106 JANUARY2020|COMPUTER SHOPPER|ISSUE383
DIYSIMULATION
If you fancy abit of hands-on experience of simulation, you
don’t need asupercomputer or adegree in computer science.
Certainly it’s not hard to knock up abit of simple code,oryou
could even use Excel to solve aset of differentialequations, but
our suggestion is simpler still, and it won’t cost you apenny.
Insight Maker is apowerfulsimulation tool that runs in your
web browser.Itsupportstwo types of simulation, which it refers
to as systemdynamics and agent-based modelling. The firstof
these is the type of simulation we’ve been looking at in this
article,which involves solving differential equations. Having got
to grips with that, however,you might like to delve intothe
agent-based approach. So,head off toinsightmaker.comand sign
up forafree account.
To make things even easier,Insight Maker hides most of the
maths fromyou. So,while you’re effectivelyentering aseriesof
differential equations, the onscreen process is much more
user-friendly.For instance,inour example of the spread of an
infectious disease (see ‘A Simple Example’, page 103), instead of
entering the differential equations, you use agraphical interface.
Just place three boxes on the screen, representing healthy,
infected and immune,drawarrows between themto represent
the flowof people fromhealthy to infected and infected to
immune,and enter afewvalues representing the initial numbers
of people in each category and the flowrates between them.
Nowjust run the simulation to see an immediategraph of the
results against time.You’ll probably get this example up and
running very quickly,but if you want abit more guidance,take
alookatthe example titled Disease Dynamics (SD).
Before dismissing the weather as asomewhat
unusual case –chaotic because the atmosphere is
such acomplicated system–the stark truth is that
even very simple systems can be hampered by chaos.
Many of us learned about the physics of apendulum
at school, which might lead us to believe that
predicting the behaviour of adouble pendulum–
basically apendulumwith ahinge at its centre –
shouldn’t be that much more difficult. Ye teven this
simple systemis chaotic, so predicting its path after
just afewswings is pretty much impossible.
HIGHFINANCE
Nowwe come to an area of computer simulation that
vies with the weather forthe titleofleast trusted:
economic modelling. One of the biggest issues,
certainly as farasthose who model electronic
circuits or racing cars are concerned, can be
gleaned by taking alook at atypical equation.
The following equation forms part of the
economic model that’s jointly maintained by
the Office of Budget Responsibility and the
Treasury:CBIUD =-169.01*dlog(MSGVA(-1))
+0.49*CBIUD(-1) +0.23*CBIUD(-2) +14.94.
We don’t need to understand what all the
variables are,but those strange numbers –
169.01, 0.49,0.23 and 14.94 –stand out. With the
exception of mathematical constants like Pi, models
used in engineering tend not to contain odd numbers.
That’s because the relationships between the variables are
exact or,asscientists would put it, they’re based on first principles.
When we turn our attention to economics, however,we’re
dealing with asocial science,whichisquitedifferent froma
physical science such as physics. There are no first principles, so
the equations are empirical, which means that theyhavebeen
established by analysing historical trends.
The objections to empirical equations by some scientists might be
ideological, but they’re not without cause.Because arelationship has
held true in the past, there’s no guarantee that it will do so in the
future.This is especially true if times are so unusual that historical
data is in short supply,and if we bearinmind that compiling an
economic model, let alone interpreting the output, invariably involves
asubjectiveelement, which leads to the possibility of political bias.
There’s another aspect to economic
modelling that’s also quiteunlike themajority of
simulation exercises. If you publish aweather
forecast, it won’t change howthe weather actually
plays out. Similarly,ifyou simulateanelectronic
circuit and then try it out, the circuit will behave in
exactly the same wayasifyou hadn’t simulated it first.
With economic modelling, however,things aresometimes quite
different. Recession is nowavery real possibility,according to
predictions made,partially on the basis of economic modelling. But
consider the possible outcome of publishing that prediction. Fearing
adropinincome or worse,peoplemight choose to postpone that
holidayofalifetime or that newkitchen. Needless to say, if sufficient
people decide to playitsafeinthisway,thatwill itself have anegative
impact on the economy,and theend result could be amuchworse
recession than if that prediction hadn’t been made.
Even if the output of the model had been totally wrong, if enough
people believe it, then it would become aself-fulfilling prophecy.For
all the benefits computer simulation has brought us, gazing intothat
digital crystal ball doesn’t always have the results we wanted.
ABOVE:The public perception of bankers and politicians
has led to distrust of economic modelling, but the
nature of themodel shares some of theblame
LEFT:It maybeanimportant element of the
Bank of England’s policy making, but economic
modelling has its issues
ABOVE:Yo uprobably won’t start with amodel this complicated, but this
example shows something of thepotential of Insight Maker
