Advances in Education, Humanities and Social Science Research ICEACE 2023

ISSN:2790-167X Volume-5-(2023)

Comparison of Markowitz Model and Index Model in

Optimization of Portfolio
Shijun Cao
Shanghai Pinghe Bilingual School, Shanghai, China
[email protected].
Abstract. To compare the differences between Markowitz Model and the Single Index Model, we
have used historical return data for ten stock which belong in groups to three different equity sectors
to practically implement the Markowitz Model and the Single Index Model. We find the stocks in the
same equity sector being noticeably positively correlated even after eliminating systemic risks, which
violates the formal condition of the IM. Also, as a result of very detailed comparison, we can conclude
that the IM serves as an accurate approximation of the MM in practical applications for large enough
number of risky assets. And adding a broad index such as S&P 500 to an existing individual equities
portfolio is improving its properties.
Keywords: Markowitz Model, Single Index Model, Sharpe Ratio.

1. Introduction
Since the 1980s, people have focused on stocks and securities in developed and emerging markets.
In the last 30 years, there have been wider choices of assets and asset classes available for use in asset
allocation. From the perspective of expected return, the stock market has great attraction, but they
also have considerable risks. From the Latin American debt crisis in the 1980s to the Technology
Bubble in 2000 and the Subprime Lending Crisis of 2008, these examples illustrate how destructive
of the investor capital the equity markets can be. To avoid repeating history, we should learn historical
lessons profoundly. This is why people are paying more and more attention to investment risk
management and investment returns.
Portfolio theory is identified as the quantitative analysis of optimal risk management. As early as
1952, Markowitz [1] introduced the mean-variance theory, which pioneered using quantitative ideas
to construct portfolios. Since then, there have been other scholars who have further investigated the
Markowitz model. For example, in 1979 Love [3] attempted to develop a model based on the
Markowitz model that would allow one to study the effect of diversification on export losses. In 1993
Gollinger et al. [4] made the first attempt to calculate the efficient frontier of a commercial loan
portfolio based on the structure of the Markowitz equity portfolio model. More recently, in 2020,
Shadabfar et al. [5] used a probabilistic approach to optimal portfolio selection using a mixture of
Monte Carlo simulation and the Markowitz Model.
To further refine the Markowitz Model, in 1963 William Shape proposed the Single Index Model
[2], which significantly promoted the practical application of portfolio theory. Other scholars have
further investigated it since then. For example, in 1986 Collins et al. [6] used Single Index Model for
risk analysis in farm planning applications. In 2010 Galea et al. [7] studied structural Sharpe models
under t-distribution. In 2018 Mallikharjunarao et al. [8] constructed optimal portfolios in two sectors:
IT and Pharma, utilizing Sharpe index models.
On this basis, other scholars have compared these two models. For instance, in 1989 Seler [9]
compared the Markowitz Mean-Variance Model and Sharpe Single Index Model to construct the
Istanbul Stock Exchange portfolios for the period 1986-1987. In 2011 Bekhet et al. [10] compared
the Markowitz and Single Index Models to construct portfolios of Amman Stock Exchange (ASE)
companies. In 2021 Susanti et al. [11] compare the best portfolio formation results of the Markowitz
and single index models for LQ index 45 in the COVID-19 pandemic.
To make a better comparison of the Markowitz Model and the single index model, we select ten
stocks from three industries over the past 20 years as the testing sample of stocks, and the Markowitz
Model and the Index Model are used to construct portfolios with different constraints. We compared

342
 Advances in Education, Humanities and Social Science Research ICEACE 2023
ISSN:2790-167X Volume-5-(2023)
the differences between the two models by calculating two crucial points on the Efficient Frontier:
the Minimal Risk portfolio and the Efficient Risky Portfolio (or Maximum Sharpe Portfolio).
Furthermore, we tested whether a broad equity index, which is a diversified stock portfolio, added to
the model will make a difference between two models.
We organized the rest of the article as follows. An introduction to the theory of risky portfolios
models is provided in Section 2. In Section 3, we preprocess the data and calculate the matrix of
correlation coefficients between stocks. In Sections 4 analyze the MM and IM model results.
Conclusions and future research directions are presented in Section 5.

2. The MM Model and The IS Model

The Markowitz Model (MM) is based on the following assumptions:
1. The investor considers each possible investment choice based on the distribution of assets
return over the time of a given position.
2. The investor estimates the risk of a portfolio based on the variance or standard deviation of
the expected return of the assets.
3. The investor’s decision is solely based on the risk and expected return of the assets and
securities.
4. For a given level of risk, the investor prefers to maximize the expected return; or for a given
level of expected return, the investor prefers to minimize the risk.

The expected return for the Markowitz Model (MM) is:

𝑅𝑅𝑝𝑝 = ∑𝑛𝑛𝑖𝑖=1 𝑤𝑤𝑖𝑖 𝑟𝑟𝑖𝑖 .
𝜎𝜎𝑝𝑝 = �∑𝑛𝑛𝑖𝑖=1 ∑𝑛𝑛𝑗𝑗=1 𝑤𝑤𝑖𝑖 𝑤𝑤𝑗𝑗 𝐶𝐶𝐶𝐶𝐶𝐶(𝑟𝑟𝑖𝑖 , 𝑟𝑟𝑗𝑗 ).
where we denoted:
𝑟𝑟𝑖𝑖 , 𝑟𝑟𝑗𝑗 : The expected return of the assets 𝑖𝑖 and 𝑗𝑗;
𝑤𝑤𝑖𝑖 , 𝑤𝑤𝑗𝑗 : The proportions of assets 𝑖𝑖 and 𝑗𝑗 in the portfolio;
𝑖𝑖 and 𝑗𝑗: The indices enumerating the two assets;
𝐶𝐶𝐶𝐶𝐶𝐶�𝑟𝑟𝑖𝑖 , 𝑟𝑟𝑗𝑗 �: The covariance between the two assets, which measures the strength of correlation
between two assets.
William Sharpe’s Single-Index Model is based on the following assumptions:
1. The risk of the portfolio is divided into systematic risk and nonsystematic or idiosyncratic risk.
The external factor (the index) will not affect the nonsystematic risk.
2. The idiosyncratic risk of one asset will not affect the idiosyncratic risk of another asset. The
correlation of two assets depends solely on the joint response of the external factors:
The expected return for the Single-Index model is:
𝑅𝑅𝑝𝑝 = ∑𝑛𝑛𝑖𝑖=1 𝑤𝑤𝑖𝑖 𝑟𝑟𝑖𝑖 .
The standard deviation for the Single-Index model is:
𝜎𝜎𝑝𝑝 = �∑𝑛𝑛𝑖𝑖=1(𝑤𝑤𝑖𝑖 𝛽𝛽𝑖𝑖 )2 𝜎𝜎𝑀𝑀 2 + ∑𝑛𝑛𝑖𝑖=1 𝑤𝑤𝑖𝑖 2 𝜎𝜎𝑖𝑖 2 .
where we denoted:
𝑟𝑟𝑖𝑖 , 𝑟𝑟𝑗𝑗 : The expected returns of the assets i and j;
𝑤𝑤𝑖𝑖 : The proportion that asset i occupies in our portfolio;
𝑛𝑛: The number of assets;
𝛽𝛽𝑖𝑖 : The risk factor of asset i;
𝜎𝜎𝑀𝑀 : The systematic risk;
𝜎𝜎𝑖𝑖 : The nonsystematic risk.
Theoretically, IM model can be regarded as a simplified version of MM model. The MM model
requires to calculate the correlation coefficient between every two assets. However, in the actual
process of portfolio construction, there are too many assets to choose, which makes the calculation

343
 Advances in Education, Humanities and Social Science Research ICEACE 2023
ISSN:2790-167X Volume-5-(2023)
of the covariance matrix of the portfolio too many parameters to be calculated, which brings more
inconvenience in the actual use process. The IM model simplifies the way of calculating the
covariance matrix of the MM model, and converts the covariance between two assets into the specific
risk of each asset and the exposure to systemic risk, which greatly reduces the number of parameters
that the model needs to calculate when constructing the optimal portfolio.

3. Data and descriptive statistics

Even with the use of IM theory, there are too many assets in the market that can be used to build
portfolios. In order to make our analysis process as simple as possible without losing
representativeness, we have selected several representative companies from the three industries of
technology, financial services and industry as the assets used to build portfolios. Companies from the
financial Services include: Bank of America Corporation(BAC), Citigroup Inc.(C), Wells Fargo &
Company(WFC), The Travelers Companies, Inc.(TRV). Companies from the Industry include:
Southwest Airlines(LUV), Alaska Air Group, Inc.(ALK), Hawaiian Holdings, Inc.(HA). Companies
from the technology industry include: Adobe Inc.(ADBE), International Business Machines
Corporation(IBM), SAP SE(SAP). In order to reduce the impact of the non-normal distribution of
specific risk on the IM model, we use the monthly yield data for analysis. The range of dates for
which we have gathered is from 05/2001 to 5/2021, which approximately corresponds to the previous
20 years of historical data. The risk-free interest rate is assumed to be 1.30%, that is, the average yield
of three-month US treasury over the past 20 years The basic information about the stocks used is
shown in the following table:
Table 1: descriptive statistics of row return
ADB
BAC C WFC TRV LUV ALK HA IBM SAP
E
Average 0.83 0.04 0.62 0.69 0.92 1.56 2.42 1.73 0.30 0.99
Return % % % % % % % % % %
Standard 11.35 12.28 8.15 5.78 9.15 10.87 17.90 9.19 6.72 9.81
Deviation % % % % % % % % % %
0.56 0.57 0.84 1.27 1.14 1.68 1.60 2.77 0.33 0.66
Median
% % % % % % % % % %
72.66 68.67 40.52 19.74 32.29 34.52 99.28 28.08 35.38 70.13
Max
% % % % % % % % % %
53.27 57.75 35.98 19.81 26.57 43.58 50.00 32.51 23.66 41.56
Min
% % % % % % % % % %

In order to better understand the correlation between the returns of these companies, we calculated
the correlation coefficient matrix between the companies. From the correlation coefficient matrix, we
can see that the returns of companies in the same industry are highly correlated, which is particularly
prominent in the financial services industry. At the same time, in order to better analyze the
relationship between market returns and individual stock returns, we added market returns
represented by SPX to the correlation coefficient matrix. We find that the financial services industry
and the technology industry have a higher correlation with market returns than industry, which in a
sense reflects that the financial services industry and the technology industry have a higher weight
than industry in the economic structure of the United States. For better visualization, we have
exhibited the data as both numerical and as a heatmap coloring the numerical data in the Table2 below.
The darker color of a table cell signifies the larger correlation coefficient between the two assets. The
lighter color of a table cell signifies the weaker correlation coefficient between the two assets.
Table 2: Correlation coefficient matrix for row return

344
 Advances in Education, Humanities and Social Science Research ICEACE 2023
ISSN:2790-167X Volume-5-(2023)
BAC C WFC TRV LUV ALK HA ADBE IBM SAP SPX
BAC 1.000
C 0.824 1.000
WFC 0.760 0.701 1.000
TRV 0.388 0.511 0.340 1.000
LUV 0.428 0.427 0.401 0.406 1.000
ALK 0.279 0.302 0.342 0.361 0.517 1.000
HA 0.334 0.342 0.354 0.240 0.422 0.402 1.000
ADBE 0.424 0.464 0.293 0.437 0.379 0.226 0.174 1.000
IBM 0.312 0.409 0.264 0.373 0.337 0.347 0.240 0.450 1.000
SAP 0.331 0.429 0.297 0.366 0.313 0.282 0.142 0.537 0.586 1.000
SPX 0.601 0.695 0.548 0.594 0.531 0.460 0.385 0.654 0.638 0.643 1.000

The IS model assumes that the risks in the company can be divided into systematic risks and
idiosyncratic risks, among which the idiosyncratic risks of different companies should be uncorrelated.
In order to extract the idiosyncratic risk of each company, we use the data of the past 20 years for
regression to get the exposure of each company to systemic risk β, and then we calculate the
idiosyncratic risk of each company. The results of regression are shown in Table 3.
Table 3: Regression results of each company
BAC C WFC TRV LUV ALK HA ADBE IBM SAP SPX
β 1.60 2.01 1.05 0.80 1.15 1.18 1.63 1.42 1.01 1.48 1.00
α -0.01 -0.14 0.01 0.03 0.01 0.09 0.15 0.09 -0.03 0.01 0.00
Residual
31.4% 30.3% 23.4% 16.0% 26.8% 33.4% 57.2% 23.8% 17.6% 25.8% 0.0%
Stdev

From Table 3, we can see that C (Citibank) has the highest exposure to systemic risk, while HA
(Hawaiian Airlines) has the highest idiosyncratic risk. From the perspective of business results, HA
(Hawaiian Airlines) has the highest α，And C (Citibank) has the smallest α.
We are more concerned about whether there is correlation between the regression residuals of each
company. For this reason, we calculated the correlation matrix of the residuals of each company, and
the results are shown in Table 4.
Table 4: Correlation coefficient matrix for residuals
BAC C WFC TRV LUV ALK HA ADBE IBM SAP SPX
BAC 1.000
C 0.707 1.000
WFC 0.644 0.532 1.000
TRV 0.049 0.170 0.021 1.000
LUV 0.160 0.094 0.154 0.132 1.000
ALK 0.001 -0.031 0.120 0.121 0.360 1.000
HA 0.138 0.112 0.184 0.014 0.277 0.274 1.000
ADBE 0.052 0.018 -0.103 0.081 0.048 -0.113 -0.111 1.000
IBM -0.115 -0.059 -0.131 -0.007 -0.002 0.077 -0.008 0.058 1.000
SAP -0.090 -0.032 -0.087 -0.026 -0.045 -0.023 -0.150 0.202 0.299 1.000
SPY 0.019 0.032 -0.021 -0.034 -0.115 -0.022 -0.014 -0.049 0.052 -0.003 1.000

From Table 4, we can see that there is basically no correlation between the residual items among
different industries, but within the financial services industry, there is still a high correlation between
the residual items among companies, which violates the premise of the IS model and, in a sense,
implies the existence of other industry-related risks for scholars.

4. The Comparison of MM and IM model

We compared the differences between the two models by calculating two crucial points on the
Efficient Frontier: the Minimal Risk portfolio and the Efficient Risky Portfolio (or Maximum Sharpe
Portfolio).
First of all, we consider a free problem. That is, when the index can be invested and all assets can
be short indefinitely, what is the difference between the minimum risk portfolio and the efficient risk
portfolio calculated by the MM model and the IS model. Figure 1 and Figure 2 show the proportion
of each asset in the minimum risk portfolio and the optimal risk portfolio calculated using MM model
and IS model respectively.

345
 Advances in Education, Humanities and Social Science Research ICEACE 2023
ISSN:2790-167X Volume-5-(2023)

160.00%
140.00%
120.00%
100.00%
80.00%
60.00%
40.00%
20.00%
0.00%
BAC C WFC TRV LUV ALK HA ADBE IBM SAP SPX
-20.00%
-40.00%

MM model IS model

Figure 1: Proportion of assets in the minimum risk portfolio under free problem

60.00%

40.00%

20.00%

0.00%
BAC C WFC TRV LUV ALK HA ADBE IBM SAP SPX
-20.00%

-40.00%

-60.00%

-80.00%

MM model IS model

Figure 2: Proportion of assets in the optimal risk portfolio under free problem
From Figure 1 and Figure 2, it can be seen that the proportion of assets calculated using different
models has a certain gap in value, but the direction is basically the same. In order to better see the
difference between the two models, we will further observe the income, standard version and Sharp
ratio of the portfolio. This result is shown in Table 5.
Table 5: the results of the problem under free problem
Model Return Stdev Sharpe Ratio
MM model 6.72% 11.75% 0.461
Min VAR
IS model 5.85% 11.95% 0.381
MM model 22.07% 21.33% 0.974
Max Sharpe
IS model 19.81% 21.99% 0.842

From the results in Table 5, we can see that the minimum risk portfolio and optimal risk portfolio
constructed by MM model has higher returns and smaller variance than IS model. But the difference
between the calculation results of the two models is not very big.

346
 Advances in Education, Humanities and Social Science Research ICEACE 2023
ISSN:2790-167X Volume-5-(2023)
To better simulate the Regulation T by FINRA, which allows broker-dealers to allow their
customers to have positions, 50% or more of which are funded by the customer’s account equity, we
add additional optimization constraint is: ∑11
𝑖𝑖=1|𝑤𝑤𝑖𝑖 | ≤ 2. Figure 3 and Figure 4 show the proportion
of each asset in the minimum risk portfolio and the optimal risk portfolio

160.00%
140.00%
120.00%
100.00%
80.00%
60.00%
40.00%
20.00%
0.00%
BAC C WFC TRV LUV ALK HA ADBE IBM SAP SPX
-20.00%
-40.00%

MM model IS model

Figure 3: Proportion of assets in the minimum risk portfolio under Regulation T

120.00%

100.00%

80.00%

60.00%

40.00%

20.00%

0.00%
BAC C WFC TRV LUV ALK HA ADBE IBM SAP SPX
-20.00%

-40.00%

-60.00%

MM model IS model

Figure 4: Proportion of assets in the optimal risk portfolio under Regulation T

From the results of Figure 3 and Figure 4, after adding regulation T as the most restrictive condition,
the calculated proportion of each asset has a certain change compared with unrestricted, but the
direction remains basically unchanged.
Table 6: the results of the problem under Regulation T
Model Return Stdev Sharpe Ratio
MM model 6.69% 11.75% 0.459
Min VAR
IS model 6.07% 11.96% 0.399
MM model 11.55% 14.40% 0.712
Max Sharpe
IS model 18.90% 21.04% 0.837

347
 Advances in Education, Humanities and Social Science Research ICEACE 2023
ISSN:2790-167X Volume-5-(2023)
From the results in Table 6, we can see that the minimum risk portfolio constructed by MM model
has higher returns and smaller variance when constructing the minimum risk portfolio. When
constructing the optimal portfolio, the optimal risk portfolio constructed by IS model has a higher
return and Sharp ratio. Compared with the case without constraints, the Sharpe ratio of the minimum
risk portfolio and the optimal risk portfolio decreased after adding constraints.
From the previous two cases, SPX has a large proportion in the portfolio. Lastly, we would like to
see if the exclusion of the broad index into our portfolio has positive or negative effect. The results
are shown in Table 7.
Table 7: the results when exclude SPX
Model Return Stdev Sharpe Ratio
MM model 9.38% 15.45% 0.523
Min VAR
IS model 9.26% 16.64% 0.478
MM model 26.53% 25.98% 0.971
Max Sharpe
IS model 23.79% 26.68% 0.843
From the table 7, we can see that when calculating the minimum risk portfolio, the standard
deviation of the minimum risk portfolio is less than 12% when there are no constraints or short
constraints, and when SPX is excluded, even when there are no other constraints, the standard
deviation of the minimum risk portfolio calculated by the two models exceeds 15%. When calculating
the optimal risk portfolio, when SPX is excluded, the maximum Sharp ratio is very close to that in
the free problem.
From this conclusion, we can see that adding SPX as an investable asset has a strong positive effect
in building the minimum risk portfolio, but when building the optimal risk portfolio, without
considering the transaction friction, we can build a portfolio similar to the index through diversified
asset portfolios, so adding SPX can also have a certain positive effect in building the optimal risk
portfolio, But the effect is not very significant.

5. Conclusion
To make a better comparison of the Markowitz Model and the single index model, we select ten
stocks from three industries over the past 20 years as the testing sample of stocks, and the Markowitz
Model and the Index Model are used to construct portfolios with different constraints. In this work
we have arrived to the following conclusions. First, the stocks in the same equity sector being
noticeably positively correlated. Even after eliminating systemic risks, there is still a high correlation
between enterprises in the financial services industry, which violates the formal condition of the IM
model. in a sense, it implies the existence of other industry-related risks. Second, as a result of very
detailed comparison between the MM and IM models, we can convincingly conclude that the IM
serves as an accurate approximation of the MM in practical applications for large enough number of
risky assets. Lastly, we find that adding a broad index such as S&P 500 to an existing individual
equities portfolio is improving its properties. Our investigation provides a lot of numerical evidence
to support these conclusions. Still, We find that MM model can calculate the effective frontier more
accurately than IM model when the number of assets is small, but as the number of assets available
increases, the difficulty of parameter estimation will make IS model more practical.
However, our investigation still has some limitations. One of the limitations is that we have only
used the data for ten stocks, which is very limited. There are thousands of assets in the market and
the data of ten stocks is not diverse enough for us to be certain that the portfolio of other stocks, funds,
or bonds will result in the same conclusions that we have drawn with 10 stocks. Additionally, the
Markowitz Model needs the data on expected return and the expected standard deviation. In our study
we have instead estimated then from sample historical data. Calculation of the actual expected values

348
 Advances in Education, Humanities and Social Science Research ICEACE 2023
ISSN:2790-167X Volume-5-(2023)
will be extremely hard. Also, the current investigation only compares the IM and MM results under
the free problem and short-sales constraints, the results in other constraints are not investigated.

References
[1] Markowitz, H.M. (March 1952). "Portfolio Selection". The Journal of Finance. 7 (1): 77–91.
doi:10.2307/2975974. JSTOR 2975974.
[2] Sharpe, William F. (1963). "A Simplified Model for Portfolio Analysis". Management Science. 9 (2):
277–93. doi:10.1287/mnsc.9.2.277. S2CID 55778045.
[3] Love, J. (1979). “A Model of Trade Diversification Based on the Markowitz Model of Portfolio Analysis.”
Journal of Development Studies, 15(2), pp. 233-241.
[4] Gollinger T.L., Morgan J.B. (January 1993). “Calculation of an Efficient Frontier for a Commercial Loan
Portfolio”. The Journal of Portfolio Management. 19 (2): 39-46. doi: 10.3905/jpm.1993.409434.
[5] Shadabfar M., Cheng L. “Probabilistic Approach for Optimal Portfolio Selection Using a Hybrid Monte
Carlo Simulation and Markowitz Model”. Alexandria Engineering Journal, (October 2020). 59 (5): 3381-
3393. doi.org/10.1016/j.aej.2020.05.006.
[6] Collins R.A., Barry P.J. “Risk Analysis with Single-Index Portfolio Models: An Application to Farm
Planning”. American Journal of Agricultural Economics, (February 1986). 68 (1): 152-161.
doi.org/10.2307/1241660, jstor.org/stable/1241660.
[7] Galea M., Cademartori D., Vilca F. “The Structural Sharpe Model under t-Distributions”. Journal of
Applied Statistics, (November 2010). 37 (12): 1979-1990. doi.org/10.1080/02664760903207316.
[8] Mallikharjunarao T., Anithadevi S., “Construction of Optimal Portfolio under Sharpe Single Index Model:
A Comparative Study on IT and Pharma Sectors”. International Journal of Management, IT & Engineering,
(October 2018). 8 (10): 114-133.
[9] Seler, İ.T., “Portfolio Selection Methods: an Application to İstanbul Securities Exchange”. Alternate title:
“Portföy Seçim Yöntemleri : İstanbul Menkul Kıymetler Borsası İçIn bit Uygulama” Bilkent Universitesi,
(Turkey) ProQuest Dissertations Publishing, 1989. 29045457.
[10] Bekhet, H.A., Matar, A., “Analyzing Risk-Adjusted Performance: Markowitz and Single-Index
Approaches in Amman Stock Exchange”. International Conference on Management (June 2011). 305-
321.
[11] Susanti E., Ervina N., Grace E., Simatupang S., “The Accuracy of Balance Model in Predicting Stock
Investment During The Covid-19 Pandemic on LQ 45 Index”. Organum (December 2021). 4 (2): 215-
227. doi.org/10.35138/organum.v4i2.193.

349