= max-ticks
[
output-print (word "Total wealth: " precision total-wealth 2)
output-print (word "Mean wealth: " precision mean-wealth 2)
stop
]
end
;
; INITIALIZATION ROUTINES
;
;
to initialize-globals
; set num-investors 100
set profit-mean 5000
; set failure-min 0.01
; set failure-max 0.1
set n-hplr 0
set n-hphr 0
set n-lplr 0
set n-lphr 0
set n-same 0
set mv-hplr 0
set mv-hphr 0
set mv-lplr 0
set mv-lphr 0
set moves 0
if failure-min > failure-max
[
user-message "ERROR: failure-min must be less than or equal to failure-max"
set failure-max failure-min
]
set time-horizon 5 ; years
set g-mean-wealth 0
set g-max-utility (profit-mean * time-horizon) * (1 - failure-min) ^ time-horizon
; set sense-radius 1
; set use-sense-radius? false
; set max-ticks 25
set-default-shape turtles "circle"
end
to initialize-patches
ask patches
[
set profit random-exponential profit-mean
set p-failure failure-min + random-float (failure-max - failure-min)
]
set g-max-utility max [expected-utility nobody] of patches
color-patches
end
to initialize-turtle
move-to one-of patches with [ not any? turtles-here ]
set wealth 0
set size 0.8
color-turtle 1.0
end
;
; TURTLE PROCESSING
;
;
to update-wealth ; turtle procedure
set wealth wealth + profit
if random-float 1.0 < p-failure [ set wealth 0 ]
end
to move
let bp find-best-patch
if [profit] of bp > profit and [p-failure] of bp < p-failure [ set n-hplr n-hplr + 1 ]
if [profit] of bp > profit and [p-failure] of bp > p-failure [ set n-hphr n-hphr + 1 ]
if [profit] of bp < profit and [p-failure] of bp < p-failure [ set n-lplr n-lplr + 1 ]
if [profit] of bp < profit and [p-failure] of bp > p-failure [ set n-lphr n-lphr + 1 ]
if bp = patch-here [ set n-same n-same + 1 ]
set moves moves + 1
move-to bp
end
to-report expected-utility [a-turtle]; patch reporter
ifelse a-turtle = nobody
[
report (g-mean-wealth + time-horizon * profit) * (1.0 - p-failure) ^ time-horizon
]
[
report ([wealth] of a-turtle + time-horizon * profit) * (1.0 - p-failure) ^ time-horizon
]
end
to-report find-best-patch ; turtle reporter
let candidates neighbors with [ not any? turtles-here ]
let alt-hi-profit-lo-risk candidates with [
(profit > [profit] of myself) and (p-failure < [p-failure] of myself) ]
let alt-hi-profit-hi-risk candidates with [
(profit > [profit] of myself) and (p-failure > [p-failure] of myself) ]
let alt-lo-profit-lo-risk candidates with [
(profit < [profit] of myself) and (p-failure < [p-failure] of myself) ]
let alt-lo-profit-hi-risk candidates with [
(profit < [profit] of myself) and (p-failure > [p-failure] of myself) ]
let best-candidate patch-here
let all-candidates (patch-set candidates patch-here)
if any? alt-hi-profit-lo-risk [ set mv-hplr mv-hplr + 1]
if any? alt-hi-profit-hi-risk [ set mv-hphr mv-hphr + 1]
if any? alt-lo-profit-lo-risk [ set mv-lplr mv-lplr + 1]
if any? alt-lo-profit-hi-risk [ set mv-lphr mv-lphr + 1]
ifelse stochastic?
[
ifelse any? alt-hi-profit-lo-risk and (random-bernoulli (p-high-profit-low-risk / 100))
[ set best-candidate one-of alt-hi-profit-lo-risk ]
[
ifelse any? alt-hi-profit-hi-risk and (random-bernoulli (p-high-profit-high-risk / 100))
[ set best-candidate one-of alt-hi-profit-hi-risk ]
[
ifelse any? alt-lo-profit-lo-risk and (random-bernoulli (p-low-profit-low-risk / 100))
[ set best-candidate one-of alt-lo-profit-lo-risk ]
[
if any? alt-lo-profit-hi-risk and (random-bernoulli (p-low-profit-high-risk / 100))
[ set best-candidate one-of alt-lo-profit-hi-risk ]
]
]
]
]
[
set best-candidate max-one-of all-candidates [expected-utility myself]
]
report best-candidate
end
;
; OBSERVER REPORTERS
;
;
to-report total-wealth ; observer reporter
report sum [wealth] of turtles
end
to-report mean-wealth ; observer reporter
report mean [wealth] of turtles
end
to-report stdev-wealth ; observer reporter
report standard-deviation [wealth] of turtles
end
;
;
; COLORING ROUTINES
;
;
to color-patches
;ifelse patch-color-mode = 1
ifelse color-patches-by = "profit"
[
let min-profit 0
let max-profit max [profit] of patches
ask patches
[
set pcolor scale-color blue profit min-profit max-profit
]
]
[
ifelse color-patches-by = "p-failure"
[
ask patches
[
ifelse (failure-min < failure-max) and (failure-min > 0)
[
set pcolor scale-color red (ln p-failure) (ln failure-max) (ln failure-min)
]
[
set pcolor scale-color red (p-failure) 1 0
]
]
]
[
ask patches
[
set pcolor scale-color magenta (expected-utility nobody) 0 g-max-utility
]
]
]
end
to color-turtle [max-wealth]
ifelse turtle-coloring-mode = "wealth"
[
ifelse max-wealth > 0
[
set color scale-color green wealth 0 max-wealth
]
[
set color black
]
]
[
set color red
]
end
to color-turtles
let max-wealth max [wealth] of turtles
ask turtles
[
color-turtle max-wealth
]
end
to-report random-bernoulli [p]
if p < 0 or p > 1 [
user-message (word
"Warning in random-bernoulli: p = " p)
]
report random-float 1.0 < p
end
to-report p-hplr
if mv-hplr > 0
[ report n-hplr / mv-hplr ]
report 0
end
to-report p-lplr
if mv-lplr > 0
[ report n-lplr / mv-lplr ]
report 0
end
to-report p-hphr
if mv-hphr > 0
[ report n-hphr / mv-hphr ]
report 0
end
to-report p-lphr
if mv-lphr > 0
[ report n-lphr / mv-lphr ]
report 0
end
to-report p-same
if moves > 0
[ report n-same / moves ]
report 0
end
to-report pct-format [x n-places]
report word (precision (100 * x) n-places) "%"
end
to get-percentages
let runs 0
let n [ 0 0 0 0 0 ]
let mv [ 0 0 0 0 0 ]
let p [ 0 0 0 0 0 ]
let total-w (list)
let mean-w (list)
repeat 1000 [
setup
set runs runs + 1
while [ticks < max-ticks] [ go ]
set total-w lput (sum [wealth] of turtles) total-w
set mean-w lput (mean [wealth] of turtles) mean-w
set n (map + n (list n-hplr n-hphr n-lplr n-lphr n-same))
set mv (map + mv (list mv-hplr mv-hphr mv-lplr mv-lphr moves))
type word runs " "
if (runs mod 50) = 0 [print ""]
]
set n-hplr item 0 n
set n-hphr item 1 n
set n-lplr item 2 n
set n-lphr item 3 n
set n-same item 4 n
set mv-hplr item 0 mv
set mv-hphr item 1 mv
set mv-lplr item 2 mv
set mv-lphr item 3 mv
set moves item 4 mv
set p (map / n mv)
set p (map [x -> precision (100 * x) 2] p)
show n
show mv
show p
output-print (word "Total wealth = " precision mean(total-w) 2)
output-print (word "Mean wealth = " precision mean(mean-w) 2)
set p-high-profit-low-risk precision (item 0 p) 1
set p-high-profit-high-risk precision (item 1 p) 1
set p-low-profit-low-risk precision (item 2 p) 1
set p-low-profit-high-risk precision (item 3 p) 1
end]]>pct-format p-hplr 1pct-format p-hphr 1pct-format p-lplr 1pct-format p-lphr 1pct-format p-same 1T
where *U* is expected utility for the patch, *W* is the investor's current wealth, and *P* and *F* are defined above. The term *(W + T P)* estimates investor wealth at the end of the time horizon if no failures occur. The term *(1 - F)T* is the probability of not having a failure over the time horizon; it reduces utility more as failure risk increases. (Economists might expect to use a utility measure such as present value that includes a discount rate to reduce the value of future profit. We ignore discounting to keep this model simple.)
### Learning
There is no learning.
### Prediction
The utility measure estimates utility over a time horizon by using the explicit prediction that *P* and *F* will remain constant over the time horizon. This assumption is accurate here because the patches’ *P* and *F* values are static.
### Sensing
The investor agents are assumed to know the profit and risk at their own patch and the adjacent neighbor patches, without error.
### Interaction
The investors interact with each other only indirectly via competition for patches; an investor cannot take over a business (move into a patch) that is already occupied by another investor. Investors execute their repositioning action in randomized order, so there is no hierarchy in this competition: investors with higher wealth have no advantage over others in competing for locations.
### Stochasticity
The initial state of the model is stochastic: the values of *P* and *F* of each patch, and initial investor locations, are set randomly. Stochasticity is thus used to simulate an investment environment where alternatives are highly variable and risk is not correlated with profit. Whether each investor fails each year is also stochastic, a simple way to represent risk. The investor reposition action uses stochasticity only in the very unlikely event that more than one potential destination patch offers the same highest utility; when there is such a tie the agent randomly chooses one of the tied patches to move to.
### Collectives
There are no collectives.
### Observation
The World display shows the location of each agent on the investment landscape. Graphs show the mean profit and risk experienced by investors, and mean investor wealth over time. An output file reports the state of each investor at each time step.
## Initialization
Four model parameters are used to initialize the investment landscape. These define the minimum and maximum values of *P* (1000 and 10,000) and *F* (0.01 and 0.1). The values of *P* and *F* for each patch are drawn randomly from uniform real number distributions with these minimum and maximum values.
One hundred investor agents are initialized and put in random patches, but investors cannot be placed in a patch already occupied by another investor. Their wealth state variable *W* is initialized to zero.
## Input data
No time-series inputs are used.
## Submodels
* *Investor repositioning.* An agent identifies all the businesses that it could invest in; the neighboring (eight, or fewer if on the edge of the space) patches that are unoccupied, plus its current patch. The agent then determines which of these alternatives provides the highest value of the utility function, and moves (or stays) there.
* *Accounting.* This action is fully described above ("Process overview and scheduling").]]>setup repeat 75 [ go ]setupgomean [wealth] of turtlesstandard-deviation [wealth] of turtlessetupgomean [wealth] of turtlesstandard-deviation [wealth] of turtlessetupgop-hplrp-hphrp-lplrp-lphrp-same