Collective Behavior in the Stock Market

Collective Behavior in the Stock Market preview image

1 collaborator

Foto elder silva (Author)

Tags

behavioral finance 

Tagged by elder silva over 9 years ago

collective behavior 

Tagged by elder silva over 9 years ago

flash crash, mini-flash crash 

Tagged by elder silva over 9 years ago

stock market 

Tagged by elder silva over 9 years ago

Visible to everyone | Changeable by the author
Model was written in NetLogo 5.0.4 • Viewed 1873 times • Downloaded 147 times • Run 0 times
Download the 'Collective Behavior in the Stock Market' modelDownload this modelEmbed this model

Do you have questions or comments about this model? Ask them here! (You'll first need to log in.)


WHAT IS IT?

The model simulate investor buying and selling shares, the investors are divided into two grups. First, they are investor who are submited to the "Disposition Effect". Second group, they use Stop Order, and they do not have influence of the "Disposition Effect". When the first group come to open a position in the market they consider the opnion of their neighbor, and the "market sentiment", to decide what they will do.

HOW IT WORKS

The investor buy and sell influencied by their neighborhood, and by a "market sentiment". We do not use "fundamental value" or shock in the profits; the series of price emerge as result of a complex system interation between the investors. These ideas you could better understand in the papers of Shiller (1992, and 2000), Kahneman (1974 and 1992), Couzin (2005).

HOW TO USE IT

If you turn off the "Index" the model will run without computation the index. When you insert a value in "I_Index", you will create a "bull / bear" market. Is the same when you receive a news that influence all the market, for example one extreme event like september eleven, or something like this. The shock could be positive or negative. This could be a manner to a central bank try to influence the stock market, for example.

THINGS TO NOTICE

This model try to be close to more empirical fact as possible. The "Index I" is an adaptation of Shiller ideas, in the sense we belief the investor could be influencied by the "market sentiment", when many other traders are buying they tend to generate a bull market. The beta solve a problem - see the code - because we do not agree in how each individual must "trust", or "relevate", the "market sentiment", so we introduce the beta. Than one individual could attribute a good share of their decision to the "market sentiment", while other will pay a little attention to the index.

THINGS TO TRY

The omega mesure the "overconfidence", when the omega is high the investor are more active and trade more frequently. This parameter come from Couzin (2005) equation, they use a similar equation to model the behavior of many species in a biology sence, here we adapt the equation and the omega could take, in our case, the overconfidence interpretation. (https://www.youtube.com/watch?v=2WqHHUxz8)

CREDITS AND REFERENCES

(W.A. Brock, C.H Hommes, Journal of Economic Dynamics and Control 22 (1998) 1235.

J. Kukacka, J. Barunik, Physica A 392 (2013) 5920.

Y. Li, L. Yang, Journal of Financial Economics 107 (2013) 715.

K.M. Kaminski, A.W. Lo, Journal of Financial Markets 18 (2014) 234.

A. Tversky, D. Kahneman, Journal of Risk and Uncertainty 5 (1992) 297.

N. Barberis, W. Xiong, Journal of Finance 64 (2009) 751.

N. Suhadolnik, J. Galimberti, S. Da Silva, Physica A 389 (2010) 5182.

I.D. Couzin, J. Krause, N.R. Franks, S.A. Levin, Nature 433 (2005) 513.

R.J. Shiller, Journal of Psychology and Financial Markets 1 (2000) 49.

V. Plerou, P. Gopikrishnan, X. Gabaix, H.E. Stanley, Physical Review E 66 (2002) 027104.

Comments and Questions

Modeling Collective Behavior in the Stock Market

This model create a simulation of stock market utilizing Prospect Theory, Financial Behavior and Collective Behavior arguments. If you need more details please keep in touch.

Posted about 10 years ago

doubt?

Contact: elder0055@gmail.com

Posted over 9 years ago

Click to Run Model

globals[
  r         ; realized return 
  D         ; demand
  I         ; indice
  p_t-1     ; price t - 1
  p_t       ; price
  buyer     ; total buyer
  seller    ; total seller
  hold      ; total hold
  t_w_DEI   ; time of hold winner shares of DEI agent
  c_w_DEI   ; count of total winner operation DEI agent
  t_l_DEI   ; time of hold loss shares of DEI agent
  c_l_DEI   ; count of total of losso operation DEI agent
  t_w_STP   ; time of hold winner shares of STOP agent
  c_w_STP   ; count of total of winner operation STOP agent
  t_l_STP   ; time of hold loss shares of STOP agent
  c_l_STP   ; count of total of loss operation STOP agent
]

patches-own[
  q            ; buy price
  B            ; information from neighbor -> buy
  S            ; information from neighbor -> sell
  H            ; information from neighbor -> hold
  xi           ; expected return 
  lambda       ; Prospect theory parameter
  delta        ; Stop loss rule parameter
  beta         ; equation 4 parameter
  Pb_B         ; probability to buy - equation 9
  Pb_S         ; probability to sell - equation 9
  Pb_N         ; probability to hold - equation 9
  DEI?         ; disposition-effect investors
  STP?         ; stop-loss investors
  buyer?       ; auxiliary variable
  seller?      ; auxiliary variable
  psell        ; auxiliary variable
  transactions ; total number of transactions
  Wealth       ; wealtg of an agent
  time1        ; auxiliar variable
]

to setup
  if random? = false [random-seed 12345]     ; fixed seed of experiment
  clear-all     ; clean previous simulations
  ask patches[
    let t random-float 1                                 ; auxiliary variable distributed between zero and one
    ifelse t < STP [ set STP? true set DEI? false] [     ; handing agents among STP e DEI
      set STP? false set DEI? true ]                     ; handing agents among STP e DEI
    set buyer? false                                     ; cleaning variable
    set seller? false                                    ; cleaning variable
    set pcolor (white)                                   ; white for agents not operated in the period
    set q 0                                              ; purchase price equal to zero - initial setup
    set psell 0                                          ; selling price of zero - initial setup
    ifelse (i. = 1) or (i. = 0) [ set xi 0.01 ] [ set xi ((x_max / i.) * random i.) + 0.01]    ; create heterogeneous expectations
    set lambda 2.25 * xi                                 ; recording individual value 
    set Wealth 100                                       ; arbitrary starting value of individual Wealth, not influence the results of experiments          
    set transactions 0                                   ; resetting number of transactions the agent
    set delta random-float .02                           ; creating a particular value of the delta, used in the STOP strategy
    set beta random-float xi ]                           ; individualde beta value
  set I 0                                                ; setup indice
  set p_t P_initial                                      ; setup initial price
  set p_t-1 P_initial                                    ; setup initial price
  set t_w_DEI 0
  set c_w_DEI 0
  set t_l_DEI 0
  set c_l_DEI 0
  set t_w_STP 0
  set c_w_STP 0   
  set t_l_STP 0   
  set c_l_STP 0   
  reset-ticks                  
end 

to fetch-information     ; Structure of "Equation 4"   
  ifelse q != 0 [ set B omega + count neighbors with [ pcolor = blue ] ] [ set B count neighbors with [ pcolor = blue ]  ]           ; value for B, adding omega own information using color to verify information from neighbors
  ifelse psell != 0 [ set S omega + count neighbors with [ pcolor = red ] ] [ set S count neighbors with [ pcolor = red ] ]          ; value for S, adding omega own information using color to verify information from neighbors
  ifelse (q = 0) and (psell = 0) [ set H count neighbors with [ pcolor = white ] ] [ set H count neighbors with [ pcolor = white ] ] ; value for H, by adding omega own information using color to verify information from neighbors
  set Pb_B (B / (B + S + H)) + beta * I                                                                                              ; probabilities buy
  set Pb_S (S / (B + S + H)) + beta * I                                                                                              ; probabilities sell
  set Pb_N (H / (B + S + H))                                                                                                         ; probabilities hold
end  

to agent-STP   ; Structure of "Equation 9"
  let t random-float 1                                                                                 ;  auxiliary variable
  ifelse ticks < 100 [ ifelse t < (1 / 3) [ buy ] [ ifelse t > ( 2 / 3) [ sell ] [ not-operate ] ] ] [ ; first 100 simulations are used to setup the model
  ifelse q = 0 [ 
    ifelse (r) > (delta) [ STP-buy ] [ not-operate ]                                                   ; opening position, "equation 9" if the last return is greater than delta agent will operate.
    ] [ 
    ifelse p_t > q [                                                                                   ; checks if the agent closes the position or not, continuing "equation 9"
      ifelse (p_t - q) > (lambda * q) [
        if ticks > 200 [set t_w_STP (t_w_STP + (ticks - time1))]
        if ticks > 200 [set c_w_STP (c_w_STP + 1)]                                                                 
        STP-sell ] [ not-operate ] ] [                                                                 ; Gain more than goal, compulsory closing the position
      ifelse (q - p_t) > (xi * q) [  
        if ticks > 200 [set t_l_STP (t_l_STP + (ticks - time1))]
        if ticks > 200 [set c_l_STP (c_l_STP + 1)]                                                                          
        STP-sell ] [ not-operate ] ] ] ]                                                               ; maximum loss reached, compulsory closing the position
end   

to buy ; structure for the agent to make a purchase
  set buyer? true          ; records the information that the agent made a purchase
  set seller? false        ; records that the agent not made a sale
  set pcolor (blue)        ; blue for agents who buy
  set q p_t                ; records of purchase price q
  set psell 0              ; sales price equal to zero means that the agent not made a sale
  set time1 ticks          ; counting time with posses share
end 

to sell ; structure for the agent making a sale
  set buyer? false         ; means not made a purchase
  set seller? true         ; records that held a sale
  set pcolor (red)         ; red agents to sell.
  set q 0                  ; purchase price equal to zero means that the agent not bought.
  set psell p_t            ; records sales price
  set time1 ticks          ; counting time with posses share
end 

to closing-operation-high ; agents MUST NOT use strategy STOP
  ;; p_t > q ---> GAIN
  ifelse (p_t - q) > (xi * q) [                                        ; this structure means: if (as I am earning) is greater than (My individual goal of GAIN)
    ifelse (random-float 1) < ((p_t - q - (xi * q)) / (xi * q)) [      ; if yes, verifies how much above the target ...
      set transactions transactions + 1                                ; ... and increases the probability as the gain exceeds the goal.
      set Wealth Wealth * (1 + ((p_t - q) / q))                        ; Wealth updates the agent
      set buyer? false      
      set seller? true      
      set pcolor (red)      
      set q 0               
      set psell 0
      if ticks > 200 [set t_w_DEI (t_w_DEI + (ticks - time1))]
      if ticks > 200 [set c_w_DEI (c_w_DEI + 1)]        
      ] [ not-operate ]     
      ] [ not-operate ]     
end 

to closing-operation-falling ; agents MUST NOT use strategy STOP
  ;; p_t < q ---> LOSS
  ifelse (q - p_t) > (lambda * q) [ ; this structure means: if (as I am earning) is greater than (individual goal - bearable loss - individual LOSS)
    ifelse (random-float 1) < ((q - p_t - (q * lambda)) / (lambda * q)) [ ; probability to end position
      set transactions transactions + 1     ; 
      set Wealth Wealth * (1 + ((p_t - q) / q))     ; 
      set buyer? false     ; 
      set seller? true     ; 
      set pcolor (red)     ; 
      set q 0              ; 
      set psell 0          ; 
      if ticks > 200 [set t_l_DEI (t_l_DEI + (ticks - time1))]
      if ticks > 200 [set c_l_DEI (c_l_DEI + 1)]
      ] [ not-operate ]    ; 
      ] [ not-operate ]    ; 
end 

to closing-shortposition-high ; agents MUST NOT use strategy STOP      - Agent conducted an asset sale, gain if the asset falls and loses if the asset rise
  ;; p_t > psell ---> LOSS
  ifelse (p_t - psell) > (lambda * psell) [                                           ; this structure means: if (as I am earning) is greater than (individual goal - bearable loss - individual LOSS)
    ifelse (random-float 1) < ((p_t - psell - (lambda * psell)) / (lambda * psell)) [ ; probability to end position
      set transactions transactions + 1     
      set Wealth Wealth * (1 + ((psell - p_t) / psell))    
      set buyer? true
      set seller? false
      set pcolor (blue)
      set q 0
      set psell 0
      if ticks > 200 [set t_l_DEI (t_l_DEI + (ticks - time1))]
      if ticks > 200 [set c_l_DEI (c_l_DEI + 1)]       
      ] [ not-operate ]
      ] [ not-operate ]
end 

to closing-shortposition-falling ; agents MUST NOT use strategy STOP   
  ;; psell > p_t ---> GAIN
  ifelse (psell - p_t) > (xi * psell) [                                                ; this structure means: if (as I am earning) is greater than (My individual goal of GAIN)
    ifelse (random-float 1) < ((psell - p_t - (xi * psell)) / (xi * psell)) [ ; 
      set transactions transactions + 1
      set Wealth Wealth * (1 + ((psell - p_t) / psell))
      set buyer? true
      set seller? false
      set pcolor (blue)
      set q 0
      set psell 0
      if ticks > 200 [set t_w_DEI (t_w_DEI + (ticks - time1))]
      if ticks > 200 [set c_w_DEI (c_w_DEI + 1)]       
      ] [ not-operate ]
    ] [ not-operate ]
end 

to STP-buy ;; agents using order STOP
  set buyer? true        
  set seller? false     
  set pcolor (blue)     
  set q p_t              
  set psell 0
  set time1 ticks            
end 

to STP-sell ;; agents using order STOP
  set transactions transactions + 1          
  set Wealth Wealth * (1 + ((p_t - q) / q)) 
  set buyer? false
  set seller? true
  set pcolor (red)
  set q 0               
  set psell 0         
end 

to not-operate ;; all agents
  set buyer? false
  set seller? false
  set pcolor (white)
end 

to traders    ; agents in the distribution of heterogeneity received minimum value of "x". Very active agents in the real market would be daytrader.
  let t random-float 1             ; t = auxiliary variable; first 100 simulations are used to setup the model
  ifelse ticks < 100 [ ifelse t < (1 / 3) [ buy ] [ ifelse t > ( 2 / 3) [ sell ] [ not-operate ] ] ] [ fetch-information  ifelse t < ((beta * I) + Pb_B) [ buy ] [ ifelse t > (1 - Pb_S + (beta * I)) [ sell ] [ not-operate ] ] ]
  ifelse q != 0 [ set transactions transactions + 1 set Wealth Wealth * (1 + ((p_t - q) / q)) ] [ if psell != 0 [ set Wealth Wealth * (1 + ((psell - p_t) / psell)) ]  ]
end 

to agent-DEI    ; Finishing structure operation of the agent subject to the disposition effect
  let t random-float 1                                                                                             ; auxiliary variable
  ifelse ticks < 100 [ ifelse t < (1 / 3) [ buy ] [ ifelse t > ( 2 / 3) [ sell ] [ not-operate ] ] ] [             ; first 100 simulations are used to setup the model
  ifelse q != 0 [                                                                                          
    ifelse p_t > q [ closing-operation-high ] [ closing-operation-falling ] ] [                                    ; checks whether the agent is winning or losing, and calls one of the functions
    ifelse psell != 0 [                                                                                            ; agent sold 
      ifelse p_t > psell [ closing-shortposition-high ] [ closing-shortposition-falling ] ] [                      ; checks whether the agent is winning or losing, and calls one of the functions
      fetch-information 
      ifelse t < (Pb_B) [ buy ] [                                                                                  ; uses equation 4 to verify that buys, sells, or does not operate
        ifelse t > (1 - Pb_S) [ sell ] [ not-operate ] ] ] ] ]
end 

to I-indice ; computing indice I
  ifelse (buyer + seller ) != 0 [ ifelse Index = true [ if ((ticks mod 50) = 0) and (ticks > 1) [ set I ((buyer - seller) / (buyer + seller + hold)) + (I_Index) ] ] [ set I 0 ] ] [ set I 0 ]  
end 

to go ; Run
  ask patches[
    ifelse xi = 0.01 [ traders ] [ 
      ifelse STP? [ agent-STP ] [ agent-DEI ] ] ] ; checking which type of agent. Calls one of its functions as above.
  set buyer count patches with [buyer?]             
  set seller count patches with [seller?]
  set hold (max-pycor * max-pxcor) - buyer - seller
  ifelse (buyer + seller) != 0 [ set D ((buyer - seller) / (buyer + seller + hold)) ] [ set D 0 ] ; computing demand period
  set p_t (((exp(D) - exp(- D)) / (exp(D) + exp(- D)) + 1 ) * p_t-1)                              ; computing the price period
  set r (ln(p_t) - ln(p_t-1))                                                                     ; computing the first return period
  set p_t-1 p_t                                                                                   ; updating price of the previous period
  I-indice                                                                                        ; calling function indice I
  tick ; new period
end 
  
  
  
  
  
  
  
  
  
  

There are 4 versions of this model.

Uploaded by When Description Download
elder silva over 9 years ago Info included. Download this version
elder silva almost 10 years ago new version Download this version
elder silva almost 10 years ago new version Download this version
elder silva about 10 years ago Initial upload Download this version

Attached files

File Type Description Last updated
Collective Behavior in the Stock Market.png preview Preview for 'Collective Behavior in the Stock Market' about 10 years ago, by elder silva Download

This model does not have any ancestors.

This model does not have any descendants.