You should read about the current versions
of the models before reading the following.
The dates and references for the MC2 model are:
MC2 Model
The order of the reading for the MC2 model should be:
- Macroeconometric Modeling (MC2 Version)
: at least skim this.
- About MC2 User Datasets.
- The MC2 Model Workbook
: Preface and Chapter 1, particularly Section 1.4.
The dates and references for the MC1 model are:
MC1 Model
Note:
Although the version of the US model used in the MC1 model has the same date
(May 9, 1997) as the US model
forecast (BASE971), the version used for the forecast is slightly different
from the version that is part of the MC1 model.
The price and wage equations, equations 10 and 16, in the
version used for the forecast differ slightly from those used in the
MC1 model.
The order of the reading for the MC1 model should be:
- Testing Macroeconometric Models (MC1
Version)
: at least skim this.
- About MC1 User Datasets.
- The MC1 Model Workbook
: Preface and Chapter 1, particularly Section 1.4.
Earlier versions of the US model are:
US Model (November 1, 1996)
Note: Appendix B and the Chapter 6 tables, which pertain to the ROW model,
are the same in both
Testing Macroeconometric Models (MC1 Version)
and The MC1 Model Workbook, but
Appendix A and the Chapter 5 tables, which pertain to the US model,
are not. The
US model dated November 1, 1996, is the one discussed in
Testing Macroeconometric Models (MC1 Version),
whereas the US model dated May 9, 1997, is the one used for the MC1 model.
US Model (September 5, 1993)
- Equations: Appendix A and the tables in Chapter 5 in the
1994 book.
- Discussion: The 1994 book.
Note:
The version of the US model that is presented in Chapter 5 and Appendix A
of the 1994 book is very similar to the version used for the
September 5, 1993, forecast. However, a few small changes were made in going
from the version used for this forecast to the final version used for
the experiments in the book.
The following lists the order of the changes
that have been made to the US model
since November 1, 1996.
As noted above, the November 1, 1996,
version of the US model is presented and discussed in
Testing Macroeconometric Models (MC1 Version).
The following changes are in the order in which they were made. Some
of the changes are superceded by later changes.
Changes to the US model from November 1, 1996, to the present
- Beginning with the February 1, 1997, version, a time trend
is included as an explanatory variable in
equation 9. The revised flow of funds data on MH showed a large fall
over time that does
not seem capable of being explained with the standard economic
variables.
- Beginning with the May 9, 1997, version, lambda is .65 instead
of .70 and eta is
.4 instead of .3 in equation 29. See the discussion
in Section 5.9 in the
1994 book
for how these values are estimated.
- Beginning with the August 4, 1997, version, the specification of
the price and
wage equations (equations 10 and 16) is similar to that in
Table 5 in Testing the NAIRU
Model for the United States.
Equation 10 differs from the price equation in Table 5 in that it uses
as the demand pressure variable log[(YS-Y)/YS + .04] instead of the
unemployment rate.
Equation 16 differs from the wage rate equation in Table 5 in that it does not
include a demand pressure variable.
- Beginning with the November 11, 1997, version, the variable
TXCR*IKFA is not included in equation 12.
- Beginning with the
November 11, 1997, version,
lambda is .65 instead of .50 in equation 19.
- Beginning with the November 11, 1997, version, revised NIPA data
from 1958:4 were
available and were used. New data on fixed reproducible tangible
wealth were also used for
the first time for this version.
- The 1997 second quarter revisions of the flow of funds data led
to the following
changes beginning with the November 11, 1997, version.
1) The Nonfarm Noncorporate
Business sector and the Farm Business sector in the flow of
funds accounts are moved
from the household sector to the firm sector in the model.
This resulted in the following
exogenous variables being dropped from the model: IVH, IVVH,
PIEH, and TFA. 2) Equation
88, which determined variable INTROW, is dropped from the
model, and INTROW is
taken to be exogenous. In addition, INTROW is now taken to be
net interest payments
instead of net interest receipts of the foreign sector,
which reverses the sign
of the variable. 3) DISBA is dropped from equation 73;
-WLDF+WLDG+WLDF+DISBA is added to
equation 67; -WLDF+WLDG+WLDS is added to equation 70;
and +WLDF-WLDG-WLDS-DISBA is added
to equation 80. 3) SUR and MRS are dropped as separate
variables in the model and
absorbed in the discrepancy variables.
- Beginning with the November 3, 1998, version,
the asset variable, log(AA/POP){-1}, is dropped from equation 7.
- Beginning with the November 3, 1998, version,
equation 9 is estimated under the assumption of no serial
correlation of the error term instead of first order serial correlation.
- Beginning with the November 3, 1998, version,
in equation 10 logPF{-1} is no longer subtracted from logPIM. This
does not affect equation 10 because logPF{-1} is a separate explanatory
variable, but it affects the coefficient restriction on the wage equation.
- Beginning with the November 3, 1998, version,
WLDF is no longer subtracted from compensation in constructing the
data for WF, which affects the identities 65, 67, and 68.
- Beginning with the January 30, 1999, version, the
interest rate variable is dropped from the
fixed investment equation, equation 12. I have been unable to find
significant cost of capital effects in the investment equation, and
the (insignificant) interest rate variable that I had been using in this
equation now has a coefficient estimate of the wrong sign. Alas.
- Beginning with the May 1, 1999, version,
the lagged value of the asset variable, AA/POP, is added to the
durable expenditures equation, equation 3. It now appears possible to
pick up a wealth effect on durable expenditures.
- Beginning with the July 30, 1999, version,
the lagged value of the log of the asset variable, AA/POP, is added to the
consumption of services, equation 1. It now appears possible to
pick up a wealth effect on consumption of services.
- The following changes were made beginning with the November 5, 1999,
version:
- Equation 5 (L1, labor force--men 25-54): The Z variable has been
dropped.
- Equation 6 (L2, labor force--women 25-54): The Z variable has been
dropped.
- Equation 7 (L3, labor force--all others 16+): The Z variable has been
replaced by the unemployment rate, UR.
- Equation 8 (LM, number of moonlighters): The Z variable has been
replaced by the unemployment rate, UR.
- Equation 10 (PF, private nonfarm price deflator): The gap variable
has been replaced by the unemployment rate, UR.
- Equation 12 (KK, stock of capital): This equation replaces the old
equation 12, which explained IKF, nonresidential fixed investment.
IKF is now explained by the identity 92, which before determined KK.
Identity 92 now is: IKF = KK - (1 - DELK)*KK-1. The new
equation 12 has on the left hand side logKK and on the right hand side
logKK-1, logKK-2, logY, logY-1,
logY-2, logY-3, logY-4, logY-5,
and RB*(1-D2G-D2S). This change overrides the change
mentioned above that was made to
equation 12 beginning with the January 30, 1999, version.
- Equation 21 (CCF, capital consumption of the firm sector):
This equation is now estimated under the assumption of first order serial
correlation of the error term.
- Equation 27 (IM, imports): The lagged value of the log of
the asset variable, AA/POP, has been added. It now
appears possible to pick up a wealth effect on imports.
- Equation 30 (RS, three month Treasury bill rate): The variable
JJS has been replaced by the unemployment rate, UR.
- The following changes were made beginning with the January 29, 2000,
version:
- Equation 30 (RS, three month Treasury bill rate):
The variable
PCGDPR (percentage change in real GDP) has been replaced by the
change in UR.
- Equation 89 (AA, total net wealth): This identity is now:
AA = (AH + MH)/PH + (PIH*KH)/PH.
This change allows capital gains
on housing to affect AA. PIH*KH is an estimate of the nominal value
of the housing stock.
When housing prices increase faster than
the overall price level, PIH increases relative to PH, and so
(PIH*KH)/PH increases. Before the second term was only KH, which was
in effect multiplying KH by PH instead of PIH.
- The following changes were made beginning with the July 31, 2000,
version:
- Equation 11 (Y, production of the firm sector): The constant term
is now dropped from the equation.
- Equation 19 (INTF, interest payments of the firm sector): There is now
a simpler specification for INTF. INTF equals (1/400)*RB*(-AF) plus
an exogenous variable (DINTF) that makes the equation fit perfectly
within the estimation period. It is difficult to link INTF, which is from
the NIPA accounts, to AF, which is from the Flow of Funds accounts, and
the current specification seems about as good as any.
- Equation 20 (IVA, inventory valuation adjustment): The constant term
is now dropped from the equation.
- Equation 25 (CG, capital gains or losses on corporate stocks held by
the household sector): The left hand side variable is now CG/GDP-1
instead of just CG, and the cash flow explanatory variable is now
also divided by GDP-1: [<>(CF-TFG-TFS)]/GDP-1.
- Equation 27 (IM, imports): The income variable has been replaced by
consumption plus fixed investment:
log[(CS+CN+CD+IHH+IKF+IHB+IHF+IKB+IKH)/POP].
The interest rate variable has been dropped. IM is now in effect run off
of consumption plus fixed investment rather than being determined by a
separate demand equation.
This change overrides the change
mentioned above that was made to
equation 27 beginning with the November 5, 1999, version.
- Equation 29 (INTG, interest payments of the government sector):
There is now
a simpler specification for INTG. INTG equals
(1/400)*(.2*RS + .8*RB)*(-AG) plus
an exogenous variable (DINTG) that makes the equation fit perfectly
within the estimation period. It is also difficult to link INTG,
which is from
the NIPA accounts, to AG, which is from the Flow of Funds accounts, and
again the current specification seems about as good as any.
- The following changes were made beginning with the October 30, 2000,
version:
- Equation 19 (INTF, interest payments of the firm sector):
The specification for INTF is now
(1/400)*(.3*RS+.7*(1/8)*(RB+RB-1+RB-2
+RB-3+RB-4
+RB-5+RB-6+RB-7))*(-AF)
plus the exogenous variable (DINTF) that makes the equation fit perfectly
within the estimation period.
- Equation 27 (IM, imports): The time trend T has been added to the
equation.
- Equation 29 (INTG, interest payments of the government sector):
The specification for INTG is now
(1/400)*(.2*RS+.8*(1/8)*(RB+RB-1+RB-2
+RB-3+RB-4
+RB-5+RB-6+RB-7))*(-AG)
plus the exogenous variable (DINTG) that makes the equation fit perfectly
within the estimation period.
- The following changes were made beginning with the April 27, 2001,
version:
- Equation 17 (MF, demand deposits and currency--f): The after-tax
interest rate is now lagged once rather than unlagged.
- Equation 19 (INTF, interest payments of the firm sector):
Equation 19 is now estimated. INTF/(-AF) is taken to depend on
a constant, the lagged value of INTF/(-AF), and the variable
.75*(1/400)*(.3*RS+.7*(1/8)*(RB+RB-1+RB-2
+RB-3+RB-4
+RB-5+RB-6+RB-7)). The coefficients on
the second and third variables are constrained so that in the long run
a k point change in the third variable leads to a k
point change in INTF/(-AF). The equation is estimated under the assumption
of a first order autoregressive error term.
- Equation 25 (CG, capital gains or losses on corporate stocks held by
the household sector): The earnings variable is now
[<>(PIEF-TFG-TFS+PX*PIEB-TBG-TBS)]/GDP-1.
- Equation 29 (INTG, interest payments of the government sector):
Equation 29 is now estimated. INTG/(-AG) is taken to depend on
a constant, the lagged value of INTG/(-AG), and the variable
.75*(1/400)*(.3*RS+.7*(1/8)*(RB+RB-1+RB-2
+RB-3+RB-4
+RB-5+RB-6+RB-7)). The coefficients on
the second and third variables are constrained so that in the long run
a k point change in the third variable leads to a k
point change in INTG/(-AG).
- The following change was made for the July 27, 2001, version:
- Equation 9 (MH, demand deposits and currency--h):
The equation is estimated under the assumption of a
fourth order autoregressive error.
- The following changes were made for the November 1, 2001, version:
- Equation 9 (MH, demand deposits and currency--h):
The equation has an
exogenous variable, T941A, added, which equals T-168 from
1994:1 on and 0 otherwise. The equation is estimated under the
assumption of no autoregressive error term.
- Equation 11 (Y, production--f): The equation
is now estimated in log form.
- The following changes were made for the January 31, 2002, version:
- Equation 1 (CS, consumer expenditures: services):
The time trend T has been added to the equation.
- Equation 5 (L1, labor force--men 25-54):
The variables log(WA/PH) and T have been dropped, and the variables
log(AA/POP)-1 and UR have been added.
- Equation 6 (L2, labor force--women 25-54):
The variable log(AA/POP)-1 has been added.
- Equation 7 (L3, labor force--all others 16+):
The Variable T has been dropped, and the
variable log(AA/POP)-1 has been added.
- Equation 9 (MH, demand deposits and currency--h):
The variable T941A, which was added for the November 1, 2001, version,
has been dropped, and the equation is now estimated under the
assumption of a fourth order autoregressive error term.
- Equation 11 (Y, production--f): The dummy variables D593, D594,
and D601 have been added to pick up the effects of the steel strike.
- Equation 12 (KK, stock of capital--f): The interest rate variable has been
dropped.
- Equation 13 (JF, number of jobs--f): All variables related to DD772 have
been dropped, and variable T has been dropped. The dummy variable D593
has been added to pick up effects of the steel strike. The excess
labor variable is now defined to be log[JF/(JHMIN/HFS)]-1.
- Equation 14 (HF, average number of hours paid per job--f):
All variables related to DD772 have
been dropped, and variable T has been dropped. The excess
labor variable is now defined to be log[JF/(JHMIN/HFS)]-1,
and logHF-1 is replaced by log(HF/HFS)-1..
- Equation 21 (CCF, capital consumption--f):
The dummy variables D811824 and D831834 have been dropped, and the
equation is now estimated under the assumption of no serial correlation
of the error term.
- Equation 27 (IM, imports): The variable T has been dropped, and the
equation is now estimated under the assumption of a second order
autoregressive error term.
- All the equations that are estimated by 2SLS have been estimated
using fewer first stage regressors. None of the equations now rejects
the overidentifying restrictions at the 95 percent confidence level. .
- The following changes were made beginning with the
April 27, 2002, version:
- Equation 12 (KK, stock of capital--f): An interest rate
variable has been added back in, which is
an estimate of the real AAA bond rate lagged twice. In addition,
another cost of capital variable has been added:
(CG-2 + CG-3 + CG-4)/
(PX-2YS-2 + PX-3YS-3
+ PX-4YS-4).
- Equation 25 (CG, capital gains or losses on corporate stocks held
by h): The divisor is now PX-1YS-1 rather
than GDP-1.
- The following changes were made beginning with the
January 30, 2003, version:
- Equation 3 (CD, consumer expenditures: durables):
The left hand side variable is now CD/POP - (CD/POP)-1,
and on the right hand side (CD/POP)-1 has been replaced
by DELD*(KD/POP)-1 - (CD/POP)-1.
This incorporates a slightly different partial adjustment assumption.
- Equation 4 (IHH, residential investment--h):
The left hand side variable is now IHH/POP - (IHH/POP)-1,
and on the right hand side (IHH/POP)-1 has been replaced
by DELH*(KH/POP)-1 - (IHH/POP)-1.
This incorporates a slightly different partial adjustment assumption.
Also, the asset variable has been dropped from the equation.
- Equation 9 (MH, demand deposits and currency--h):
A dummy variable has been added that is 1 in 1998:1 and 0
otherwise.
- Equation 12 (KK, stock of capital--f):
The left hand side variable is now log(KK/KK-1), and
the right hand side variables are in addition to the two cost of capital
variables: the constant term, log(KK/KKMIN)-1,
log(KK-1/KK-2, and five change in output terms.
log(KK/KKMIN)-1 is an estimate of the amount
of excess capital on hand.
- Equation 17 (MF, demand deposits and currency--f):
A dummy variable has been added that is 1 in 1998:1 and 0
otherwise.
- Equation 21 (CCF, capital consumption--f):
Nine dummy variables have been added to account for tax law
changes. The constant term has also been added.
Some of the changes from November 5, 1999, on
have lessened the reliance of the model
on peak to peak interpolations. Variables Z, JJS, and the output gap
variable have been replaced by UR. This means that Z and JJS
are no longer needed in the model.
This change also eliminates equations 96 and 97.
These variables
and equations have been retained in
the software, but they play no role in the solution of the model.
The different treatment of INTF and INTG beginning with the July 31, 2000,
version means that three variables have
been dropped from the model---BF, BG, and TI.
The Version of the ROW Model Prior to the 1994 Book
The first version of the ROW model was presented in
Fair (1984),
Chapter 4. Some of the main changes that were made between the model in
the 1984 book and the model in the 1994 book are the
following. First, the number of countries (not counting the United
States) for which
structural equations were estimated was decreased from 42 to 37,
and the trade share matrix was changed from 65 x 65 to 59 x 59.
Some countries were dropped because of poor data, and some were added.
Second, OECD data were used whenever possible rather than IFS
data. The OECD
has better NIPA and labor data than is available from the IFS data.
Third, annual data
were used for countries in which only annual NIPA data existed. In the
1984 version,
quarterly data were constructed for all the countries by interpolating
the annual data.
Fourth, wage, employment, and labor force equations were added to the
model (equations
12-15). Fifth, estimates of the capital stock of each country were made,
and the capital
stock variable was used in the investment equation. (This change, however,
was later nullified because of the difficulty of getting good capital stock
data.)
Finally, a few more coefficient constraints were imposed.
|