Master Économie et Affaires InternationalesCours “Modèles de Simulation”Paris Dauphine –October 2012
Prof. Dr. Ramón MahíaApplied Economics Department
www.uam.es/ramon.mahia
SIMULATION MODELS IN ECONOMY SOME BASICS
SIMULATION MODELS:
SOME BASICS OUTLINE
Part I: WHAT DOES SIMULATION MEAN? And
WHY DO WE NEED SIMULATION MODELS?
Part II: EXAMPLES OF (OWN) REAL
SIMULATION MODELS
Part III: BASIC ELEMENTS, STAGES AND
ADVICES FOR BULDING UP A SIMULATION
MODEL
SIMULATION MODELS: SOME BASICS
WHAT DOES SIMULATION MEAN? And WHY
DO WE NEED SIMULATION MODELS?
PART I of III
SIMULATION MODELS:
SOME BASICS WHAT DOES SIMULATION MEAN?
• A simulation shows the expected working of a
system based on a model (simulation model).
Simulation means to “run”, to put in practice a
“simulation model”
• A “simulation model” is a technical tool that
help us to understand real complex
systems…in order to take or evaluate
decisions.
SIMULATION MODELS:
SOME BASICS WHAT DOES SIMULATION MEAN?
Using a simulation tool, we can experiment in
real systems:
To Understand how the system works: how “inputs”
become “outputs”
To Evaluate alternative decisions
….or to find out the best set of inputs (decision) for
achieving a particular result / goal = Optimization
SIMULATION MODELS:
SOME BASICS WHY DO WE NEED SIMULATION MODELS?
A real system use to be complex (not chaotic) : different “agents”
affecting lots of variables (elements) greatly interrelated in a way that
…
…even if we can understand (or model) every single relationship, it is
difficult to anticipate and figure out the joint result
Of course we can try to to anticipate the result of a given decision
just relying on experience, intuition or theoretical conceptions…
but IDEALLY …
.. to understand the system and/or evaluate decision’s outputs, we
would need IDEALLY to “try out”, to experiment with reality...
…But obviously, most of the times we CAN’T make real tries for
evaluating alternative decisions because it is simply impossible or
very risky and/or expensive.
SIMULATION MODELS:
SOME BASICS MORE ON SIMULATION DEFINITION
Simulations Vs. Optimization
There are not Simulation Vs Optimization models but different
ways of use models :
“what if” = Simulation is an open strategy that uses the links
between inputs and outputs without setting an objective a priori or
the conditions for an optimum solution.
“how to”= Optimization systems concentrates mainly on reaching a
well predefined objective given a set of restrictions.
That’s why we usually say that simulation models are “run” and
optimization models are “solved”.
Most of the times, simulation looks like a natural previous stage for
optimization….
SIMULATION MODELS:
SOME BASICS MORE ON SIMULATION DEFINITION
Example: Simulation Vs. Optimization: Replace a quota
regime by a “tariff only” system:
1.- OPTIMIZATION LIKE: Which is the tariff level
equivalent to an existing quota regime?
2.- SIMULATION LIKE: Different tariff levels help us to
evaluate different impacts on domestic producers (as a
basis to negotiate other EU compensations), foreign
producers, NON EU exporters, EU re-exporters, changes
on export prices, wholesale prices, consumer prices…..
SIMULATION MODELS: SOME BASICS
EXAMPLES OF REAL SIMULATION
MODELS
PART II of III
SIMULATION MODELS:
SOME BASICS 3 REAL EXAMPLES
Simulating the impact of migration on pension
system for 2007- 2025 (CES Project 2006-07): Very complex and simultaneous interrelations between migration,
native demographical trends, structural economics, short terms conditions, ..politics (show or draw picture)
Very dynamic exercise: outcomes in “t” affects “t+1”, “t+2”,… etc “k” variables x “t” periods = “k” x “t” inputs and/or outputs
Once again,… impossible to try out and impossible to risk a single forecast output .
Lack of a single theoretical framework to be applied Different qualitative issues (politics) to be considered: migration policy
design and application, future welfare state design ….. LINK to International Migration Jouurnal Review:
"An Estimation of the Economic Impact of Migrant Access on GDP: the case of the Madrid Region"
SIMULATION MODELS:
SOME BASICS 3 REAL EXAMPLES
Removal of EU import barriers and evaluation of
effects for third countries (exporters) (FEMISE – EC
projects 2003,2004,2005,2006): Econometric models help us to anticipate new trade flows (changes
in prices ►new import demand ► export flows) IO Models help us to evaluate chained sector impacts in third
countries (you will learn how) obtaining detailed VA (GDP) and employment impacts.
A complementary Computable General Equilibrium model (CGE) could help us to spread simulation through the whole economy of the third country.
Two links for examples: “An equilibrium model for Free Trade Area creation economic impacts estimation” "
A Euro-Mediterranean Agricultural Trade Agreement: Benefits for the South and Costs for the EU"
SIMULATION MODELS:
SOME BASICS 3 REAL EXAMPLES
A simulation of the economic impact of renewable
energy development in Morocco (2012) An evaluation of RES economic impact in Morocco 2010 -2040
We identify the renewable energy source (RES) demand scenarios for Morocco ► the needs of RES installed capacity according to those scenarios and ► the detailed FDI plans needed to achieve such installed capacity supply.
Then, using a dynamic variant input–output model, we simulate the macroeconomic impact of the foreign investment inflows needed to make available these Moroccan RES generation capacity plans in the medium and long term.
Alternatives of CSP, PV and WP are compared Link to “Energy Policy” article:
"A Simulation of the Economic Impact of Renewable Energy Development in Morocco".
SIMULATION MODELS: SOME BASICS
BASIC ELEMENTS, STAGES AND
ADVICES FOR BULDING UP A
SIMULATION MODEL
PART II of III
SIMULATION MODELS:
SOME BASICSBASIC ELEMENTS & STAGES FOR BIUILDING UP A SIMULATION MODEL
(i) Real system “draft”
(ii) Operative system “representation” (design)
(iii) Identification and specification of
“variables (Inputs – Outputs) and “links”
(simulation flow)
(iv) Modeling (Technical core)
(v) Interface (platform of use)
(vi) Results (use of the model)
SIMULATION MODELS:
SOME BASICSBASIC ELEMENTS & STAGES FOR BIUILDING UP A SIMULATION MODEL
(i) Real (whole) system to be analyzed: The complete
collection of elements and interactions to be analysed by means
of the simulation.
My advice: The largest part of the technical decisions regarding the
estimation, calibration, design of scenarios and interface rely on and
are conditioned by a good comprehension of the elements and
interrelations of the whole system to be analysed….so
You will need to STUDY IN DEPTH until you get a complete sketch
of the real framework of the whole system: different parts (sub-
systems) should be recognized, every element and every relevant
connection properly acknowledged even if your fundamental interest
is focused in just a single part.
SIMULATION MODELS:
SOME BASICSBASIC STAGES FOR BIUILDING A SIMULATION MODEL: ELEMENTS AND DECISIONS
SIMULATION MODELS:
SOME BASICSBASIC ELEMENTS & STAGES FOR BIUILDING UP A SIMULATION MODEL
(ii) System “representation”: Simplified and
limited version of the real system A good simulation model BALANCE the compromise between
realism and simplicity…
…Then, in a second stage, you SHOULD identify the “reduced”
representation of the system that best fit YOUR simulation aims:
leave out some complete parts, reduce elements of interest and drop
useless relationships (never forget, of course, those rejected
variables and links, in case you need them later on, and bear them
always in mind for a broad and wide range comprehension of the
final results).
SIMULATION MODELS:
SOME BASICSBASIC STAGES FOR BIUILDING A SIMULATION MODEL: ELEMENTS AND DECISIONS
SIMULATION MODELS:
SOME BASICSBASIC ELEMENTS & STAGES FOR BIUILDING UP A SIMULATION MODEL
(iii) Variables:
Inputs: (***) Stimulus Inputs (decision or critical): main variables to be
changed when simulating
Exogenous Inputs (out of model, usually fixed or very limited in
variation, frequently qualitative, ideally not critical,..)
Outputs:
Intermediate outputs (state and auxiliary variables, or estimated
parameters)
(***) Final outputs
SIMULATION MODELS:
SOME BASICSBASIC STAGES FOR BIUILDING A SIMULATION MODEL: ELEMENTS AND DECISIONS
SIMULATION MODELS:
SOME BASICSBASIC ELEMENTS & STAGES FOR BIUILDING UP A SIMULATION MODEL
(iii) Simulation flow structure: Structured
scheme that illustrate the connection between
different variables: cause – effect chains Simplify the flow along the cause – effect chains (reduce
dimensionality, look for a semi - linear design)
Rationalize chain flows: prioritize inputs and outputs, give them
hierarchical order, and then…
Divide the system in homogeneous parts for planning the work
across areas. Locate the links between the different areas and order
the stages, identifying the priorities, bottlenecks and crucial points.
…(cont)
SIMULATION MODELS:
SOME BASICSBASIC ELEMENTS & STAGES FOR BIUILDING UP A SIMULATION MODEL
(iii) Simulation flow structure: (cont.)..
Plan a preliminary time work modeling schedule
according to:
“In model” factors: the previous identification of lines,
crossing points and bottlenecks
“Out of model” factors: existing organization of areas,
the resources available, the difficulty of different
tasks..
SIMULATION MODELS:
SOME BASICSBASIC STAGES FOR BIUILDING A SIMULATION MODEL: ELEMENTS AND DECISIONS
SIMULATION MODELS:
SOME BASICS
BASIC STAGES FOR BIUILDING A SIMULATION MODEL: ELEMENTS AND DECISIONS
t t+1 t+2 t+3 t+4 t+5 t+6 t+7 t+8
SIMULATION MODELS:
SOME BASICSBASIC STAGES FOR BIUILDING A SIMULATION MODEL: ELEMENTS AND DECISIONS
(iv) Technical structure: Quantitative definition of
elements (variables) and links (equations)
between them including:
1.- Collection of data for every variable (element)
2.- Mathematical (for deterministic links) and/or
statistical models (for randomness)
3.- Mathematical and/or statistical algorithms to describe
and validate convergence and/or equilibrium of simulation
or optimization solutions.
SIMULATION MODELS:
SOME BASICSBASIC STAGES FOR BIUILDING A SIMULATION MODEL: ELEMENTS AND DECISIONS
NATIONAL PRODUCERS YIELDS
TARIFFS
IMPORT PRICES
IMPORT DEMANDDOMESTIC
GROWTH
ECONOMETRIC MODEL
DOMESTICDEMAND
SUBSIDIES
DOMESTICPRICES
ECONOMETRIC MODEL
IDENTITY
REST OF THE MODEL
SIMULATION MODELS:
SOME BASICSBASIC STAGES FOR BIUILDING A SIMULATION MODEL: ELEMENTS AND DECISIONS
(v) Technical Structure (ADVICES):
Concentrate on data (Carpenters say "Measure twice, cut
once“).
Carefully supervise your “raw material”: use homogeneous data,
ensure the future availability of them, choose the samples
carefully, be extremely scrupulous in the handling of data.
Use the data provided by the end user, agree with them if data
responds truthfully to “their” perception of reality.
Explore the analytical - mathematical – statistical
procedures that best adapt to the system and your aims. (Cont.)
SIMULATION MODELS:
SOME BASICSBASIC STAGES FOR BIUILDING A SIMULATION MODEL: ELEMENTS AND DECISIONS
(v) Technical Structure (ADVICES):
Try to adapt the analytical technique to the problem
and not the other way round (models MUST be useful
and suit the problem, not technically attractive or
handsome)
Let simplicity guide your decisions. Do not complicate
the technical models if doesn't lead to sound benefits
from the user perspective (“If your intention is to discover
the truth, do it with simplicity and lave the elegance for
the tailors“ A. Eisntein) (Cont.)
SIMULATION MODELS:
SOME BASICSBASIC STAGES FOR BIUILDING A SIMULATION MODEL: ELEMENTS AND DECISIONS
(v) Technical Structure (ADVICES):
Be cautious with stochastic components: If you can, try to avoid critical dependency on stochastic
estimations: if inferential statistics are used, not only the final,
BUT the INTERMEDIATE outcomes would vary in a confidence
interval so you should carefully check the “sensitivity” of the
WHOLE system to EVERY coefficient change
... Think “seriously” about if/how re-estimations will be addressed
in the future.
Try (never easy) to offer results in an confidence interval – way
(providing values and probabilities).
SIMULATION MODELS:
SOME BASICS BASIC ELEMENTS OF A SIMULATION MODEL
(vi Interface: Platform for using the model
Sometimes is not necessary (self use)
Call for software professionals (if you have lots of money)
Let simplicity guide the design of the interface: The
interface is wished for using the model, not for understanding
the model: The “model” COULD be COMPLEX, but the
interface MUST be FRIENDLY:
Prioritise the wishes of users in all the stages and take
their advices
Set different levels of use: Decision makers, medium level
technicians, high skilled technical experts, etc... “There is no
inept user, only badly designed systems”.
SIMULATION MODELS:
SOME BASICS BASIC ELEMENTS OF A SIMULATION MODEL
(vi) Interface: Platform for using the model
SIMULATION MODELS:
SOME BASICS BASIC ELEMENTS OF A SIMULATION MODEL
(vi) Interface: Platform for using the model
SIMULATION MODELS:
SOME BASICS BASIC ELEMENTS OF A SIMULATION MODEL
(vi) Using the model: (**) Scenario: a set of inputs and parameters considered for
a simulation exercise
When several inputs are taken, lots of potential variant
scenarios arises
For reducing dimensionality:
Try to identify tree-structures (if possible) identifying
hierarchical connections of different inputs
“Pode the tree”: Drop impossible, hardly probable, not
interesting and not different scenarios.
Order the final list, select baseline and alternatives
(Cont.)
SIMULATION MODELS:
SOME BASICS BASIC ELEMENTS OF A SIMULATION MODEL
(vi) Using the model: Give probabilities to different scenarios (use conditional
probabilities if a tree scheme is used)
Evaluate the output:
Offer a kind of result that jointly evaluates the probability of
the outcome and the magnitude of it
Once you get results for each given scenario, clearly
identify the sensitivity of results to changes in every inputs.
Identify (and don’t underestimate) qualitative issues (or
simply out of model facts) that could affect results.
SIMULATION MODELS:
SOME BASICS BASIC ELEMENTS OF A SIMULATION MODEL
INPUTS VALUESHost country demographics High fertility variant Medium fertility variant Low fertility variantHost country economic growth High growth Medium growth Poor growth CrisisImmigration restrictions None Medium HighTime interest Short term Medium term Long termTOTAL SCENARIOS 108
Time DemographicsEconomic growth Restrictions Scenario Prob.
Short term Medium Medium Medium 1 15% Poor High 2 85%Medium Term Medium Medium. Medium 3 50% Poor High 4 30% Crisis High 5 20%Long Term High High None 6 30% Medium Medium None 7 40% Low Poor Medium 8 15% Crisis Medium 9 10% High 10 5%
Possible combinations 108 Selected = 10 # 2,4,8 = Baselines