.199020324 0.198634709 0.200482016 0.200285181 0.20157777

$ $ $ $ $ $

2.84838679

Você também pode gostar