2. Agent
based
models
and
stock
flow
consistent
models:
a
coherent
alterna@ve?
Stephen
Kinsella
University
of
Limerick
3. • Funded
with
a
series
of
grants
from
the
Ins@tute
for
New
Economic
Thinking,
INET,
Rannis,
Sta@s@cs
Iceland,
and
Irish
Research
Council.
• Overarching
goal
is
to
build
a
workable
model
comparable
to
models
used
in
CBs/Govt
Departments
for
policy
evalua@on/
counterfactual
scenario
tes@ng.
4. Today
• Context.
• Stock
flow
consistent
methodology.
What
is
it?
• SFC+ABM:
Why
connect
SFC
models
to
agent
Based
Models?
• 2
Applica@ons
– Irish
INET
model
basics
• Irish
economic
situa@on
from
a
balance
sheet
perspec@ve
– SFC
ABM
(Kinsella
et
al,
EEJ,
2011)
• Plan
of
Further
Work
8. SFC:
Horrible
name,
good
idea.
• Tobin
(1982)
in
his
Nobel
Lecture
and
Godley
and
Lavoie
(2007),
illustrated
the
generality
of
these
concepts
by
se`ng
out
a
model
of
the
economy
based
on
a
flow-‐of-‐funds
matrix.
• Each
column
shows
a
sector’s
balance
sheet
(for
stocks)
or
sources
and
uses
of
funds
(flows).
• Meanwhile,
a
row
shows
the
stock
or
flow
of
an
asset
as
it
is
distributed
among
the
supplying
and
demanding
sectors.
• Approach
now
common
in
simula@ng
models,
but
macro-‐
econometric
applica@ons
are
scarce
because
of
the
consistency
of
the
data
mainly
from
balance
sheet
with
those
of
the
real
economy
(Na@onal
Accounts).
9. Stock
flow
consistent
models
• Morris
A.
Copeland
(1949)
is
the
father
of
the
Flow
of
Funds
accoun@ng.
(Federal
Reserve
Bureau
Z.1
Release).
• Copeland’s
idea
was
to
enlarge
the
social
accoun@ng
perspec@ve
-‐
up
to
that
moment
used
mainly
in
the
study
of
na@onal
income
-‐
to
the
study
of
money
flows.
• Essen@ally
trying
to
find
an
answer
to
fundamental
economic
ques@on:
‘when
total
purchases
of
our
na@onal
product
increase,
where
does
the
money
come
from
to
finance
them?
When
purchases
of
our
na@onal
product
decline,
what
becomes
of
the
money
that
is
not
spent?’
(Copeland,
1949,
p.
254)
10. Tobin
1982,
Nobel
Lecture
These
models
should
have
1.
Precision
regarding
@me.
2.
Tracking
of
stocks.
3.
Several
assets
and
rates
of
return.
4.
Modeling
of
financial
and
monetary
policy
opera@ons.
5.
Walras’s
Law
and
adding
up
constraints.
J.
Tobin.
Money
and
finance
in
the
macroeconomic
process.
Journal
of
Money,
Credit
and
Banking,
14(2):171–204,
1982.
11. Joan
Robinson
“Before
a
model
can
be
confronted
with
empirical
tests,
it
has
to
be
examined
for
internal
consistency
and
for
the
a
priori
plausibility
of
its
assump@ons”
-‐-‐-‐Joan
Robinson,
What
are
the
quesFons?
JEL
14(4)
1977,
pp.
1319-‐1320.
12. Godley
&
Lavoie
• Sectoral
models
• Set
up
balance
and
transac@on
matrices
• Build
a
model’s
equa@ons
from
the
balance
sheet
rela@ons
(Behavioural
and
Iden@ty
rela@ons)
• Solve
for
steady
state
• Shock
using
‘policy
experiments’
through
simula@on.
• Lem
open
the
ques@on
of
es@ma@ng
these
models.
• This
is
my
group’s
central
problem.
13. Evolu@on
of
stock
flow
models:
sectors
Godin
et
al,
2013
Stock
flow
consistent
modeling
through
the
ages,
Levy
Ins@tute
WP
745
17. Issues/Problems
to
solve
• Consistency/ • Es@ma@ng
SFC
Frequency/Bubble
models
is
very
issues/Transfer
hard,
especially
pricing
porpolio
balance
equa@ons.
Data
Es@ma@on
Applica@on
Equa@ons
• Want
this
to
be
as
• Constantly
policy-‐relevant
as
balancing
possible
completeness
off
against
complexity
18. Real
world
balance
sheet.
2011Q1 NFC FC G HH ROW
Balance sheet A L A L A L A L A L
G & SDRs 841 841
Deposits 34,461 358,423 17,907 122,776 183,280 1
Bonds 233 451,093 69,945 455 381,371 -1
Loans 84,852 602,826 46,207 184,912 286,855 0
Equities 150,940 557,115 17,539 46,261 644,255 0
ITR 3,511 208,755 125,895 79,349 0
Other 10,489 1,045 2,304 5,553 14,783 0
Wealth (A-L) -208,542 -70,578 -78,402 104,922 253,441 -841
Sum (A-L) 0 0 0 0 0 0
19. Simplified
2011Q1 NFC FC G HH ROW
Balance sheet A L A L A L A L A L
Deposits 34,461 358,423 17,907 122,776 183,280 1
Bonds 233 451,093 69,945 455 381,371 -1
Loans 84,852 602,826 46,207 184,912 286,855 0
Equities 150,940 557,115 17,539 46,261 644,255 0
Wealth (A-L) -201,564 138,381 -80,706 -15,420 159,309 0
Sum (A-L) 0 0 0 0 0 0
21. Simula@on
studies
• Kinsella
&
Khalil
2011
Debt
Defla@on
Traps
within
Small
Open
Economies
• Kinsella
&
Khalil
2011
Bad
Banks
Choking
Good
Banks:
Simula@ng
Balance
Sheet
Contagion
• Kinsella
&
Godin
2012
Leverage,
Liquidity
and
Crisis:
A
Simula@on
Study
22. Es@ma@on
Studies
• O’Shea
&
Kinsella
(2010)
Solu@on
and
Simula@on
of
Large
Stock
Flow
Consistent
Monetary
Produc@on
Models
Via
the
Gauss
Seidel
Algorithm
• Godin
et
al
(2012)
Method
to
Simultaneously
Determine
Stock,
Flow,
and
Parameter
Values
in
Large
Stock
Flow
Consistent
Models
• Work
in
progress
w/
Rudi
Von
Arim
(UTAH)
on
‘solving’
and
studying
SFC
matrices
numerically.
23. Agent
based
Studies
• Kinsella,
Greiff
&
Nell
Income
Distribu@on
in
a
Stock-‐Flow
Consistent
Model
with
Educa@on
and
Technological
Change
Eastern
Economic
Journal,
Vol.
37,
Issue
1,
pp.
134-‐149,
2011
• New
IRC
&
INET
grants
w/
Mauro
Gallega@
&
Joe
S@glitz
to
bring
ABM
approach
closer
to
SFC
&
Vice
versa.
24. Pure
Dynamic
Empirical
Es9mated
SFC
Pure
Sta9c
Simula9on
simula9on
Model
Simula9on
• Calibra@on
• Empirical
• No
balance
• Calibra@on
• Dynamic
calibra@on
sheets,
par@al
• Sta@c
parameters
∆
• Real
world
data
es@ma@on
parameters
period
by
• Dynamic
• No
balance
• No
empirical
period
parameters
sheets,
full
data
• No
Empirical
• Natural
macro
es@ma@on
• Coherent
data
ra@o
coherent
• Par@al
Balance
macro
ra@o
criteria,
eg.
• Coherent
• More
sheets,
full
Debt/GDP
macro
ra@o
constraints
in
es@ma@on.
criteria,
e.g.
calibra@on
• Full
balance
Debt/GDP
• Use
country
sheets,
full
balance
sheets.
es@ma@on.
25. A
word
on
closures.
Lance
Taylor
(1991:
41):
‘Formally,
prescribing
closure
boils
down
to
sta@ng
which
variables
are
endogenous
or
exogenous
in
an
equa@on
system
largely
based
upon
macroeconomic
accoun@ng
iden@@es,
and
figuring
out
how
they
influence
one
another
...
.
A
sense
of
ins@tu@ons
and
history
necessarily
enters
into
any
serious
discussion
of
macro
causality’
26. Adjustment
Processes.
The
adjustment
processes
within
the
model
towards
the
steady
state
will
be
based
on
simple
reac@on
func@ons
to
disequilibria.
27. Note
that
the
empirical
values
for
adjusted
GDP,
and
GNP,
are
not
directly
comparable
to
standard
SNA
95
defini@ons.
An
example
will
show
you
why.
34. SFC
ABM
Much
more
developed,
connec@ons
to
Sectoral
complexity/network
theory/etc
No
black
holes
Individual
rather
than
sectoral
Avoids
lots
of
neoclassical
modeling
Porpolio
es@ma@on/simula@on
v.
easy
problems
Focus
on
closures
Models
agent
interac@ons
more
naturally
Focus
on
empirical
regulari@es
eg
power
Needs
solu@on
methods
laws
36. SFC
model
with
interac@ng
agents
• 4
Sectors,
households,
firms,
banks,
government.
• Workers...
– search
for
work.
– work
for
a
wage
or
get
dole.
– spend
money
on
consump@on.
– spend
money
on
educa@on.
• Firms...
– hire
workers.
– pay
wages.
– receive
revenue
from
selling
output.
• Government:
collects
taxes
and
provides
dole.
• Banks
lend
out,
can
go
broke.
• Model
allows
for
changes
in
educa@on/employment/income/
wealth
37. Nice
features
• no
representa@ve
agent
• no
u@lity
func@on
• no
ra@onal
expecta@ons
• large
number
of
heterogeneous
agents
• individual
behavior
is
unpredictable
• individuals
follow
simple
rules
• indeterminacy
at
the
micro
level
(random
selec@on
from
a
given
distribu@on)
• SFC
Adding
up
constraints.
41. Cool
stuff:
Measuring
Mobility
• Via
G.S.
Fields
&
E.A.
Ok,
“Measuring
Movement
of
Income”,
Economica
(1999).
• Mb=1/N*∑
|log
m_{0}−log
m_{1}|
• Implies
Higher
savings
→
lower
mobility.
42.
43. Conclusion
&
Further
Work
• Promising
connec@ons/crossovers
• Benchmark
model
to
be
built,
an
INET
group
exists
for
this
now.
• Lots
of
unexplored
areas,
open
ques@ons,
low
hanging
and
high-‐hanging
fruit.
• Fun
@mes
ahead!