Application of Markov Chain

Following to the article of “EMFPS: Efficient Portfolio of Assets (The Optimization for Risk, Return and Probability)” on the link: http://emfps.blogspot.com/2011_10_09_archive.html

I would like to remind you about Step (7) – Part (A) where we fix Rp and change probabilities into limited range. But how can we fix the probabilities distribution into a limited range? It means that we should anticipate the probabilities distribution for the future. Here I have used from homogeneous Markov Chain as a tool to foresee the probabilities distribution.

Now, let me start again the problem with the example mentioned on above article as follows:

We had the data for the forecasted returns of assets A, B, C, D, E, and F from 2012 to 2017 as follows:

				Assets Return (%)
Year		A	B	C	D	E	F
2012		7	19	7	25	8	17
2013		9	16	11	21	10	15
2014		11	14	13	19	12	13
2015		14	12	16	15	14	11
2016		18	10	20	12	16	9
2017		21	8	23	9	18	6

I consider the people in city or country or the location (M) who are the owners of assets A, B, C, D, E, and F. I assume the owners of each asset as a particular area or state are dealing as follows:

Owners of Assets                 Particular Area (State)

A                                                             S1

B                                                             S2

C                                                             S3

D                                                            S4

E                                                             S5

F                                                             S6

At the first, we need to obtain recent data of the total transactions for each particular area for instance, how many people who are the owners of asset “A “would like to deal only into area “A” or liquid their asset to go area B, C, D, E or F (in the period of exact time). The approach will be certainly based on the balance of Risk and Expected Return rate. Therefore, we have to highlight them that the conditions included Risk and Expected Return is equal for all six assets.

How can we find statistics data for the whole of the transactions?

It is clear; there are two types of the data to collect:

1) Primary data: By using of distributing the survey and questionnaire among the people to know their interest.

2) Secondary data:   Refer to historical data collected from internet and so many finance websites for instance, the Volume of transacted shares in a Stock Index during the period of the last time.

Definitely all people do not fill our questionnaire accurately. Therefore, we should have a cross –section of primary and secondary data. On the other hand, we should have a good PEST analysis and then a good Industry analysis for each asset to confirm the combination of primary and secondary data in which we should perceive all Economic, Politic, Society -Cultural and Technology indicators which are affecting on all transactions.

Now, assume we have made our final collected data as follows:

-In Area S1: 30% of people are interested in dealing into area S1, 12% of people are interested in liquating their asset and deal into area S2, 18% deals into S3, 20% deals into S4, 15% deals into S5, 0.05% deals into S6

- In Area S2: 10% of people are interested in dealing into area S1, 20% of people are interested in liquating their asset and go into area S2, 30% go to S3, 0.05% go to S4, 17% go to S5, 0.18% go to S6

-In Area S3: 20% of people are interested in dealing into area S1, 10% of people are interested in liquating their asset and go into area S2, 12% go to S3, 21% go to S4, 18% go to S5, 21% go to S6

-In Area S4: 10% of people are interested in dealing into area S1, 18% of people are interested in liquating their asset and go into area S2, 15% go to S3, 25% go to S4, 10% go to S5, 20% go to S6

-In Area S5: 10% of people are interested in dealing into area S1, 13% of people are interested in liquating their asset and go into area S2, 25% go to S3, 16% go to S4, 14% go to S5, 25% go to S6

-In Area S6: 20% of people are interested in dealing into area S1, 0.09% of people are interested in liquating their asset and go into area S2, 13% go to S3, 17% go to S4, 25% go to S5, 15% go to S6

To simplify above information, we can use from a matrix which is named as transition or movement matrix (S) as follows:

Matrix (S) =

	S1	S2	S3	S4	S5	S6
S1	0.3	0.12	0.18	0.2	0.15	0.05
S2	0.1	0.2	0.3	0.05	0.17	0.18
S3	0.2	0.1	0.12	0.21	0.18	0.21
S4	0.1	0.18	0.15	0.25	0.1	0.2
S5	0.1	0.13	0.25	0.16	0.14	0.25
S6	0.2	0.09	0.13	0.17	0.25	0.15

We have the starting of the movement (deals) on rows and the ending of the movement (transactions) on columns in the period of the exact time.

Let me model the problem as a Markov Chain (homogeneous) to reach the fixed probabilities distributions in the future (at least until 2017 year in this example).

Of course, we should know that there are three conditions (properties) for each problem to be considered as a Markov Chain as follows:

-Each one of the deals done in this system stays a Risk and Expected Return exactly equal after the transaction for all assets in the period of the distinct time.

-The total percentage transactions made by the people into each area must be equal to 1.

- The primary probabilities distribution is not changed over the distinct time (Matrix (S)).

I as well as know that my assumptions are not exactly accurate. But to solve any problem, we should be able to simplify a complicated problem into boundaries conditions in which we need to sure if our assumptions are reasonable. In fact, there is the fundamental difference between the accurate and the reasonable.

If we consider that Matrix (S) is for the first transaction, we will be able to anticipate the probabilities for the second transaction as follows:

S^2 =

0.1794	0.138	0.1856	0.1863	0.1513	0.1594
0.1597	0.1293	0.1874	0.1633	0.1768	0.1835
0.1675	0.1337	0.1806	0.1832	0.1643	0.1707
0.16	0.1414	0.1821	0.177	0.1646	0.1749
0.1605	0.1289	0.1731	0.1779	0.1757	0.1839
0.1648	0.1328	0.1879	0.1818	0.1597	0.173

After the third transaction:

S^3 =

0.167449	0.135237	0.182456	0.179637	0.16344	0.171781
0.165079	0.132657	0.182574	0.178067	0.166625	0.174998
0.16546	0.134598	0.182678	0.179218	0.164359	0.173687
0.164409	0.134689	0.183509	0.17763	0.165285	0.174478
0.164464	0.133764	0.182849	0.178746	0.165509	0.174668
0.165867	0.134181	0.181737	0.179471	0.164906	0.173838

After the fifth transaction:

S^5 =

0.165483	0.134269	0.182655	0.178807	0.164941	0.173846
0.165455	0.13424	0.182622	0.17884	0.16496	0.173882
0.165456	0.134254	0.182647	0.178809	0.16496	0.173875
0.165446	0.134232	0.182638	0.178808	0.164981	0.173895
0.165443	0.134243	0.182634	0.178818	0.164971	0.173891
0.165457	0.134257	0.182643	0.178814	0.164955	0.173874

After the eighth transaction:

S^8 =

0.165457	0.134249	0.182641	0.178815	0.164962	0.173877
0.165456	0.134249	0.182641	0.178815	0.164962	0.173877
0.165456	0.134249	0.18264	0.178815	0.164962	0.173877
0.165456	0.134249	0.18264	0.178815	0.164962	0.173877
0.165456	0.134249	0.18264	0.178815	0.164962	0.173877
0.165456	0.134249	0.18264	0.178815	0.164962	0.173877

After the thirteenth transaction:

S^13 =

0.165456	0.134249	0.18264	0.178815	0.164962	0.173877
0.165456	0.134249	0.18264	0.178815	0.164962	0.173877
0.165456	0.134249	0.18264	0.178815	0.164962	0.173877
0.165456	0.134249	0.18264	0.178815	0.164962	0.173877
0.165456	0.134249	0.18264	0.178815	0.164962	0.173877
0.165456	0.134249	0.18264	0.178815	0.164962	0.173877

As we can see, after 13 deals and the ending of the period of distinct time, the probabilities distribution is equal for all areas of S1 to S7 as follows:

Areas	S1	S2	S3	S4	S5	S6
Assets	A	B	C	D	E	F
Pr	16.55%	13.42%	18.26%	17.88%	16.50%	17.39%

It has been presented us that asset “C” will be the best option to invest in the future. And so we can take above fixed probabilities distribution for our analysis in said article.

Another application of this Markov Chain (homogeneous) is to track the number of the petitions and deals divided into each area after ending time of all transactions.

Assume we have 100 petitions to deal for each area in starting time of the transactions. We replace the number of petitions into a row vector (transposed vector) which is named “X” as follows:

S1          S2        S3       S4        S5      S6 

100       100      100     100       100    100

X = [100,100,100,100,100,100]

^{After the first deal, we will have below situation for the number of deals:}

^{X * S = [98, 82, 113, 104, 99, 104]}

^{After the thirteenth transaction, we will have below situation for the number of deals:}

^{X * (S^13) = [99, 81, 110, 107, 99,104]}

^{As we can see, the number of deals into each area in the starting time will be approximately equal to the ending time of the transactions.}

Note:  “All spreadsheets and calculation notes are available. The people, who are interested in having my spreadsheets of this method as a template for further practice, do not hesitate to ask me by sending an email to: soleimani_gh@hotmail.com or call me on my cellphone: +989109250225.   Please be informed these spreadsheets are not free of charge.”

 

^{To be continued ………}