Lecture 07

IIT Kanpur July 201
20 followers
Oct 22, 2023
In this lesson, we'll revisit the problem of portfolio construction, computation of
expected returns, risk, correlation and covariance.
We'll also summarize our learnings of portfolio optimization with two security cases and multi-security
cases.
We'll construct portfolio possibility curve and examine the feasibility region where all
the possible combination of securities in the form of portfolios are available to investors.
We'll examine various properties of this feasible region with and without short sale and minimum
variance portfolio.
We'll try to find the most efficient portfolio on these available set of portfolios.
Subsequently, we'll introduce risk-free lending and borrowing and examine the implication
of the efficient set of portfolios.
Lastly, we'll introduce the concept of market risk and beta.
Portfolio Construction Recap 1 In this video, we'll discuss the expected
return, risk and correlation structure for a multi-security portfolio.
The actual return on a portfolio of assets for any given period is the weighted average
of the actual return on the individual securities for that given period.
The weights are the fraction of the portfolio invested in securities.
Now we'll try and compute the expected returns of the portfolio.
But what is the need to compute the expected returns?
Because it is the expected returns that are matched with the risk profile in various portfolio
applications, not the actual returns.
Now keeping this rule in mind, we know that expected returns on any security for any period
are the simple average, if all the return observations are equally likely, of the returns
sampled from the past periods.
What does it mean and what are its implications?
This simply means the following.
If you have three observations, for example, 10%, 20% and 30%, and also that the returns
on a security are sampled in an unbiased manner over equally spaced periods, like daily, weekly
and so on, then the expected returns for any given period are the simple averages, for
example, 10 plus 20 plus 30% divided by 3, which is equal to 20%.
Mathematically, this can be very easily written as return Ri bar, which is the expected returns
equal to 1 upon t, summation t equal to 1 to t, Ri t.
Now consider the actual return on a portfolio for period t, as shown here, Rp t equal to
summation Xi Ri t over i equal to 1 to n, where i depicts one of the n securities and
Xi is the weight invested in the security i.
Now the expected returns of the portfolio can be easily written as Rp bar equal to expected
Rp t or expectations of Xi Ri t summation over i equal to 1 to n.
This can be also written in the following manner, summation of i equal to 1 to n, expectations
of Xi Ri t or summation of i equal to 1 to n, Xi times expectation of Ri t because Xi
is constant and this can be effectively written as Rp bar equal to summation i equal to 1
to n, Xi times Ri bar as shown here in equation 2.
Even if you are not comfortable with the derivation above, keep the last result in
mind.
In simple English, this result suggests that the expected returns on the portfolio are
simply the weighted average of the expected returns on the individual securities.
Here the weights are the proportion of the portfolio invested in the securities.
Please note that while equation 2 looks similar to equation 1, there is a fundamental difference.
Equation 1 describes the relationship between the actual return on portfolios and individual
securities in that portfolio, while equation 2 describes the expected returns.
Now let us talk about computing the risk or standard deviation for a portfolio of assets.
Now that we have understood the return part, we will move to the risk part.
However, it is not so simple for a multi-security case.
So we will start with the two securities case with returns being R1, R2.
Therefore, sigma p square is equal to expectation of Rp t minus Rp bar raised to the power 2,
which can be further expanded with expectations of X1R1t plus X2R2t minus X1R1 bar plus X2R2
bar raised to the power 2, which is also equal to expectation of X1 times R1t minus R1 bar
plus X2 times R2t minus R2 bar raised to the power 2.
This expression can be further expanded to sigma p square times expectation of X1 square
times R1t minus R1 bar square, then X2 square R2t minus R2 bar square plus 2 times X1X2R1t
minus R1 bar square into R2t minus R2 bar, which can be further simplified like this
X1 square times expectation of R1t minus R1 bar raised to the power 2, X2 square expectation
of R2t minus R2 bar raised to the power 2 plus 2 times X1X2 expectation of R1t minus
R1 bar into R2t minus R2 bar.
This third term here, which is denoted as expectation of R1t minus R1 bar into R2t minus
R2 bar is also called the covariance between securities 1 and 2 and it can be easily depicted
as sigma 1 2, which is the covariance between security 1 and 2 and please note a very fundamental
property that sigma 1 2 equal to sigma 2 1.
Also, we must notice that this expectation of R1t minus R2 R1 bar raised to the power
2 is nothing but sigma 1 square, this is by definition sigma 1 square, similarly expectation
of R2t minus R2 bar raised to the power 2 is nothing but sigma 2 raised to the power
2.
Therefore, our resulting expression can be written here, this is our resulting expression
of risk for the two security portfolio that is sigma p square equal to X1 square sigma
1 square plus X2 square sigma 2 square plus 2 times X1X2 into sigma 1 2.
Now, this expression can be further extended for a three security portfolio and here it
is sigma p square plus X1 square sigma 1 square plus X2 square sigma 2 square plus X3 square
sigma 3 square plus 2 times X1X2 sigma 1 2 plus 2 times X1X3 sigma 1 3 plus 2 times X1X3
sigma 2 3, which is very simply an extension of three security case.
Let us have few words on the covariance.
Please note that this covariance is a product of two deviations.
For example, as you can see here, it is the expectation of R1t minus R1 bar into R2t minus
R2 bar.
Since the deviations are from two different securities, unlike the variance, we may expect
it to be negative as well, though that is a less practical scenario.
If both the securities move together, that is large positive outcomes and large negative
outcomes are observed for both of these securities together, then the covariance is expected
to be high and positive.
Conversely, if large positive outcomes of one security occur together with a large negative
outcomes of the other security, then the covariance is expected to be high and negative.
If the two securities do not move together, then the covariance is expected to be low.
This covariance between security i and k is standardized in the following manner to obtain
the correlation coefficient as shown here.
We simply divide the covariance between security i and k, that is sigma ik by their respective
standard deviation sigma i sigma k to obtain the correlation between these two securities
rho ik.
Here we have simply divided the covariance by individual deviations, standard deviations
to obtain the standardized measure which is also known as correlation coefficient.
This correlation coefficient is very easy to interpret and it is free of any scale bias.
It varies between plus one and minus one.
To summarize, in this video, we discussed the expected return risk and covariance of
securities in a portfolio.
We also discussed that covariance can be standardized to obtain a more clean measure of correlation
between securities in a portfolio and this measure ranges from plus one to minus one.
Portfolio construction recap two.
In this video, we'll talk about the risk of a more generic N security portfolio case.
Now moving to a more generic case of N securities, while solving these cases requires computer
programming.
However, a small discussion will hurt nobody.
If we carefully examine the terms in the equation four shown earlier, we'll find that these
terms can be segregated into two categories.
Terms that are similar to x1 square sigma 1 square.
For example, you can see here in the three security case, you have sigma p square as
x1 square sigma 1 square plus x2 square sigma 2 square plus x3 square sigma square.
These are the terms and another set of terms like 2 times x1 x2 sigma 1 2 2 times x1 x3
sigma 1 3 and so on.
Terms like x1 square sigma 1 square.
These are called variance terms and terms like x1 x2.
These are called covariance terms.
Now for N security variance, the generalized term can be simply written as xi square summation
of xi square sigma square i equal to 1 to N.
Let us confirm the correctness of this term by applying it for two or three security case
or even more.
A three security case is already provided here.
As we can see, understanding the second term, which is the covariance term, is slightly
more involved.
The term looks like this.
It is summation of j equal to 1 to N, summation of k equal to 1 to N, xj xk times sigma jk.
Please note, this j is not equal to k for all the cases where j equal to k, it becomes
the variance terms.
If you are looking for the coefficient of security 2, which is multiplied with all the
covariance terms, please remember that j equal to 2, k equal to 3 and j equal to 3 and k
equal to 2 will give the same two results, that is sigma 2 3 equal to sigma 3 2.
And that is why we have a multiplication of 2 here.
So this would have been x1 x3 times sigma 1 3 plus x1 x3 times sigma 3 1.
And that is why summation of these two will become 2 times x1 x3 sigma 1 3.
Also the symbol j is not equal to k ensures that the k is not the same as j, which otherwise
would result in the variance term.
So now we can write the full equation for the generalized N security case very simply
as follows, sigma p square equal to summation, sigma p square equal to summation i equal
to 1 to N, xi square sigma i square plus summation j equal to 1 to N times k equal to 1 to N,
xi x, xj xk, xj k, j is not equal to k.
Let us examine this formula further, assume that we are investing equal amounts in each
of these securities.
And in that case, x1 equal to x2 equal to xN equal to 1 upon N.
This means that the variance terms will become 1 upon N squared summation i equal to 1 to
N sigma square or we can take 1 upon N outside to make it 1 summation i equal to 1 to N sigma
square upon N.
The term summation of i equal to 1 to N sigma square upon N represents the average of variances
or average variance, that is sigma i bar square.
So we can write this term as sigma i bar square sort of average variance and thus the summation
of the variance term can also be written in the form of average variance as sigma i bar
square upon N.
This is average variance multiplied by 1 upon N.
So for a portfolio with large number of securities, this variance term will become close to zero
or very small.
As N tends increases, this overall term or summation of variances will become very small.
In fact, if you assume N tends to infinity, this term will become close to zero.
What about covariance terms?
Summation of covariance terms.
Since there are N securities, there will be N into N minus 1 covariance terms.
For example, with 3 securities, there will be 6 such terms.
However, since half of them are the same, that is sigma ij, that is sigma ij equal to
sigma ji, we use a multiple of 2 and only 3 terms are left.
Therefore, if we present the covariance terms like this, that is j equal to 1 to N summation
into k equal to 1 to N summation and xjxk sigma jk, which is equal to ultimately summation
of j equal to 1 times k equal to 1 to N into 1 upon N square sigma jk.
This also assumes equal investment in all the securities.
That means x1 equal to x2 and so on equal to 1 upon N equal investments.
And this can be further simplified like this, N minus 1 upon N into summation j equal to
1 to N into summation k equal to 1 to N times 1 upon N to N minus 1 times sigma jk.
The term j equal to 1 to N times summation k equal to 1 to N into 1 upon N to N minus
1 times sigma jk is the summation of covariances divided by the number of covariances.
Remember, there are N into N minus 1 covariance terms.
So this is also called the average covariance or sigma jk bar, thus all the covariance terms
can be denoted using the average covariance term as N minus 1 upon N into sigma jk bar.
In this case, please note as long as much as you can increase N, this term will approach
to unity.
So if N tends to infinity, this term will approach to 1 and therefore the resulting
variance or summation of all the covariance terms will approach sigma jk bar, which is
precisely the average variance covariance term and it is not equal to zero as happened
in the case of variances.
So this summation of all the covariance term with a very high value of number of securities,
this approaches average covariance sigma jk bar.
So what happens to the overall risk of the portfolio that is summation of variance terms
plus covariance term when we have large number of securities.
So look at this overall total standard deviation of the security portfolio, it converges to
sigma p square equal to 1 upon N times sigma e bar square, which is the average variance
plus N minus 1 upon N times sigma jk bar, which is average covariance.
Now please note for a very large number of securities, that is N increasing to a very
large number, this term will approach to zero or a very small value and this N minus 1 upon
N will approach to unity and therefore we can say that sigma p square will simplify
or approach to this value sigma jk bar.
This is precisely what we observe in this diagram.
Let us discuss this diagram in slightly more detail.
The diagram gives us the intuition that as the number of securities are increased, as
you can see here, the number of securities are increasing, the variance term that represent
the risk of the security is offset.
For example, here till here you have the variance term component and here you have covariance
component.
As the number of securities are increased, this variance term drastically decreases and
for even a small number of securities like 30, it becomes close to zero.
So what is left is the covariance risk that cannot be diversified away.
That means this is also called the bedrock of risk as we discussed earlier and as more
and more securities are added as figure shown here depicts how the specific risk or the
variance term, this variance term dies away and only the covariance risk, this risk remains.
These are the covariance terms that are left to contribute to the risk of portfolio or
the average covariance.
With a fairly large number of securities, this remaining risk is equal to the average
of the portfolio, often called as market risk or systematic risk or non-diversifiable risk.
To summarize, in this video, we discussed the risk or standard deviation of an N security
portfolio.
We found that as the number of securities increase, the variance term or the idiosyncratic
stock specific risk is eliminated and the only risk remaining is non-diversifiable and
is on account of covariances, which is driven by correlations across securities.
This is the benefit of diversification.
Simply by adding more securities to a portfolio, we are able to eliminate the stock specific
idiosyncratic risk.
Mean Variance Framework.
In this video, we'll introduce the Harry Markowitz Mean Variance Framework and its role in portfolio
optimization.
Since 1952, Harry Markowitz's seminal work on portfolio selection, the practice of portfolio
diversification is well known and documented.
In short, by choosing stocks that do not move together or with low correlation, investors
can reduce the standard deviation of portfolio returns.
If a sufficiently large number of past return observations are taken, their distribution
is plotted, then this distribution is fairly close to a normal distribution that we saw
earlier.
Though there are certain issues at tail, for example, there are observations related to
extreme positive and negative values that may cause deviations from the normal distribution
and other issues such as excess kurtosis and fat tails, the normal distribution is
still is considered to define the return distributions very well and uses only two particular parameters,
which is mean and standard deviation or variance.
In nutshell, all the past returns can be defined by these two numbers, that is mean and standard
deviation.
And that is the reason we have discussed the normal distribution in some of the previous
discussions.
Now let us combine two securities to construct a portfolio.
Consider two securities A and B. Security A offers 8% expected return and B offers 18.8%
return.
Standard deviation of A is 13.2% and that of B is 31%.
A portfolio comprising these two securities has its expected return and risk governed
by these two equations, equation 1 and 2, which can be seen here.
For example, expected return of the portfolio can be easily written as W1 times R1 bar,
which is the expected return on security 1 and W2 times R2 bar, which is equation 1
for expected return on the portfolio.
This we have seen earlier.
Similarly, the variance of the portfolio is W1 square sigma 1 square plus W2 square sigma
2 square plus 2 times W1 W2 rho 12 into sigma 1 sigma 2, this equation 2, which also we
have seen earlier.
Now we'll examine all the behavior of returns and standard deviation for different levels
of correlations and different investment proportions, that is W1 and W2 for these two securities.
We have plotted all the possible portfolios of correlations corresponding to rho equal
to 1 in blue, rho 12 equal to 0.5 in red, rho 12 equal to 0 in yellow and rho 12 equal
to minus 0.5 in green and rho 12 equal to minus 1 in black.
Here each curve with a given color indicates all the possible combinations of investment,
that is each possibility is a point here in either of these securities from W1 equal to
0 to W1 equal to 1, where W1 plus W2 equal to 1.
So basically we are varying the proportionate amounts, that is W1 and W2 between 0 and 1
and ensuring that W1 plus W2 equal to 1 constraint is followed.
These graphs are essentially governed by the two return and risk equations that we have
already seen, which is RP equal to W1 R1 plus W2 R2 and sigma P square is equal to W1 square
sigma 1 square W2 square sigma 2 square plus 2 times W1 W2 sigma 1 sigma 2 rho 12, the
equation 1 and 2 that we saw earlier.
Rho 12 here was the correlation coefficient between the two securities 1 and 2 and a very
important parameter.
Now please consider the blue line here with rho 12 equal to 1, a correlation.
In this special case, the equation becomes a straight line with sigma P equal to W1 sigma
1 plus W2 sigma 2, this is the blue line case, this case and it should be clear from
the graph itself that across all the graphs, the lowest amount of diversification, that
is highest portfolio risk sigma P square is achieved for a given level of return associated
with this line.
Why so?
Please notice that for this blue line, for all the combination of W1, W2, the amount
of risk for a given level of return is maximum for all the combination.
For example, if we look at this level of returns, the maximum risk is obtained on this blue
line for all the different correlation combinations.
The minimum correlation, we will discuss the minimum and other combinations also, but it
is very easy to see here that the maximum risk is observed or obtained for this particular
line.
And in this particular line, there is no diversification and the equation of risk sigma P is simply
the weighted average of risk between two securities, that is W1 times sigma 1 plus
W2 times sigma 2, something similar to the weighted average of returns as we computed.
Next, we examine the other extreme case corresponding to rho1 to equal to minus 1 and that correlation
is shown here in green.
This is the case where highest diversification is observed as it carries the highest levels
of return for a given level of risk or minimum level of risk for a given level of return.
In this case, the equation for risk effectively becomes or described by this equation, which
is sigma P square is equal to W1 sigma 1 minus W2 sigma 2 raised to the power 2.
Please remember, this particular equation has two solutions, each representing a straight
line that is line number one is sigma P equal to W1 sigma 1 minus W2 sigma 2, which is this
line and W1 sigma 1 minus W2 sigma 2, which is this line.
So in this case, when sigma P equal to W1 sigma 1 minus W2 sigma 2, when W1 into 1 sigma
1 minus W2 sigma 2 is greater than equal to 0 and sigma P equal to minus W1 sigma 1 minus
W2 sigma 2, when W1 sigma 1 minus W2 sigma 2 is less than 0.
So these two line segments can be represented by these first this black line and the other
one is by this black line.
So these are two line segments and these two lines intersect at sigma P equal to 0, where
precisely W1 equal to sigma 1, W1 sigma 1 equal to W2 sigma 2.
Now this is a very special but an impractical case where you have attained complete diversification
with zero risk.
Such cases of negative correlation are rarely observed and almost okay from theoretical
point of view, but not observed practically.
However, if we assume them to be true in theoretical sense, then there is a possible combination
of weights W1 and W2 such that we can obtain this condition W1 sigma 1 equal to W2 sigma
2 and the risk of the portfolio, such hypothetical portfolio, complete risk becomes zero or neutralized.
This is a special case where minimum risk sigma P equal to 0 is observed.
For all the other cases where rho 12 lies between minus 1 to plus 1 are concave kind
of curves in between these two lines.
For example, notice all the cases will be between this blue or black line.
You will not observe any case such as this, such cases will not be observed.
So all these are concave kind of cases between blue and black line, the extreme cases that
we have discussed.
Any point here would suggest even higher level of risk for a given law of return as compared
to maximum correlation rho equal to 1, which is not feasible.
And any point beyond this black curve would suggest a risk which is lower than even this
negative minus 1 correlation, which is also not possible.
So all the possible combinations for different correlation levels are expected to lie only
between these two black and blue curves and of concave nature.
So there are no convex cases like this outside, which would be outside as the lines will strictly
lie between these two extreme cases of black and blue lines.
Also imagine a concave curve, it would suggest a completely opposite situation to what we
have discussed about diversification.
It would suggest the existence of curves where expected returns will be lower for a given
level of risk as we have considered portfolio of securities.
It would suggest more risk here, if it is of this nature, it would suggest if the curve
is of this nature, it would suggest more risk for a given level of return or less expected
return for a given level of risk, which is spurious and not possible.
So our region of possibility is strictly lying between these black and blue lines.
So this is the region of possibility for all possible combination of correlations and W1,
W2 weights.
One important observation here is that the risk of the portfolio for a given level of
returns is sometimes even less than the least risky security in the portfolio.
So the risk is sometimes even less than, for example, this is the least risky security,
then there are certain combination where risk is even less or on the left of it.
And this is particularly the case when the correlation between the securities is very
low.
For example, on this side, where correlation is low.
On this side, we are expecting low correlation points.
In actual terms, there may be a security that exhibits returns outside this region
of possibilities.
Then what happens?
Surely expectations cannot stay away from the actual returns for a long time if markets
are reasonably efficient.
That means if any security is lying outside of this feasibility region, then market forces
of arbitrage and market efficiency will drive down these points or these returns back to
the normal levels by correcting the points.
So there will be a correction witnessed in the point that will drive down the observed
level of prices to align them with the expected returns.
So we have seen that adding more and more securities to the portfolio here surely lowers
the specific risk of the portfolio.
Even with 15 to 20 stocks, we can achieve a considerable amount of diversification by
neutralizing the stock specific or variance risk.
What happens when we add more and more securities to this two stock portfolio?
How does the feasible region as we saw here earlier, this region changes, which is the
area of all the possibilities.
In the next video, we'll discuss these possibilities when we add more securities to this two stock
portfolio, how this feasibility region changes.
To summarize, in this video, we saw the mean variance framework of Harry Markowitz.
We found that for a correlation equal to one, there's no diversification in risk and the
total risk of the portfolio is simply the weighted average of individual risk or standard
deviations of the two securities, weights being the proportionate amount invested in
these.
We also saw that for a theoretical and hypothetical case where correlation is strictly equal to
minus one, maximum diversification is achieved, which is represented here by two line segments
in the black color.
Theoretically, there is a possibility of a point where overall risk becomes zero for
a given correlation equal to minus one.
However, that is more of a theoretical possibility.
For all the possible combinations of correlation and proportionate amounts invested in the
portfolio, the feasibility region lies between these two curves, that is between this blue,
which is corresponding to sigma equal to correlation equal to one and the black curve, which is
corresponding to correlation equal to minus one.
So all the feasible points lie in this region.
Any other points that are outside this curve, whether on this side or this side are not
possible.
And if there is a stock for which prices are such that the stock lies on this point, market
forces of arbitrage and market forces of efficiency will drive that point or price
of that security within this region in a short frame if the markets are reasonably efficient.
Portfolio Possibilities Curve.
In this video, we'll talk about the feasible region and various portfolio possibilities
with multi-security case.
We start with four hypothetical portfolios.
Different combinations of ABCD, that means ABCD, AC and so on, with different investment
mixtures can lead to a large number of portfolio possibilities as shown here.
These possibilities are shown in the form of curves in the figure.
As we add more and more securities, we get the first diagram like a net cut shape.
As we keep on forming these combinations infinitely, we can add more and more securities.
In addition to these securities, we'll get the convex cut shape as provided here.
This region of possibilities is shown here in blue.
The blue area is effectively the region of expected return and risk possibilities that
an investor can attain.
Each point here represents a combination of risk and returns that is available to investors
in the form of investment into portfolios.
Together, all these points or portfolios comprise region of possibilities or the feasible region.
As we can see here, this is your feasible region of all the possible different combinations.
We started from four securities, A, B and C and D.
Then we kept adding more and more securities to obtain this region of possibility.
Now on this region of possibility, our goal is clear.
We want to move up or increase the returns.
That means we want to move in this direction where for a given risk, we are increasing
the expected returns and we want to move to the left.
That means for a given level of return, we want to reduce the risk.
As we do this, we reach the top surface of region of possibilities, which is this surface
or we can say SS dash on this region of feasibility.
There are no more points where we can move further left or up on this curve.
So this is the extreme available to us, the best possible combination.
And that is why this region is called efficient frontier.
All the points on this region offer the highest return for a given level of risk or lowest
risk for a given level of return.
For example, if you pick this level of return, the lowest amount of risk available for all
the points available in feasible region would be on this, that is on the surface SS dash.
Similarly, if you pick a level of risk, let's say this level of risk, the maximum level
of return available would be on this SS dash line here.
Now on this SS dash or efficient frontier, each investor, depending upon their risk preference,
may choose a specific risk level.
For example, an investor may choose this point if he prefers this level of risk, or if he
prefers a lower level of risk, they can move left or left on this particular efficient
frontier.
Once the investor decides a specific risk level, they will have a given level of fixed
return on the surface SS dash.
So we hope, and we can clearly see here that once the investor reaches this surface of
risk return, they cannot do any better by moving left or up.
Now if you focus on this efficient frontier, which is SS dash, two points are particularly
important for us here.
First point S, this point is the minimum risk point.
So if you look at this efficient frontier on the feasible region, this point has the
minimum risk.
There is no point on the left of this in the feasible region.
So this is the leftmost point and therefore represents the point of minimum risk.
Similarly, point S dash, which is at the top of it, is the point of maximum return.
So there is no point on this feasible region, which is above this point S dash.
And therefore this point offers you the maximum return.
All the combinations of this surface SS dash are called efficient frontier on this entire
feasible region because these points offer you the best combinations of risk and return
possible in the entire feasible region.
So this becomes our efficient frontier.
One interesting point to note here or observe that if you go below this point S, there are
some points that offer you lower return, but these points are not efficient because for
this level of return, you can also obtain this level of risk, you can also obtain this
level of return.
So as compared to this point, you would like to be on this point.
So in this manner, the SS dash remains the efficient frontier, the best combinations
of risk return that are available to you.
To summarize, in this video, we saw that as we keep on adding multiple securities to a
portfolio, we obtain a region of possibilities, which is called feasible region as shown here
in the blue.
This appears to be an egg cut shape.
In order to improve our position or profile on this feasible region, we would like to
move on the left to improve our or lower our risk for a given level of return, or we want
to move up to increase our expected return for a given level of risk.
As we keep on doing that, we obtain a region called efficient frontier or a surface, which
is efficient frontier, which offers us the best possible combinations of risk return
scenarios.
And this surface or this region, which is called efficient frontier starts from the
point with the minimum risk, global minimum risk portfolio, which is S here and the portfolio
which offers the maximum expected return S dash here and this SS dash presents as the
efficient frontier.
Feasible frontiers.
In this video, we'll examine the shape of efficient frontiers that are feasible.
We'll also discuss some of the shapes of efficient frontiers that are not feasible.
Look at the possibility curve shown here.
From what we already know, the figure B is not a possibility because it is convex in
nature.
Notice the convex nature of figure B here.
We already know that there cannot be points to the right of this line joining MB and C.
We know that there cannot be any points to the right of line joining global minimum variance
portfolio, which is MB and C here, which is above MB.
Why so?
Because for any point that lies on this curve, we'll have a risk which is higher than those
points that have perfect correlation, which is equal to 1 that fall on this line.
And therefore, in that case, the points on the convex curve will have more risk or higher
standard deviation than the straight line.
But we already know that the straight line joining MB and C corresponds to the highest
risk, which is correlation equal to 1 and hence it is not possible.
What about figure C?
This figure, please note again, our reasoning remains the same, U and V are also portfolios.
So any portfolio that can be formed by combining U and V would look something like this.
And therefore, any shape which is like this convex shape like this should not be possible.
So the straight line joining this UV points will be one extreme set of portfolios with
the highest risk.
And therefore, the shape of the curve joining UV can only be at max or at best a straight
line or a concave curve, not this kind of convex curve as we have discussed.
And therefore, this figure C is also not a possibility.
Let us also now discuss the points that are below the minimum variance fit here, like
here shown in A, B and C.
With the same reasons as we discussed earlier, only A is the feasible option.
This one is the feasible one.
The curve below the point MV should be convex like this.
And therefore, figure B is also not possible because it would again represent those points
that are having higher risk than the straight line with correlation equal to 1.
Now if U and V are combinations of MV and S, for example, these U and V, they are combination
of MV and S, if the portfolios U and V are mixed, the combination can be a convex curve
at or max a straight line joining UV like this.
So it can be like this or the straight line joining there.
It cannot be a concave curve as shown here, so it cannot be a concave curve and therefore
C is also not a possibility.
So now we have understood the risk-return profile of a combination of two assets and
how feasible frontier would look like.
To summarize, in this video, we discussed the shape of feasible frontiers.
For those points that are lying above the minimum variance portfolio MV, this MV, the
shape of efficient frontier should be a concave curve that is towards the left of the straight
line joining point MV and S, which suggests that there cannot be any portfolio with higher
risk for a given level of return.
Similarly, for those points that are below MV, the curve has to be convex and not concave.
The shape of feasible frontier should be convex not concave because all the concave curves
would indicate those points that have higher risk for a given return than the correlation
Efficient Frontier Scenarios Multi-Security Case 1
In this video, we will talk about different efficient frontiers for multi-security case
first without short sales and then with short sales.
Let us start by examining a case with multiple securities that are distributed in a rather
heterogeneous manner as shown in the figure here.
So here we have six securities.
These are A, B, C, D, E, these six securities, they are different combinations are plotted
here on risk and expected return space.
Before we start our discussion, let us remember two important goals.
One higher return for a given risk that means for a given amount of risk, you would like
to have higher expected returns that is you would like to move up and lower the risk for
a given level of return that is you would like to move on the left.
These are desirable portfolio characteristics.
Now if we compare portfolios A and B here, portfolio A and B in this possibility space,
which is this portfolio possibility space, B of course would be preferable to A since
it offers a higher return for the same level of risk as we can see here similarly C would
be preferable to A because it offers same level of return with lower level of risk.
So it may appear to us by this logic that no portfolio dominates portfolio C, B or for
that matter portfolio E. So these are the dominating portfolios.
This kind of elimination process where we compare the portfolios for their risk return
profile, we can easily see that portfolio C, B, E, they cannot be eliminated and therefore
they are part of the efficient frontier.
Also we can clearly see here that C is the point of global minimum variance that means
there is no portfolio that has lower risk that is on the left of this point.
So no point on this feasibility region is on the left of point C. So no other portfolio
has a lower risk than this portfolio C and therefore we can say the surface C, E, B would
comprise the set of portfolios that offer the highest return for a given level of risk
or alternatively the lowest risk for a given level of return.
Now this curvature, the set of portfolios that lie on this curvature C, E, B are called
efficient set of portfolios or this is our efficient frontier.
Here we also find that this C portfolio is the portfolio with minimum variance and portfolio
B is the portfolio with maximum return.
Also we can call C as global minimum variance portfolio.
Now if we take out the portion of this efficient frontier, it would look something like the
graph shown on A, this A curvature, we know from our previous discussions that the shape
can only be concave, not convex.
The lower end of this frontier represents the portfolio with the minimum variance.
This one, the minimum variance, the higher end of this portfolio is the portfolio with
maximum return across all the portfolios that is security with the highest returns.
Please remember in this case, we cannot have a portfolio looking like B as we have discussed
already.
Any convex nature like between U and V is not possible.
Why?
Because U and V are also portfolios and therefore we have already discussed if all the points
are above minimum variance frontier, that is above this minimum variance point, any
two portfolios, U and V, their combination can be at best a straight line.
And in that case, the straight line would represent portfolios with correlation of one
that is maximum risk.
So any portfolio on the right side of it would represent the set of portfolios that carry
higher risk than those with correlation of one, which is not possible.
And therefore, any set of portfolios on the right side would not be possible.
And therefore, this convex nature of curve is not possible.
Therefore, B is not a feasible scenario.
The only feasible scenario for efficient frontier in this case is this one.
Now, let us talk about a case where short sales is allowed.
Short sale is essentially selling a security one does not own, which is equivalent to taking
a negative position in the security.
This negative position allows one some additional liquidity or funds that one can invest further
in taking a long position or investing in any other purposes.
Now, if you're making the short selling, you would like to sell securities that have lower
expected returns and you would like to buy securities from those proceeds where you have
higher expected returns.
And therefore, those not so efficient securities, which are lower in the efficient frontier
here, you have this opportunity now to short sell these securities because these securities
are offering us lower expected returns.
This suggests that one can keep on short selling securities that are here while investing in
long in securities that are here and extending their position beyond B to C and further high.
Theoretically, this suggests that one can achieve infinite expected returns or very
large amount of expected returns.
But at the same time, when you do that, you also extend the risk of your position.
That is, you also extend the risk or standard deviation to very large levels.
And therefore, one can see that now our actual efficient frontier is not restricted at point
B and one can take positions that can go beyond B and even further and further.
So now on aggregate, our position becomes MVB and beyond to C.
So one can extend their portfolio possibility curve to MVBC, which becomes the new efficient
frontier.
And notice that this is a concave curve, which is similar to earlier without short sale scenario.
But now with short sales, our efficient frontier extends from MV.
Not only it moves up to the security with maximum return, but it also extends beyond
that to large expected returns, which have no finite upper bound.
However, please note the caveat.
In order to achieve that, we are able to or we are assuming a large amount of short sales.
Every market has certain restrictions on short sales, but still to a reasonable extent, we
can assume that one can extend their efficient frontier beyond this point B, which was point
of maximum return further with no upper bound.
To summarize, in this video, we discussed the shape of efficient frontier with and without
short sales.
We found that in the absence of short sales, the efficient frontier extends from portfolio
of minimum variance, portfolio with global minimum variance to portfolio of maximum return,
in this case MV to B.
However, when the restrictions on short sales are removed, one can short sell securities
with poor expected return and invest them in higher expected returns that are on efficient
frontier and extend their possibilities beyond this upper point B.
And in that case, the expected returns would be very higher with no upper bound on this
efficient frontier.
However, at the same time, if one were to obtain such positions, the risk or standard
deviation of such positions would also be very high.
