chapter 7 implementing the dynamic gtap model in the
TRANSCRIPT
1
Chapter 7
Implementing the Dynamic GTAP model in the RunDynam Software
Ken Itakura, Elena Ianchovichina, Csilla Lakatos and Terrie Walmsley1
1. Introduction
The purpose of this chapter is to introduce the reader to the publicly available software –
RunDynam – which is used to carry out the applications presented in Part III of this book. The
RunDynam program is based on the GEMPACK suite of software (Harrison and Pearson, 1999),
which is specifically designed to solve non-linear general equilibrium models. Other general
equilibrium models solved using the GEMPACK software suite include the standard GTAP model
and the Monash model of Australia. The RunDynam program has been specially tailored to the
needs of the GDyn model and other dynamic models. It offer the user a great deal of flexibility in
constructing simulations. It is available from the Centre of Policy Studies at Monash University,
Australia.2
Using the RunDynam software, you can examine the data, construct and modify
simulations, solve simulations, and examine results. Users who wish to alter the underlying theory
of the model will need to acquire additional software from the Centre of Policy Studies at Monash
University, Australia. Also for those who wish to make their own aggregations they will have to
purchase the GTAP Data Base from the Center for Global Trade Analysis, Purdue University,
USA3. Altering the standard model and/or the data aggregation are not discussed in this chapter.
The RunDynam program requires a PC running Microsoft Windows XP or later, with at
least 512MB of RAM and at least 1GB of free disk space;
1 The authors would like to thank Anna Strutt and Horacio Santander for their help with the welfare decomposition section of this chapter. 2. http://www.gempack.com.au 3. http://www.gtap.agecon.purdue.edu/
2
The chapter is organized as follows. Section 2 demonstrates how to install the program and
load up the applications. Section 3 shows how to view the data. Section 4 provides an introduction
to running simulations. Section 5 examines how to view simulation results. Section 6 outlines the
notation used to denote the chapters presenting the applications in Part III of this book.
2. Installing the RunDynam Software and loading up applications
Installing RunDynam
The following steps allow us to install RunDynam:
Double click on the install EXE in the RunDynam CD-ROM.
RunDynam Installation program starts and a Welcome pop up box appears on your screen.
Click Next.
You will be asked which directory you wish to install RunDynam. Default destination
folder is c:\RunDynam. Click Next.
If you are ready to install, then click Next.
Once installation is completed, click Finish.
The RunDynam icon will then appear on your desktop ready for use.
Downloading the RunDynam application archive
Once you have installed the RunDynam software you can obtain the applications for this
Chapter and the other Chapters in this book, from the GTAP web site
(https://www.gtap.agecon.purdue.edu/models/Dynamic/applications.asp). Each application is a
RunDynam zip file. Download the zip files from the web site and place into the following
RunDynam sub-directory c:\RunDynam\archive.
Opening RunDynam
The first step is to open up RunDynam. This is achieved by double clicking on the
RunDynam icon on your desk top. If a dialogue box appears to ask if you would like to load a
model and simulation from zip archive: click No.
T
O
toolbar (F
have seen
menu allo
help. The
Other file
contains
Model/D
Restoring
A
the GDyn
ingredien
using the
O
A
In
A
d
O
A
The following
Once inside R
File, Zip, Ta
n in other W
ows you to c
e second too
es) looks lik
an essential
Data page). E
g the Ingred
At this stage,
n model to s
nts of this si
e following i
On the main
A drop down
ngredients f
A “ZIP File to
irectory whe
OPEN button
An informatio
g should app
RunDynam y
asks, View,
Windows pro
carry out var
olbar (Introd
ke a notebook
element of t
Each label is
dients of a S
we will use a
imulate the e
imulation ha
nstructions:
n menu firs
menu will ap
from ZIP A
o Restore Fr
ere you dow
n.
on box appe
pear at the to
you will noti
Options, Ru
ograms. This
rious functio
duction, Mod
k or card ind
the simulatio
referred to a
Simulation
a simulation
effects of a p
ave been sav
st choose:
ppear. Move
Archive and t
rom” box wi
wnloaded the
ars and show
3
op of your sc
ce two sets o
un Preference
s menu is re
ns such as op
del/Data, Sim
dex. Each of
on design. E
as a tab. You
n which was c
productivity
ved to a zip
Zip | Rest
e down in the
then click.
ill appear. Ch
em: Ch7HO
ws history of
creen:
of toolbars at
es and Help
eferred to as
pening, savi
m Overview
f the pages i
Each card is r
u click on a t
created prev
shock on th
file and can
tore Ingred
e drop down
hoose the fo
O3x3_gdyn_
f the ZIP arc
t the top of t
p) is similar t
s the main m
ing, viewing
, Closure/Sh
in the notebo
referred to a
tab to go to t
viously for Ru
he Rest of W
n be restored
dients from
n menu until
ollowing file
_v3_97.zip,
chive. Click
he page. The
to those you
menu. This
g data and see
hock, Result
ook or card i
as a page (i.e
the required
unDynam. It
World (ROW)
d into RunDy
m ZIP Arc
you find Re
e from the ar
and click o
OK.
e first
u may
main
eking
ts and
index
e., the
page.
t uses
). The
ynam
chive.
estore
chive
on the
A
th
C
sp
A
A
D
T
C
N
C:\RunD
C:\RunD
C
L
Another infor
he files in th
Choose the fo
pecified doe
A confirmatio
After the file
Details into R
The software
Click OK.
Note that
Dynam\HO3
Dynam\HO3
Click the tab
Listed at the t
rmation box
e ZIP archiv
ollowing dir
s not exist, t
on box appea
es are unzipp
RunDynam n
e reminds yo
under
3x3\data,
3x3\tabetc, h
labeled Mod
top in blue s
appears and
ve. Click OK
rectory: C:\R
the software
ars. Click O
ped, you wi
now?” Click
ou that the r
the direc
have automa
del/Data to
should be the
4
d asks you to
K.
RunDynam\
will create i
OK.
ll be asked
k Yes.
restored app
ctory you
C:\RunDy
atically been
turn to the M
e model. Che
o specify the
\HO3x3. Cl
it for you. C
“Do you w
plication is n
u specified
ynam\HO3x
n created to s
Model/Data p
eck that the
directory yo
lick OK. If t
Click Yes.
want to load
now the cur
d three
x3\model
store the file
page.
model is gdy
ou want to re
the directory
these Simul
rrent applica
new fol
es.
yn.exe.
estore
y you
lation
ation.
lders,
and
A
in
N
data inpu
3. V
In
Two exam
we first l
the data f
file (gdyn
A
(extensio
files con
respectiv
and some
Also make su
ngredients sh
Notice again
ut files are st
Viewing the
n this section
mples are un
ook at how t
files themsel
n.tab). The T
As we saw ab
on *.HAR) o
ntaining the
vely. The hea
e additional
ure you hav
hould be as f
that the mod
tored in the d
Data
n, we use Ru
ndertaken to
the data is us
lves. The und
TABLO file
bove, four da
one containin
parameters
ader array fil
data require
ve all the dat
follows:
del (gdyn.ex
directory - d
unDynam to
examine the
sed by the un
derlying equ
is the human
ata files are
ng the sets a
s for the st
le containing
d for the dyn
5
ta ingredien
xe) is stored
data.
look at the G
e sets data an
nderlying mo
uations of the
n readable v
required by
and the othe
andard GTA
g the base da
namic exten
nts in the wh
in the direc
GDyn model
nd the core da
odel or TAB
e GDyn mod
version of the
the GDyn m
r containing
AP model a
ata includes
sion.
hite box in m
ctory - mode
l and the asso
ata. In each o
BLO file and
del are define
e executable
model: four h
g the base da
and the dyn
both the stan
middle. The
el and the va
ociated data
of these exam
then we exa
ed in the TA
e file gdyn.ex
header array
ata and othe
namic exten
ndard GTAP
e data
arious
files.
mples,
amine
ABLO
xe.
y files
r two
nsion,
P data
Example
T
m
T
coefficien
gdyn.tab
1
2
3
Y
equations
In
F
File
e 3.1: Viewin
The TABLO
menu. TABm
The TABLO
nts and equa
, there are a
. All variab
must first
. All coeffi
have been
. In order to
these com
You will need
s, variables a
n TABmate
File. You sho
ng the Set D
code can be
mate opens a
file contain
ations. In ord
number of r
bles, coeffic
be defined i
icients must
n previously
o assist the u
mments are a
d to refer to
and coefficie
(or other ed
ould find the
Data.
e viewed by
copy of the
ns the essen
der for the s
rules which m
cients, sets, p
in the TABL
either be rea
defined and
user of the m
always placed
the TABLO
ents or when
itor), select
following s
6
choosing V
gdyn.tab file
ntial element
oftware to u
must be adhe
parameters,
LO file.
ad from a fil
d read;
model comme
d between tw
O file constan
n interpreting
Search | Fin
tatement (F3
View | Main
e, and the co
ts of the mo
understand th
ered to. Som
and files re
le or derived
ents have be
wo exclamat
ntly when se
g your result
nd from the
3 will repeat
n TABLO fi
opy may be l
odel includi
he model eq
me of these in
eferred to in
d from other
een added to
tion marks.
eeking any in
ts.
e main menu
t the search)
ile from the
labeled tab1
ng the varia
quations set o
nclude:
the TABLO
r coefficient
the TABLO
nformation a
u to find the
):
main
1.tab.
ables,
out in
O file
s that
O file;
about
word
7
GTAPSETS # file with set specification #;
This statement defines a file with logical name GTAPSETS. The words between the # #’s
are the label given to this logical file name. The name indicates to the user what is contained in the
file i.e., set information.
If you now use Search | Find to find Set, you should find a list of set statements, such as:
Set
REG # regions in the model #
maximum size 10 read elements from file GTAPSETS header "H1";
The second line of each set declaration defines the set and gives it a name. The third line
states that the set should contain no more than 10 elements and is read from header H1 of the file
with logical filename GTAPSETS.
If you now use Search | Find to find REG you will see that many of the variables,
coefficients and equations are defined over the set REG. You will also see variables
defined in terms of the other sets (e.g. TRAD_COMM).
You can now close the TABmate by clicking File | Exit.
Back in RunDynam, you should still be on the page labeled Model/Data. If you look at the
white box containing the data files you will see the name of the file corresponding to the
logical name GTAPSETS.
File GTAPSETS = C:\RunDynam\HO3x3\data\gdset.har
This tells us that gdset.har is the file, with logical name GTAPSETS, containing the set
information.
To open this file highlight the GTAPSETS file in the white box and click the right hand
button of your mouse to get a menu. Then select View this file.
O
table corr
In
third to th
in the se
length 1
character
(REG) sh
seen betw
Y
C
o
Once in View
responds to
n the table of
he size of th
cond column
12 in the thi
rs. As we kn
hould be loca
ween the ##’
You can view
Click on Con
f sets.
wHAR you w
a set.
f sets the firs
e data and th
n means tha
ird column
now from th
ated under h
’s in the TAB
w the regions
ntents on the
will see a men
st column re
he last colum
at the inform
means that
he TABLO c
header H1. T
BLO code.
s by double
e main menu
8
nu bar and a
elates to the h
mn is a brief
mation conta
there are 3
code (tab1.t
The commen
clicking any
u bar or click
table contai
header, the s
f explanation
ained in the
elements w
tab, copy of
nt in the last
ywhere on th
k anywhere o
ining the sets
second to the
n of what is i
header are c
which may c
f gdyn.tab) t
column is th
he row label
on the data to
s. Each row i
e type of dat
in the set. Th
characters. T
contain up t
the set of re
he same com
ed header H
o return to th
in the
a, the
he 1C
The 3
to 12
gions
mment
H1.
he list
9
Now look at the traded commodities (TRAD_COMM) in this data set.
You can now close the header array file by clicking File | Exit.
Example 3.2: Viewing the Core data
Before examining the data let us look again at the TABLO file (tab1.tab, a copy of
gdyn.tab).
Choose View | Main TABLO file from the main menu. The core data can be found in the
file with logical file name GTAPDATA.
Using Search | Find in TABmate, search for GTAPDATA. You should find the following
statement.
File
GTAPDATA # file containing all base data #;
This statement defines a Header Array file with logical file name GTAPDATA. Now let us
look at an equation containing some data.
In TABmate, select Search | Find and look for TOTINCEQY. This will take you to the
following equation labeled TOTINCEQY:
Equation TOTINCEQY
# This equation determines the change in total income from equity#
(all,r,REG)
yqh(r)
= [YQHFIRM(r) / YQHHLD(r)] * yqhf(r)
+ [YQHTRUST(r) / YQHHLD(r)] * yqht(r);
The first line is the equation name. The second line is a comment providing some
information on what the equation does. The third line states that this equation holds for each region
r in the set REG (i.e., USA, EU and ROW as seen in Example 1). The rest of the lines define the
equation.
In standard GTAP notation, the convention is to use upper case for levels and lower case
10
for percentage changes or deviations from base case. Here, we are interested in the initial database
or the levels coefficients (shown in upper case). We can find out what these coefficients are and
how they are calculated by searching the TABLO file. Since all variables and coefficients must be
defined prior to use we should be able to find a definition by searching upwards.
Select Search | Find from the main menu of the TABmate and look for YQHHLD.
Remember to change the find box to search upwards. This is done by clicking on Back in the
box labeled Direction (alternatively you can use CTL-Home to move you back to the
beginning of the TABLO file, search from top or gloss). Keep searching (F3 to search again)
until you find the following definition and formula for YQHHLD(r):
Coefficient (all, r, REG)
YQHHLD(r) # regional household equity income #;
Formula (all, r, REG)
YQHHLD(r) = YQHFIRM(r) + YQHTRUST(r);
This tells us that YQHHLD(r) is defined as a coefficient and is the income earned on equity
by the regional household. It is equal to the sum of two other coefficients YQHFIRM(r) and
YQHTRUST(r). This coefficient is often referred to as a derived coefficient as it is not read directly
from the database but is derived from other coefficients.
You can now use Search | Find to find YQHFIRM(r) and YQHTRUST(r). You should find
the following statements in the TABLO file:
Coefficient (ge 0)(all,r,REG)
YQHFIRM(r) # income of region r from local firms #;
Update (all,r,REG)
YQHFIRM(r) = yqhf(r);
Read
YQHFIRM from file GTAPDATA header "YQHF";
Coefficient (ge 0)(all,r,REG)
YQHTRUST(r) # regional income from global trust #;
Update (all,r,REG)
Y
Read
Y
T
as inco
YQHH
respect
If y
als
Yo
Ba
containin
GTAPDA
To
mo
You sh
Lo
can
eac
T
YQHTRUST(
YQHTRUST
These stateme
ome earned
HLD(r) these
tively, in the
you scroll up
o read from
ou can now c
ck in RunDy
ng the data fi
ATA.
view this f
ouse to get a
hould see the
ok for heade
n view the d
ch of the thre
The two param
(r) = yqh
from fil
ents tell us th
by region r
e coefficien
e file with lo
p or down fro
the file GTA
close the TA
ynam you sh
iles you shou
file highlight
menu. Then
e following t
er YQHF con
ata by doub
ee regions. R
meter files c
ht(r);
le GTAPDA
hat YQHFIR
from local
nts are read
ogical file na
om here you
APDATA.
ABLO file by
hould still be
uld see the na
t the GTAP
n select View
table contain
ntaining the v
ble clicking o
Repeat the sa
can be viewe
11
ATA heade
RM(r) and YQ
firms and fr
d directly f
ame GTAPD
u will notice
y clicking Fil
e on the pag
ame of the fi
PDATA file
w this file.
ning the base
values of equ
on the heade
ame for head
ed in the sam
er "YQHT"
QHTRUST(r)
rom the glob
from heade
DATA.
a number of
le | Exit.
ge labeled M
le correspon
and click th
e data:
uity income
er “YQHF”.
der “YQHT”
me way.
T";
) are also co
bal trust resp
ers “YQHF”
f other coeff
Model/Data.
nding to the l
he right han
earned from
Data should
”.
efficients de
pectively. U
” and “YQ
ficients whic
In the white
logical file n
nd button of
m local firms
d be availab
efined
Unlike
QHT”,
ch are
e box
amed
f your
. You
le for
12
4. Running a Simulation
The purpose of this section is for you to gain some hands-on experience at simulating with
RunDynam. An increase in total factor productivity in the Rest of World is used to illustrate this.
For this purpose we examine the remaining pages of RunDynam. The next two pages, Sim
Overview and Closure/Shocks, relate to the elements required to undertake a simulation with the
GDyn model. The third page, Results, provides an easy way to view the results of the simulation.
The final page, Other Files, lists the data files used in the simulation and informs you as to
whether they are updated during a typical simulation.
First, we provide a recipe outlining the basic ingredients required to conduct a simulation
using RunDynam. The simulation undertaken here is a productivity shock. Two of the main
elements of the simulation have already been discussed in the previous section, these were the
Model (gdyn.tab) and the Data. Therefore, the focus here is on the two pages labeled: Sim
Overview and Closure/Shocks. We examine the Sim Overview page, the Closure/Shocks page,
a base case shock file, a policy shock file and a closure file. Finally, we run the simulation.
Simulation Overview Page
Click on the Sim Overview tab on the second level of the toolbar to move to the Sim
Overview page.
This page contains a number of important aspects about the simulation, including the label
for the starting year, the number of periods being examined, the length of these periods and the
solution method. The page should look something like the figure provided below.
Startin
Th
ma
col
In t
to d
Ch
Fol
per
5 y
Fol
thr
you
the
sim
dif
ng from the t
e box labele
ay assume thi
llected. In th
the next box
determine ho
heck that this
llowing this
riod may dif
years.
llowing this
ee simulatio
u are require
e names BAS
mulations res
fferently to a
op and movi
ed Start from
is is the curr
his case we s
x labeled Nu
ow many per
s figure is co
is a drop d
ffer from 1 y
is a list labe
ons are under
ed to give sim
S, BRR and
spectively. L
avoid writing
ing down:
m data for y
rent year or it
tart at year 1
umber of Pe
riods you wa
orrect.
own menu f
year. In this c
eled Simula
rtaken, the b
mulation nam
d HO3 for th
Later when y
g over previo
13
year (4 digit
t may be the
1997. Check
riods for ba
ant to examin
for the lengt
case, the leng
tion and Sim
base case, the
mes to all thr
he Base Ca
you do more
ous simulatio
ts) is the firs
year corresp
k that this say
ase case (1 o
ne. In this ca
th of each p
gth of each p
m names (3
e base re-run
ree of these e
ase [B], Bas
simulations
on results.
st year of the
ponding to w
ys 1997.
or 2 digits),
se we are loo
eriod. In the
period, exce
3 Chars). In
n and the pol
experiments
e Rerun [R
, you may w
e simulation
when the data
, you are req
oking at 5 pe
e GDyn mod
ept the last o
the GDyn m
licy. At this
s. In this case
R] and Polic
want to label
. You
a was
quired
eriods.
del, a
ne, is
model
stage
e, use
y [P]
these
14
The three simulations are discussed below:
1. The Base Case simulation represents how we might expect the economy to look
without the policy shock (i.e., without the productivity shock). Depending on what
we know about the future state of the world economy, this could include our beliefs
about population and labor growth rates, or the state of tariff reductions etc. over the
simulation period. How to develop a baseline was discussed in Chapter 5.
2. The Base Rerun takes the policy closures and the base case shocks. It is a calibration
simulation. If there are any variables which were endogenous in the base case
closure and are exogenous in the policy closure the program automatically takes the
values of these variables from the base case simulation, and includes them as
exogenous shocks to variables in the base rerun simulation. The usefulness of the
base rerun simulation is that it allows you to reverse any calibration done in the
baseline. For instance in Chapter 5 we exogenized real GDP (qgdp) to ascertain
changes in region-wide technological change (afereg) in order to obtain that change
in the real GDP. In the base rerun simulation the values for the region-wide
technology obtained in the base case are now applied in the base re-run to afereg to
endogenizely determine real GDP. If the base re-run has worked correctly, the
resulting changes in real GDP in the base re-run should equal those applied in the
base case, subject to small differences due to path dependency. The fact that real
GDP is now endogenous means that in the policy simulation, real GDP can now
respond endogenously to the policy shock.
3. The Policy simulation examines the effects of the policy shock, which is applied in
addition to the other base case shocks. The policy shock then interacts with the other
changes expected to occur in the world economy. Both simulations are undertaken so
that the difference between the two scenarios can be calculated, and hence the effects
of the policy shock isolated.
The next step requires us to specify the current working directory: C:\RunDynam\HO3x3\.
Check that this is in fact the working directory. If not you can change this by clicking on the
cha
Ne
of v
the
Fin
ext
to t
but
che
Closure/S
Now we
Cli
sim
Th
that the l
ange button.
ext you are a
variables you
e mapping fil
nally, you ar
trapolation.
the solution
tton. In the b
eck the optio
/Shock Page
can proceed
ick on the tab
milar to the o
e table lists t
labels in the
asked to spec
u would like
le by clickin
e asked to sp
. You must a
method. If n
bottom left h
ons (keep the
e
d to the closu
b labeled Clo
one provided
the shock an
first column
cify a mappi
e to place into
ng on the but
pecify a solu
also include
not, automati
hand corner
e default opt
ure and shoc
osure/Shock
d below:
nd closure fil
n of the tabl
15
ng file: GDy
o a spreadsh
tton labeled
ution method
automatic ac
ic accuracy c
click on th
tions). Then
ks file.
k to move to
les for each
le relate to th
yn.map. Th
heet file for fu
Edit.
d. In this cas
ccuracy. It sh
can be chose
he box label
click OK.
o the next pag
period of th
he informati
he mapping f
further analys
se we use Gr
hould appea
en by clickin
led automat
ge. This pag
e simulation
ion provided
file contains
sis. You can
ragg: 2-4-6
ar in brackets
ng on the Ch
tic accuracy
ge contains a
n. You will n
d on the prev
a list
view
steps
s next
hange
y and
table
notice
vious
page: Sim
periods. F
2002.
Ba
shocks ar
policy sh
If t
indicated
directory
(constitu
To
han
T
which ar
indicates
this case
skilled an
m Overview
For example
se case and P
re specified i
hocks).
these are se
d in black fo
y specified.
ent) file whi
view a base
nd button on
The shock fil
re imposed i
what we mi
there are fo
nd unskilled
w. The label
e, the first pe
Policy, shoc
in files (with
et up correc
font. If any o
If any of th
ich does not
e case shock
n the mouse a
e contains a
in the base
ight expect t
orecasts for
d labor growt
ls 2002, 200
eriod starts a
ks and closu
h extensions
ctly, all the
of them are
hem are in g
exist.
file place th
and select E
list of varia
case scenari
to happen in
the rate of
th (qfactsup
16
07, 2012 and
at 1997 and
ures, must be
.CLS for clo
files (with
in red then
green then t
he cursor on
dit.
ables and the
io. As ment
n the world e
real GDP g
), shocks to
d 2020 are t
continues fo
e specified fo
osure, .BSH
extension .
n these files
the files exi
the file you
e correspond
tioned previo
economy dur
growth (qgdp
import tarif
the last year
or 5 years to
or every peri
for base sho
CLS, .BSH
do not exis
ist, but they
u wish to ope
ding shocks t
ously the ba
ring the simu
p), populatio
ffs (tms) and
rs of each o
the beginni
iod. Closure
ocks and .PS
H, and .PSH
st in the wo
y refer to an
en, click the
to these vari
ase case sce
ulation perio
on growth (p
d export subs
of the
ing of
es and
H for
H) are
orking
nother
right
iables
enario
od. In
(pop),
sidies
(txs).4 Fi
first simu
changes o
4. If there are.
inally, time i
ulation is fro
over a 5 yea
are any variabl
is shocked b
om the begin
ar period.
le names you ar
by 5 to indica
nning of 1997
re uncertain of,
17
ate that a 5 y
7 to the begi
, you can alway
year period i
inning of 200
ys go back to th
is being simu
02, and all th
he TABLO file
ulated, henc
he shocks re
to find out wha
e this
fer to
at they
18
In some cases the shocks are located in separate files (e.g. ENDW002.shk) which are
referred to hereafter as the base case shock file. These files can be viewed by selecting File | Open
from the main menu in TABmate. If one of these files were missing, Y97_02.BSH would appear
in green in the table on the Closure/Shocks page.
The policy shock files are given in the final column of the table on the Closure/Shocks page.
Again the policy shock file can be viewed by clicking on the file (AFEREG.PSH) with the
right click and selecting edit.
There are two differences between the policy shock file and the base case shock file:
1. Only the policy shocks in this case a single shock to productivity (afereg) are
included, even though in the policy simulation all shocks from both the base case
shock file and the policy shock file are imposed.
2. the term ashock is used to shock this variable as compared to the base case shock
file where the term ‘shock’ was used. The term ashock stands for additional shock.
Thus afereg(“row”) this variable is shocked in the base case and by an
additional -5% in the policy simulation in the period 2002 to 2007.
Finally the closure file (POL.CLS) can be viewed by clicking on the file with the right click
and selecting edit.
The closure file is used to set out which of the variables are exogenous or fixed and which
of the variables are determined endogenously within the model. To close the model, the number
of endogenous variables must equal the number of equations; otherwise the model will not solve.
In the closure file, there is a list of exogenous variables, followed by the statement “Rest
Endogenous”. The file POL.CLS specifies the standard Gdyn closure.
A comparison of the policy and base case closures will reveal that we are calibrating
region-wide technological change to target forecasted real GDP.
Running the Simulation
In o
to r
An
If s
Cli
Altering
A
simulatio
and time
shocks yo
the time h
of time w
shock an
need to a
it is easy
When yo
1
2
order to run
run the base
n information
successful a
ick OK. Not
the Simulat
As you beco
ons. We sugg
periods and
ou simply ed
horizon simp
will automati
d closure fil
alter the shoc
to forget.
ou start maki
. Rename t
simulatio
changing
. You shou
the simul
the simulatio
case, base r
n box will th
message wil
tice you have
tion
ome more c
gest making
d then eventu
dit the files w
ply edit the s
cally adjust t
es yourself.
ck to time in
ing your own
three letter
on overview
these extens
uld also edit t
lation overv
on begin by
rerun and the
hen appear te
ll appear:
e jumped on
confident wi
small chang
ually making
we have bee
simulation o
the closure/s
Note that if
n the shock fi
n simulation
extensions
w page. This
sions will sto
the simulatio
view page.
19
choosing Ta
e policy simu
elling you to
nto the Resul
ith the prog
es to current
g your own a
n examining
verview pag
/shock page,
you alter the
iles – the pro
ns you should
given to th
will stop y
ore the resul
on descriptio
This allows
asks | Run B
ulations con
o “Beginning
lts page.
gram you w
t application
aggregations
g. To alter th
ge. Any chan
however yo
e length of t
ogram does
d also:
he base, ba
ou from ove
lts under alte
on by clickin
s you to wri
Base, Base R
nsecutively.
g simulation(
will wish to
ns – changing
s. To alter th
he length of p
nges to the le
u will have t
the time peri
not do this a
ase rerun an
erwriting pre
ernative nam
ng on edit sim
ite a short d
Rerun and P
(s)”. Click O
make your
g closures, sh
he closures a
periods or ex
ength or exten
to alter/make
iods you wil
automatically
nd policy on
evious resul
mes.
m descriptio
description o
Policy
OK.
own
hocks,
and/or
xtend
nsion
e new
l also
y and
n the
lts, as
on on
of the
20
simulation which will help you to remember what this simulation was when you come
back to it at a later date.
3. It is important to save the ingredients of your simulations.
First you can save the details of the simulation by selecting File | Save simulation details
as. Then provide a name for the simulation. These details can then be reloaded at any time
by selecting File | Load simulation details and selecting the relevant file. The disadvantage
with this is that it only saves the simulations details5, not the ingredients6 themselves, thus if
you change an ingredient this will change the simulation.
An alternative method is to zip up the ingredients of the simulation. This is done by selecting
File | Save ingredients as Zip archive. We suggest you do this here. Save the ingredients
as Example1f.zip. The disadvantage/advantage of this method is that it does not save the
results. With all the ingredients and simulation details saved, the results are just a click away.
If you wish to keep the results you need to zip them up yourself.
5. Viewing the Results
Results are obtained for each period of the Base Case, Base Rerun and the Policy
simulations and can be viewed in a variety of ways. The Results page is divided into two parts.
The first section allows you to look at the results for all periods at once; while the second is for
viewing individual periods.
Viewing the Results for All Periods.
The first section looks something like this:
5 The simulation details refer to the setup in RunDynam and includes information recorded on the sim overview and closure pages. 6 The ingredients are the closure and shock files.
There are
1. R
re
T
co
2. T
a
A
S
3. T
p
v
ch
Example
Th
data. Vie
the result
Cli
wh
Ne
7 Note that
e three parts
Results for th
erun and the
The differenc
omparable to
The results fo
spreadsheet
AnalyseGE p
preadsheet
The third part
eriod-on-per
ariable occu
hange in tha
e 5.1: Viewi
e ViewSOL
ewSOL is mu
ts using View
ick on the D
hite circle ne
ext select the
t when using A
to this first
he base case
e policy can
ce between t
o the output
or some peri
t or within t
program. Si
box or the G
t relates to st
riod) or cum
urring during
at variable oc
ng the Resu
program is s
uch more ad
wSOL you m
Difference –
xt to the sele
style: year-
AnalyseGE you
section:
, the base re
be viewed b
the two simu
of a compar
od(s) or all p
the ViewSO
imilarly, yo
Graph/View
tyle. Results
mulative7. Ye
g that period
ccurring betw
ults Using V
similar to the
dept at viewi
must first sel
Policy v Re
ected option
on-year or c
u can only view
21
erun, the pol
by selecting
ulations show
rative static m
periods can
OL program.
ou can see
wSOL box.
can be displ
ear-on-year r
d, while cum
ween the init
ViewSOL.
e program V
ing the resul
lect the simu
erun option
n.
umulative re
w cumulative re
licy and the
one of these
ws the effec
model.
be displayed
Clicking th
the simulat
layed year-on
results show
mulative resu
tial period a
ViewHAR us
lts than the s
ulation you w
n. A small bl
esults. Selec
esults.
difference b
e options in
cts of the po
d in the Ana
he AnalyseG
tion results
n-year (or in
the percenta
ults show the
and the perio
sed previous
spreadsheet.
wish to view
lack dot sho
ct year-on-ye
between the
the first col
licy shock a
alyseGE prog
GE box start
by clicking
n the GDyn m
age change i
e total percen
od specified.
sly to examin
. In order to
w.
ould appear i
ear by clickin
e base
lumn.
and is
gram,
ts the
g the
model
in the
ntage
ne the
view
in the
ng on
Ye
Th
The choi
results w
available
No
You are n
On
fou
var
col
dif
sel
nee
To
| Sh
Ve
the
8. Note: If
ear-on-year
e choice of s
ice of simula
will be availab
e to view onc
ow click on t
now in View
nce in ViewS
ur columns.
riable or its d
lumn provid
fference resu
ections need
ed to select T
view the ye
how.. YonY
ery little appe
e top right h
f you opened th
in the Style
style at this s
ation is how
ble, howeve
ce in ViewSO
the box label
wSOL. The s
SOL you wil
The first on
dimensions,
des a short d
ults appear
d to be made
Time series.
ear-on-year r
Y | Year-on-y
ears to chang
and corner h
he base case sim
box.
stage is not im
ever importa
er if you sele
OL.
led Graph/V
screen will lo
ll see a menu
ne shows the
the third sh
description o
regardless
e once again
... | Show.. P
results for th
year
ge, except y
has changed
mulation in Ru
22
mportant as b
ant. If you s
ect either Pol
ViewSol.
ook somethi
u bar and a
e variable na
hows the num
of the variab
of the optio
inside View
Pert | Pertur
he policy sim
ou will notic
d to read “SE
unDynam this o
both are ava
elect Base C
licy or Diffe
ing like the f
table of resu
ame, the sec
mber of varia
ble. You wi
ons you se
wSOL. To vi
rbed Solutio
mulation you
ce that the p
EQ4 Pert Yo
option would n
ailable from w
Case then on
erence, all sim
following:
ults. The tab
cond display
ables in the
ll notice tha
lected in R
ew the polic
on8
u need to sele
phrase “SEQ
onY p.” Ch
not be available
within View
nly the Base
mulations w
ble is divided
ys the size o
list, and the
at the cumul
RunDynam.
cy simulation
ect Time ser
4 Diff Cum
heck that thi
e to you
wSOL.
Case
will be
d into
of the
e final
lative
Your
n you
ries...
d” in
is has
cha
Yo
and
Yo
Yo
afe
sce
Th
Cu
P.”
Yo
Example
C
well suite
model co
AnalyseG
A
and the p
9. If you ca
anged.9
ou can now v
d find qgdp
ou can now c
ou can also c
ereg. Remem
enario and in
e cumulativ
umulative. A
”
ou can now e
e 5.2: Viewi
Cumulative re
ed for analys
ode (tab file)
GE by select
AnalyseGE w
policy for the
annot see this c
view one of th
the variable
click on Con
heck that the
mber that th
n the policy s
ve results can
Again check t
exit from Vie
ng the Resu
esults can al
sis (see Pears
, the underly
ting differen
will load the
e final period
click on the sma
he variables
for percenta
ntents in the
e value of th
his shock w
scenario.
n also be vi
that the phra
ewSOL.
ults Using A
so be examin
son, Hertel a
ying data and
nces, all and
model tab f
d (2020) and
all box (next to
23
by double cl
age change i
main menu
he shock to a
ill depend o
iewed by se
ase in the top
AnalyseGE.
ned in Analy
and Horridge
d the solution
cumulative
file, the cum
the initial da
o the cross) in th
licking on th
in real GDP.
to move bac
afereg was as
on the shoc
electing Tim
p right hand c
yseGE. Anal
e, 2002). The
n file. Open
and then cli
mulative diffe
ata base; wh
he very top righ
he variable n
.
ck to the tab
s expected. F
ck imposed
me series...
corner reads
lyseGE is a p
e program br
the cumulat
ck on the An
ferences betw
hich in this ca
ht hand corner t
name. Move d
le of variabl
Find and clic
in the base
| Show.. Yo
“SEQ4 Pert
program wh
rings togethe
tive differenc
nalyseGE bu
ween the bas
ase is the up
to enlarge the s
down
les.
ck on
case
onY |
t Cum
hich is
er the
ces in
utton.
seline
dated
screen.
data for p
features i
Sea
On
Th
ins
mo
Yo
var
Cli
Th
period prece
in AnalyseG
arch for rorg
nce you have
e line numb
stances wher
ove to that lin
ou can now v
riable (left cl
icking on “E
e decompos
eding the fin
GE you can a
ge, the expec
e found rorge
bers are give
re the variab
ne and the eq
view the valu
lick) and the
valuate (sele
se options on
nal period (i
analyze the r
cted rate of r
e, select the
en in red on
ble is on the
quation.
ue of rorge (
en using righ
ection or coe
n the other
24
i.e., 2017). U
esults for ea
return.
Gloss button
the left han
left hand sid
cumulative d
ht click to ca
eff/variable a
hand will a
Using the gl
ach equation
n to see all o
nd side and t
de of the equ
difference) b
all up the me
at cursor) wi
allow you to
loss and eva
. For examp
occurrences o
the highlight
uation. Click
by placing y
enu.
ll give you th
o decompose
aluate/decom
le:
of that varia
ted lines ind
k on line 17
your cursor o
he value of r
e the equatio
mpose
able.
dicate
754 to
on the
rorge.
on or
sel
Yo
·
On
Th
rorge
Int
lab
equ
par
ection. Righ
ou will be ask
nce selected v
is is a decom
e(r)
= - R
- 100
+ sro
elligent dec
bel (TempCo
uation are us
rt of the equ
ht click on ro
ked how to d
viewhar will
mposition of
RORGFLEX(
0.0 * LAM
orge(r);
omposition
oeff) to each
sed for label
ation - ROR
orge and sele
decompose,
l open up an
f the equation
(r) * [qk
MBRORGE(r
breaks the e
section of th
ling purpose
RGFLEX(r)
25
ect “decomp
select RHS,
nd you will o
n for rorge.
k(r) - 10
r) * ERRR
equation up
he equation.
es. For exam
) * [qk(r
pose all or on
intelligent a
obtain the fol
00.0 * KH
RORG(r) *
according t
The coeffici
mple 1 e1_RO
r) - 100.
ne side of thi
and first togg
llowing:
HAT(r) *
* time
to the bracke
ients from ea
ORGFLEX
.0 * KHAT
is equation”
gle position.
time]
ets and assig
ach section o
refers to the
T(r) * ti
.
.
gns a
of the
e first
ime]
26
and e1_LAMBRORGE refers to the second part 100.0 * LAMBRORGE(r) *
ERRRORG(r) * time
So this tells us that in NAM qk rises relative to KHAT. Further decomposition will show that
in this case qk is 1.66, since the cumulative difference in time and hence 100.0 *
KHAT(r) * time is zero.
While intelligent decomposition is usually best, this is not always the case. Try decomposing
complete by variable to see the other features of AnalyseGE.
Close AnalyseGE and return to the RunDynam results page.
Viewing the Results for Some years.
Example 5.3: Viewing some of the Results using AnalyseGE
In some cases you may not want the cumulative differences for all of the years.
On the results page of RunDynam if you select ‘some’ and then click on AnalyseGE,
RunDynam will ask for an rsl file. To obtain an RSL file you will need to cancel and go back
to RunDynam.
Next select Tasks | Run SS Jobs for selected Years
Ru
unDynam willl then ask yyou to select
27
the starting
and end yeaars, and the ssimulation.
Sel
pre
Ru
for
Viewing
T
as well a
you to e
second p
Example
To look a
Fir
cas
Sec
con
inc
the
by
Th
Fin
lect 2002 as
erequisites an
unDynam wil
r 2002 to 201
the Results f
This section a
as the update
examine the
art of the Re
e 5.4: Viewi
at the 2012 u
rst select wh
se, select Ba
cond, select
ntaining the
cluding any e
e GTAPDAT
clicking on
ird, select th
nally click on
s the initial y
nd click ok.
ll then run th
12 which you
for Individu
allows you to
ed data and l
GDYNView
esults page l
ng the Upda
updated data
hich of the si
ase.
the output
results for
errors which
TA containin
GTAPDAT
he label corre
n the view b
year and 20
he SS job. N
u can load u
ual Periods.
o view the ba
og files for t
w, GDYNV
looks like th
ated Data
abase for the
imulations y
type you wi
the variable
h may have o
ng the updat
TA.
esponding to
utton to view
28
012 as the en
Now when yo
up.
ase case and
the end of ea
Vol and Wel
his.
base case si
you would lik
ish to view.
es, the Log
occurred or A
ted database
o the period
w these resu
nd year, the
ou select An
d policy simu
ach period. M
lfare results
imulation yo
ke to view -
You can ch
file contain
Accuracy Su
. In this case
you are inte
ults.
en click ‘ok’
nalyseGE the
ulation result
Moreover th
for individ
ou must:
- Base Case
hoose from
ning a log o
ummary of
e, select the
erested in. In
’. Yes, chec
ere is an RSL
ts for each p
his section a
dual periods.
or Policy. In
the Solution
of the simul
the simulati
updated data
n this case 20
ck the
L file
period
llows
. The
n this
n file
lation
ion or
abase
012.
29
The ViewHAR program for viewing header array files is automatically opened. This
should look familiar since you have used the same tool to look at the core database.
Exit from ViewHAR by selecting File | Exit.
Release 9 of GEMPACK also makes it possible to undertake post-simulation processing.
Hence, in addition to the Solution file and GTAPDATA users now have access to a more detailed
updated data file through WELVIEW, GDYNView, TAXRATES and GDYNVol10.
When taking a closer look, you’ll be able to find that in fact GDYNView as well as
GDYNVol and TAXRATES are all placed at the end of the TABLO file inserted between the
following two statements:
PostSim (Begin);
PostSim (End);
Post-simulation processing thus makes it possible to carry out calculations with values of
variables and coefficients which depend on previous simulation results within a single tablo file.
GDYNView gathers certain parts of the global data, with important information about
certain macro variables and others regarding trade, transport and protection.
Follow the same steps to examine the GDYNView file and the GDYNVol file for selected
periods. These files are similar to those in RunGTAP (Pearson and Nin Pratt, 1999), with some
minor modifications to include useful data for the GDyn model, such as rates of return and foreign
income. The formulas for GDYNView, GDYNVol and WELVIEW are included in the standard
GDyn tab file as post-sim processing.
Example 5.5: Viewing the Welfare Decomposition
The welfare decomposition (WELVIEW) results are viewed in the same way as the
GTAPView and GTAPVol results. Note however that the simulation will give zero results if time
is shocked. Valid results will only appear in WELVIEW when time = 0; i.e., when you undertake
the comparative static simulation to create a valid welfare decomposition. Users wanting to
30
undertake the welfare decomposition for the dynamic simulation should re-run this simulation
(CH7HO3x3_gdyn_v35_97.zip) but set automatic accuracy at 90%11 and then refer to Chapter 6
for details on the special closure and shocks used in the baseline.
Below is a summary of the steps required to undertake welfare decomposition simulation:
Re-run the dynamic simulation (CH7HO3x3_gdyn_v35_97.zip) but set automatic accuracy
at 90%. Open the results in ViewSOL and leave open for later use.12
Model/Data Page: Change base data to the updated data file from the final year of the base
re-run (eg dat-basb-brrr-2020.har)
Sim Overview page: Change labels so as not to over-write your files
- Year and time period (you need just one year, start 2019 and length 1 year)
- Alter File names (WDB, WDR, WDP). Although you will not need all three it is so
easy to mistakenly over-write one of the simulations by clicking the wrong button.
Closure/Shock page: Change closure file by swapping the following (Remember to save
the new closure under a different name and load into the baseline):
- Time variables
swap srorge = rorge ;
swap SDKHAT = DKHAT ;
swap sqk = qk ;
swap swqh = wqh ;
- Numeraire, e.g.
10 GDYNVOL output is placed in the updated GTAPDATA file, there is no special GDYNVOL file produced. 11 Note higher automatic accuracy assists with ensuring accuracy of the comparative static welfare decomposition simulation. The application CH6HO3x3_gdyn_v35_97.zip and the welfare decomposition of that application Ch6HO3x3_gdyn_v35_97wd.zip can be downloaded from the GTAP website at: https://www.gtap.agecon.purdue.edu/models/Dynamic/applications.asp. 12 Closing the file will not delete it, however you will need to re-open is later and switching between results can be time consuming. An alternative way of loading up other sequences of results is to use time series | load sequence. See
31
swap psavewld = ps("food","ROW") ;
Closure/Shock page: Change shocks file (again remember to save and load into
RunDynam)
- Shock the following variables by the amount from the solution file for the cumulative
difference between the policy and the base re-run.
rorge
DKHAT
qk
wqh
- ps for one commodity and region e.g., ps("food","ROW")
- Shock all the policy variables (by the cumulative differences from the dynamic policy
simulation)
Run base
Open WELVIEW for the base year to view the welfare decomposition. (Note that if all the
welfare results are zero then there is a shock to time and you have made a mistake.)
This will give the (more-or-less) correct EV and decomposition. To check you can compare the
resulting database, from the comparative static welfare simulation, with the final database from the
dynamic simulation. There will be differences however they should be relatively small, less than
0.01% of the values for a simple simulation like this one. If differences are larger than this then try
increasing the accuracy of the dynamic and comparative simulations.
6. Book Applications
Several applications are provided with this book and with the RunDynam software. The
help for details.
32
file name containing the application is labeled Ch#_model_v#_##.zip, where Ch# refers the
chapter number in this book (e.g. Ch7 refers to the application used in Chapter 7), model refers to
the model variant used (e.g., GDyn, Savings or Farm variant etc), v# refers to the tab file version
used if relevant (e.g. v3 is the latest version of GDyn.tab at the time of writing) and ## refers to the
year of the initial GTAP 5 Data Base used (e.g. 97 is 1997 which corresponds to the GTAP 5 Data
Base, as it has a reference year of 1997). These files are available on the GTAP website at:
https://www.gtap.agecon.purdue.edu/models/Dynamic/applications.asp.
Chapter 6 Ch6HO3x3_gdyn_v35_97.zip13
C6HO3x3_gdyn_v35_97wd.zip
Chapter 7 Ch7HO3x3_gdyn_v35_97.zip
Chapter 8 Ch8_gdyn_r19_95.zip
Chapter 9 Ch9_gdyn_r19_95.zip
Chapter 10 Ch10_dfarm_95.zip
Chapter 14 Ch14_gdyns_34_97.zip
Appendix A Ch7HO7x7_gdyn_v31c_97.zip
7. References
Harrison, J. and K. Pearson (1998): “Getting started with GEMPACK: Hands-on
Examples” GEMPACK document No. 8, Centre of Policy Studies and Impact Project, Monash
University, Clayton, Vic, 3168, Australia.
Pearson, K., Hertel, T. and M. Horridge (2002): “AnalyseGE: Software Assisting
Modellers in the analysis of their results” Centre of Policy Studies and Impact Project, Monash
University, Clayton, Vic, 3168, Australia.
Pearson, K. and A. Nin Pratt (1999): “Hands-on Computing with RunGTAP and WinGEM
to Introduce GTAP and Gempack” Center for Global Trade Analysis, Purdue University, West
13 This is the same application as Ch7HO3x3_gdyn_v35_97.zip but with automatic accuracy of 90%.
33
Lafayette, IN, 47907-1145, USA. Prepared for the July 1999 Short Course at Purdue University.