SMALL INFLATION MODEL OF

MONGOLIA (SIMOM) †

Batnyam Damdinsuren
Bank of Mongolia, e-mail: batnyam@mongolbank.mn

Gan-Ochir Doojav
Bank of Mongolia, e-mail: gan_ochir.d@mongolbank.mn

Tomasz Łyziak
National Bank of Poland, e-mail: Tomasz.Lyziak@mail.nbp.pl

April 2008

†
Opinions expressed in this paper are of the authors and do not necessarily correspond to the opinions of
the institutions they work for. The authors wish to thank Boldbaatar Dagva, Ryszard Kokoszczyński and
Ewa Wróbel for their comments and suggestions to the model. All the further comments and suggestions
are very welcome.
ABSTRACT

This paper describes a preliminary version of the small inflation model of Mongolia (SI-
MOM), a successor of the inflation model built at the Bank of Mongolia (in cooperation
with the National Bank of Poland) by Urgamalsuvd Nanjid in July 2007. The model is
rooted in theoretical concepts, however – due to the fact that it will be used for regular
inflation forecasting at the Bank of Mongolia – it is the empirical adequacy and consis-
tency, which played the most important role in solutions applied while building it. The
SIMOM consists of ten estimated equations – domestic currency loan rate and foreign
currency loan rate equations, the IS curve, the LM curve, the exchange rate equation, the
Phillips curve in terms of the net inflation (excluding food prices and fuel prices), food
price dynamics and fuel price dynamics equations as well as approximating (ad-hoc) equ-
ations for the real GDP and public wage. In the paper we discuss each equation and
present in detail estimation results.

The intended primary use of the model is analysis of the monetary transmission mechan-
ism and the inflation process in Mongolia, estimation of dynamic responses of selected
variables to different shocks hitting the Mongolian economy as well as forecasting ma-
croeconomic categories (e.g. exchange rate, output gap, inflation) over a medium term,
consistent with the lags in the monetary transmission mechanism.

Conclusions from the paper are the following: Mongolian inflation is driven by a large
number of shocks, both internal and external. At the same time the effectiveness of the
monetary transmission mechanism is relatively weak (although stronger than previously
perceived). The exchange rate channel seems to be the most important channel of mone-
tary transmission mechanism in Mongolia.

-2-
I. INTRODUCTION

It is widely recognized that due to the lags in the monetary transmission mechanism,
monetary policy actions should be forward-looking. Therefore inflation forecasts consti-
tute extremely important input to the decision making process at central banks. They are
also principal communication tool of central banks.

Central banks use many sources of information to predict future inflation rate. For this
purpose they usually make inflation forecast by using different types of economic models,
characterized by different degrees of theoretical and empirical coherence (Pagan, 2002,
2003). Moreover, central banks – especially inflation targeting ones – collect information
on private agents’ (producers’, consumers’, financial markets’) inflation expectations,
which affect the inflation process (Łyziak, 2005).

In recent years, the Bank of Mongolia has also paid attention to enlarging the information
set used in the decision making process and making it more forward looking. Research
studies on the Mongolian economy and econometric models aimed at forecasting ma-
croeconomic performance have been developed.

In this paper, we present a new small scale model SIMOM built to analyse features of the
monetary transmission mechanism and of the inflation process in Mongolia. Moreover,
we formulate some recommendations on the future improvement in terms of modeling
and forecasting at the Bank of Mongolia, especially in the context of likely adoption of
the inflation targeting strategy in future.

Evaluating the model we took into account the three following characteristics:
⋅ Firstly, we analysed statistical properties of the model equations and their economic
consistency.
⋅ Secondly, we derived responses of selected variables to different types of shocks and
assessed their adequacy on the basis of our knowledge concerning the Mongolian
economy.
⋅ Thirdly, we performed additional checks of the forecasting properties of the endo-
genous part of the model.

This paper is structured in the following way. Section II presents a detailed description
of the model. Section III shows estimation results of parameters of the model equations.
Section IV presents simulations of the model, while section V is focused on forecasting
accuracy of the model. Finally, section VI offers some conclusions and recommenda-
tions.

II. DETAILED DESCRIPTION OF THE MODEL

In this section we present a small scale highly aggregated model of the Mongolian econ-
omy, SIMOM. Such small-scale models have been used in many central banks to under-
stand monetary transmission mechanism, derive optimal policy rules or forecast inflation

-3-
(e.g. Bank of England, 1999; Batini and Haldane, 1999; Łyziak, 2002; Kłos et al., 2005).
The SIMOM was designed to capture specific features of the inflation process in Mongo-
lia.

The model is both theoretically and empirically based, although empirical consistency
was the priority while building it. There are following estimated equations in the model:
⋅ Domestic currency loan rate equation: domestic currency loan rate depends on: (1)
its lag; (2) monetary policy rate; (3) reserve money (M0); (4) seasonal dummy va-
riables.
⋅ Foreign currency loan rate equation: foreign currency loan rate depends on: (1) its
lag; (2) monetary policy rate; (3) reserve money (M0); (4) a seasonal dummy variable.
⋅ IS curve: output gap depends on: (1) real average loan rate; (2) real effective ex-
change rate; (3) foreign demand (Chinese economic growth); (4) change in exported
commodity prices (copper and gold prices); (5) reserve money (M0); (6) a dummy va-
riable.
⋅ LM curve: reserve money (M0) depends on: (1) its lag; (2) monetary policy rate; (3)
real GDP; (4) Fiscal expenditure; (5) seasonal dummy variable; (6) dummy variable
related to changes in regulation on central bank bill auctions.
⋅ Exchange rate equation: USD/MNT exchange rate depends on: (1) its lag; (2) inter-
est rate disparity; (3) PPP deviation; (4) change in exported commodity price (gold);
(5) fiscal expenditure relative to GDP (proxy for the risk premium); (6) cross ex-
change rate of Yuan against US dollar.
⋅ Phillips curve: net inflation depends on: (1) its lag; (2) fuel price inflation; (3) output
gap, (4) public wage growth; (5) import price growth; (6) a seasonal dummy variable.
⋅ Food price dynamics: food price inflation depends on: (1) its lag; (2) net inflation; (3)
headline inflation; (4) dummy variable related to significant increase in oil price and
exchange rate; (5) a seasonal dummy variable.
⋅ Fuel price dynamics: fuel price inflation depends on: (1) growth of oil price in the
world market; (2) changes in the USD/MNT exchange rate; (3) a seasonal dummy.
⋅ Real GDP equation (approximating equation, ad-hoc specification): GDP in real
term depends on: (1) its lag; (2) the output gap; (3) seasonal dummy variables.
⋅ Public wage equation (approximating equation, ad-hoc specification): public
wage depends on: (1) its lag; (2) real GDP; (3) change in exported commodity prices
(copper and gold prices); (4) seasonal dummy variables.

-4-
The equations and identities of the model are the following:

INTEREST RATES

Domestic currency loan rate equation:

−1 = c lr dc + α 1 (lrt −1 + α 2 m 0 t −1 ) + α 3 (i t − 2 − i t − 3 ) + α 4 S + α 5 S , α 1 < 0
2 3
lrt dc − lrt dc dc
(1)

Foreign currency loan rate equation:

lrt fc − lrt −fc1 = clr fc + β1 ( lrt −fc1 + β 2 m0t −1 ) + β3 (it −1 − it − 2 ) + β 4 S 2 , β1 < 0 (2)

Average loan rate in nominal and real terms:

lrt ave = wtfc lrt fc + (1 − wtfc ) lrt dc (3)
lrqave
,t = lrt
ave
4 (4)
rlrqave ave
,t = lrq ,t − π t −1 (5)

IS CURVE
yˆ t = c yˆ + γ 1rlrqave r
( China
, t −3 + γ 2 et −3 + γ 3 ∆ 4 y t − 2 )
+ γ 4 ptcop( cop
)gold
( gold
−3 − pt − 4 + γ 5 p t − 0 − pt −1 + )
(6)
+ γ 6 (m0 t −3 − m0 t −4 ) + γ 7 DQ 4,03

LM CURVE
m 0 t = φ1 m 0 t −1 + φ 2 it + φ 3 y t + φ 4 rfe t + φ 5 DQ 3, 07 + φ 6 S 4 (7)

EXCHANGE RATE EQUATION

etUSD / MNT = ϕ1etUSD
−1
/ MNT
+ ϕ 2 etUSD
−2
/ MNT
( )
+ ϕ 3 lrt dc − it f + ϕ 4 dev _ pppt −1 +
(8)
(
+ ϕ 5 ptgold gold
− 2 − p t −3 )
+ ϕ 6 rfet −3 + ϕ 7 etRMB
−1
/ USD

PRICE SYSTEM

Phillips curve (net inflation equation):

π tN = cπ N + θ 1π tN−1 + θ 2π tO + θ 3 yˆ t −3 + θ 4 ( p tIM
−2
_D
− p tIM
−3
_D
) + θ 5 (wtP−2 − wtP−3 ) + θ 6 S 2 (9)

Food price dynamics:

π tF = cπ F + δ 1 (π tF−1 + π t −1 ) + δ 2 π tN−1 + δ 3π tF− 2 + δ 4 S 2 (10)

Fuel price dynamics:

π t0 = cπ 0 + λ1 ( p toil− 2 − p toil−3 ) + λ 2 ( p toil − p toil−1 ) + λ3 (etUSD
−3
/ MNT
− etUSD
−4
/ MNT
) + λ4 S 1 (11)

Quarterly CPI inflation:

π t = wtF π tF + wtOπ tO + (1 − wtF − wtO ) π tN (12)

-5-
Annual CPI inflation:
Π t = (1 + π t )(1 + π t −1 )(1 + π t − 2 )(1 + π t −3 ) − 1 (13)
Π N
t (
= 1+ π )(1 + π )(1 + π )(1 + π ) − 1
t
N N
t −1
N
t −2
N
t −3 (14)
Π F
t = (1 + π )(1 + π )(1 + π )(1 + π ) − 1
t
F F
t −1
F
t −2
F
t −3 (15)
Π Ot = (1 + π )(1 + π )(1 + π )(1 + π ) − 1
O
t
O
t −1
O
t −2
O
t −3 (16)

APPROXIMATING EQUATIONS (AD-HOC SPECIFICATIONS)

Real GDP equation:

yt − yt −1 = ϑ1 yˆ t + ϑ2 S 1 + ϑ3 S 2 + ϑ4 S 4 (17)

Public wage equation:

&cop
wtP = c w P + η1 wtP−1 + η 2 y t + η 3 p tgold
−1 + η 4 S 1 + η 5 S 2 + η 6 S 3 or: (18)
wtP = wtP−1 + π t −1 (19)

DEFINITIONS

Prices (CPI):
pt = pt −1 + π t (20)
F F F
p =p
t t −1 +π t (21)

Import prices:
ptIM _ D = ptIM _ F − etUSD / MNT (22)
( ) (
ptIM _ F = wtRMB ptchina − Index RMB + 1 − wtRMB ptUSA , ) wtRMB = 0.05 (23)

Exchange rates:
etn = etUSD / MNT + wtRMB etRMB / USD , wtRMB = 0.05
(24)
etr = etn + p − p IM _ F (25)

PPP deviation:
dev _ pppt = etUSD / MNT − pppt (26)
( ) ( )
pppt = pt − wtRMB ptchina − Index RMB + 1 − wtRMB ptUSA , wtRMB = 0.05 (27)

Nominal GDP:
ytn = yˆ t + p (28)

Fiscal expenditure relative to GDP:

rfet = fet − ytn (29)

-6-
Other prices:
ptgold &cop = wtcop ptcop + (1 − wtcop ) ptgold (30)
ptChina = ptChina
−4 + π tChina (31)
USA USA USA
p t =p
t −4 +π t (32)

The following symbols have been used:

lr dc - domestic currency loan rate in nominal terms (annual);
i - short-term BoM bill rate (monetary policy rate) in nominal terms (annual);
m0 - reserve money M0 (in logs);
d
S - d-quarter seasonal dummy;
fc
lr - foreign currency loan rate in nominal terms (annual);
lr ave - weighted average of domestic and foreign currency loan rates in nominal terms
(annual);
lrqave - weighted average of domestic and foreign currency loan rates in nominal terms
(quarterly);
rlrqave - weighted average of domestic and foreign currency loan rates in real terms
(quarterly);
wtfc - weight of foreign currency loans in total loans;
π - inflation, quarter on quarter;
ŷ - output gap;
er - real effective exchange rate (in logs);
∆y China - real GDP growth rate in China;
p cop - copper price (in logs);
gold
p - gold price (in logs);
DQ 4, 03 - dummy variable for the 4th quarter 2003;
DQ 3,07 - dummy variable for the 3rd quarter 2007 (related to the changing of required
reserve regulation);
yt - real GDP (in logs);
etUSD / MNT - USD/MNT exchange rate (in logs);
it f - LIBOR 3M in nominal terms;
ppp - USD/MNT exchange rate (in logs) consistent with purchasing power parity
(PPP);
dev _ ppp - deviation of the USD/MNT exchange rate (in logs) from ppp;
p gold &cop - weighted price index of gold and copper (in logs);
wtcop - copper weight in the weighted average price index of copper and gold;
rfe - fiscal expenditures related to the GDP;

-7-
πN - net inflation (price growth of consumer goods and services excluding food and
fuels), quarter on quarter;
πF - food price dynamics, quarter on quarter;
πO - fuel price dynamics, quarter on quarter
F
w - weight of food in the CPI basket;
wO - weight of fuels in the CPI basket;
ptIM _ F - import prices in foreign currency (in logs);
p IM _ D - import prices in domestic currency (in logs);
wP - wage in the public sector (in logs);
p oil - USD oil price per barrel in international markets (in logs);
Π - CPI inflation, year-on-year;
ΠN - net inflation, year-on-year;
ΠF - food price dynamics, year-on-year;
O
Π t - fuel price dynamics, year-on-year;
p - headline consumer price index (CPI);
pF - food price index;
e RMB / USD - RMB/USD exchange rate (in logs);
en - nominal effective exchange rate (in logs);
RBM
w - Chinese RMB weight used to calculate the nominal effective exchange rate
(5%);
p china - Chinese CPI (in logs);
pUSA - US CPI (in logs);
n
y - nominal GDP (in logs);
fe - fiscal expenditures, in nominal terms (in logs);
π tChina - Chinese CPI inflation, year-on-year;
π USA - US inflation, year-on-year.

There are 12 endogenous and 10 exogenous variables in the model. The relationships be-
tween these variables are summarized using the flow-chart (Figure 1), with endogenous
variables shown in the grey boxes. We can use the flow-chart to understand the main lin-
kages in the model.

-8-
Figure 1. Flow-chart scheme of the SIMOM
GDP growth in China Domestic currency loan Domestic short term
rate interest rate

REER Reserve Money

Output Gap Real Activity

Foreign currency loan

rate Fiscal Expendi-
Copper Gold ture
Price Price

CPI Inflation in
Exchange rate
China

CPI Inflation in
USA
Public Wage Foreign short term inter-
est rate

CPI Inflation Net Inflation Import price

Oil
Price Fuel Inflation
Food Inflation

Cross rate of
NEER RMB/USD

Comments:
- Starting point
- Endogenous variables
- Exogenous variables
- Identities
-9-
III. ESTIMATION RESULTS

We estimated the model with OLS technique equation by equation using quarterly
time series covering the period 2001Q1-2008Q1. Model variables are not subject to
seasonal adjustment – the exception is the GDP series, which is used to determine the
output gap. However, in some of the equations there is a need to include seasonal
dummies in order to capture strong seasonality in the Mongolian economy. The model
uses a statistical measure of the output gap (obtained with the use of Hodrick-Prescott
filter). Data sources are presented in the Annex 1 and equations’ estimations with
principal diagnostics are show in Annex 2.

Table 1 presents estimated parameters of the SIMOM.

Table 1. Estimated parameters of the SIMOM

Value
Coefficient Interpretation Equation
(S.E. if estimated)
clr dc 1.28 Intercept
α1 -0.62 ECM parameter

α2 0.14 reserve money (long run) (1) Domestic

currency loan
α3 lagged change in monetary rate equation
0.12
policy rate
α4 -0.02 2nd quarter seasonal dummy
α5 0.01 3rd quarter seasonal dummy
clr fc 0.73 Intercept
β1 -0.69 ECM parameter
(2) Foreign cur-
β2 0.07 reserve money (long run) rency loan rate
β3 lagged change in monetary equation
0.12
policy rate
β4 0.01 2nd quarter seasonal dummy
c ŷ -1.27 Intercept
lagged weighted average
γ1 -0.26
loan rate in real terms
lagged real effective ex-
γ2 -0.15
change rate
γ3 lagged Chinese GDP
2.79
growth
(3) IS curve
growth in the copper price,
γ4 -0.07
3 lags
γ5 lagged growth in the gold
-0.26
price
γ6 0.14 lagged reserve money
γ7 dummy variable for
-0.04
2003Q4
φ1 0.61 lagged reserve money
φ2 -1.26 monetary policy rate

- 10 -
 Value
Coefficient Interpretation Equation
(S.E. if estimated)
φ3 0.16 real activity (real GDP) (4) LM curve
φ4 lagged relative fiscal ex-
0.23
penditures
φ5 -0.22 4th quarter seasonal dummy

φ6 dummy variable for

-0.06
2007Q3
ϕ1 lagged exchange rate (1
0.50
lag)
ϕ2 0.43 Interest rate disparity
ϕ3 0.09 lagged PPP deviation
ϕ4 0.06 change in the gold prices (5) Exchange
ϕ5 lagged relative fiscal ex- rate equation
0.14
penditures (risk premium)
ϕ6 lagged exchange rate of
-0.04
Yuan against US dollar
ϕ7 lagged exchange rate (2
-0.17
lags)
cπ N 0.01 Intercept
θ1 -0.33 lagged net inflation
θ2 0.08 fuel price inflation (6) Phillips curve
θ3 0.15 lagged output gap (net inflation eq-
θ4 0.27 lagged import price growth uation)
θ5 0.12 lagged public wage growth
θ6 -0.03 2nd quarter seasonal dummy
cπ F -1.90 Intercept
δ1 -0.41 Relative price
δ2 1.19 Net Inflation (7) Food price
δ3 -0.42 Food price inflation dynamics
dummy variable for
δ4 0.11
2007Q4
cπ O 0.04 Intercept
change in the oil price per
λ1 0.41 barrel in the world market,
2 lags
change in the oil price per (8) Fuel price
λ2 0.18 barrel in the world market, dynamics
3 lags
λ3 change in USD/MNT ex-
-2.05
change rate
λ4 -0.09 1st quarter seasonal dummy
ϑ1 0.84 output gap (9) Real GDP
ϑ2 -0.40 1st quarter seasonal dummy equation (ad-hoc

- 11 -
 Value
Coefficient Interpretation Equation
(S.E. if estimated)
ϑ3 0.44 2nd quarter seasonal dummy specification)
ϑ4 0.08 4th quarter seasonal dummy
cwP -6.06 Intercept
η1 0.74 lagged public wage
η2 0.52 real GDP
(10) Public wage
η3 Weighted average price of equation (ad-hoc
0.04
copper and gold specification)
η4 0.22 1st quarter seasonal dummy
η5 0.07 2nd quarter seasonal dummy
η6 -0.11 3rd quarter seasonal dummy

The estimated coefficients seem to be in line with our expectations, economic intui-
tion and data typical for small open economies. Signs of estimated coefficients are
theoretically and empirically consistent. Diagnostic tests of the SIMOM equations are
satisfactory (see Annex 2).

IV. SIMULATIONS OF THE MODEL

After estimating the model we checked its dynamic properties by conducting different
simulations. They allowed us analyzing responses of different variables to a number
of shocks and comparing them with our intuition. Below we present the response of
annual inflation to the following shocks:
(1) Chinese GDP growth impulse: increase of the Chinese GDP growth by 1 pp for
4 quarters;
(2) Fiscal expenditure impulse: increase of fiscal expenditures by 10% for 1 quar-
ter;
(3) Domestic interest rate temporary impulse: increase of the domestic short-term
interest rate by 1 pp for 4 quarters (or for 8 quarters);
(4) Domestic interest rate permanent impulse: increase of the domestic short-term
interest rate permanently;
(5) Oil price impulse: increase of oil prices in international markets by 10% for 4
quarters;
(6) Copper price impulse: increase of the copper price by 10% for 4 quarters;
(7) Gold price impulse: increase of the gold price by 10% for 4 quarters;
(8) Chinese inflation impulse: increase of Chinese inflation by 1 pp for 4 quarters;
(9) RMB/USD cross exchange rate impulse: increase of the RMB/USD exchange
rate (USD appreciation) by 1% for 1 quarter;
(10) US inflation impulse: increase of the US inflation by 1 pp for 4 quarters;
(11) Food price impulse: increase of the q-o-q food inflation by 1 pp for 1 quarter;

- 12 -
(12) Foreign interest rate impulse: increase of the foreign short-term interest rate by
1 pp for 4 quarters.

Below we inspect in detail the results of the monetary transmission mechanism simu-
lation, in which we change the short-term domestic interest rate by 1 pp for 8 quarters
(Figure 2) as well as present annual inflation responses to the shocks defined above
(Figure 3).

In the monetary transmission mechanism simulation we present inflation response to

the interest rate changes including both interest rate channel and exchange rate chan-
nel effects and separately we assess the importance of the interest rate channel. It
seems that the exchange rate channel constitutes the dominant channel of monetary
transmission mechanism in the Mongolian economy.

The maximum impact of monetary policy changes on inflation manifests in the 9th
quarter after an 8-quarter interest rate impulse. It equals approximately 0.3 pp. Mone-
tary policy actions are relatively good in terms of their impact on inflation. As far as
the maximum impact is concerned, the effects of interest rate changes defined above
are roughly the same as:
⋅ oil price change by less than 1% (average absolute quarterly change: 6%);
⋅ change in the Chinese GDP by 0.1 pp (average absolute quarterly change: 0.6 pp);
⋅ increase of fiscal expenditures by 7.5% (average absolute quarterly change: 20%);
⋅ change in copper price by 2.5% (average absolute quarterly change: 8%).

- 13 -
Figure 2, part 1: Reaction of different variables to the increase of the domestic
short-term interest rates by 1 pp for 8 quarters

Short-term interest rate (I, in pp) Loan rate on loans in domestic currency (LR, in pp)
1.2 0.5

1 0.4
0.8
0.3
0.6
0.2
0.4
0.1
0.2

0 0

-0.2 -0.1
0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60

Loan rates on loans in foreign currency (LR_F, in pp) Money base (M0_L, in %)
0.25 0.5
0
0.2
-0.5
0.15 -1
-1.5
0.1
-2
0.05 -2.5
-3
0
-3.5
-0.05 -4
0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60

USD/TOG exchange rate (USDTG_L, in %, +=appreciation) Nominal effective exch. rate (NEER_L, in %, +=appreciation)
0.2 0.2

0.15 0.15

0.1 0.1

0.05 0.05

0 0

-0.05 -0.05

-0.1 -0.1

-0.15 -0.15
0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60

Real effective exch. rate (REER_L, in %, +=appreciation) Output gap (GAP, in pp)
0.1 0.2
0.15
0.05 0.1
0.05
0
0
-0.05
-0.05
-0.1

-0.1 -0.15
-0.2
-0.15 -0.25
0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60

- 14 -
Figure 2, part 2: Reaction of different variables to the increase of the domestic
short-term interest rates by 1 pp for 8 quarters

Annual net inflation (ANINF, in pp) Annual food inflation (AFINF, in pp)
0.1 0.1

0.05 0.05

0 0

-0.05 -0.05

-0.1 -0.1

-0.15 -0.15
0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60

Annual fuel inflation (AOINF, in pp) Annual CPI inflation (AINF, in pp)
0.25 0.1
0.2
0.15 0.05
0.1
0.05 0
0
-0.05 -0.05
-0.1
-0.15 -0.1
-0.2 all channels interest rate channel
-0.25 -0.15
0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60

Figure 3, part 1: Annual CPI inflation responses to different impulses

(9) RMB/USD impulse (10) US inflation impulse

0.7 0.3
0.6 0.25
0.5 0.2
0.4 0.15
0.3 0.1
0.2 0.05
0.1 0
0 -0.05
-0.1 -0.1
-0.2 -0.15
0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60
quarter after impulse quarter after impulse

(11) Food inflation impulse (12) Foreign interest rate impulse

0.5 0.1
0.4 0.08
0.3
0.06
0.2
0.04
0.1
0.02
0
-0.1 0

-0.2 -0.02
-0.3 -0.04
0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60
quarter after impulse quarter after impulse

- 15 -
Figure 3, part 2: Annual CPI inflation response to different impulses

Definition of impulses:
(1) increase of the Chinese GDP growth by 1 pp for 4 quarters; (2) increase of fiscal expenditures by 10% for 1 quarter; (3) increase of the domestic short-term interest rate by 1 pp for 4
quarters; (4) increase of the domestic short-term interest rate permanently; (5) increase of oil prices in international markets by 10% for 4 quarters; (6) increase of the copper price by 10% for 4
quarters; (7) increase of the gold price by 10% for 4 quarters; (8) increase of Chinese inflation by 1 pp for 4 quarters; (9) increase of the RMB/USD exchange rate by 1% for 1 quarter; (10)
increase of the US inflation by 1 pp for 4 quarters; (11) increase of the q-o-q food inflation by 1 pp for 1 quarter; (12) increase of the foreign short-term interest rate by 1 pp for 4 quarters.
(1) Domestic interest rate impulse, temporary (2) Domestic interest rate impulse, permanent

0.06 0.04
0.04 0.02
0.02 0
0
-0.02
-0.02
-0.04
-0.04
-0.06
-0.06
-0.08 -0.08
-0.1 -0.1
-0.12 -0.12
-0.14 -0.14
0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60
quarter after impulse quarter after impulse

(3) Chinese growth impulse (4) Fiscal expenditure impulse

2.5 0.25
2 0.2
1.5
0.15
1
0.1
0.5
0.05
0
-0.5 0

-1 -0.05
-1.5 -0.1
0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60
quarter after impulse quarter after impulse

(5) Oil price impulse (6) Copper price impulse

0.6 0.2
0.4 0.15

0.2 0.1
0.05
0
0
-0.2
-0.05
-0.4
-0.1
-0.6 -0.15
-0.8 -0.2
0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60
quarter after impulse quarter after impulse

(7) Gold price impulse (8) Chinese inflation impulse

1.5 0.015

1
0.01
0.5
0.005
0
0
-0.5

-1 -0.005

-1.5 -0.01
0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60
quarter after impulse quarter after impulse

- 16 -
V. FORECASTING ACCURACY

We checked the forecasting accuracy of the SIMOM in the following way:

⋅ We started by cutting our sample and preparing a forecast of endogenous va-
riables starting from 2002Q1. We assumed that all the exogenous variables have
their true values (perfect forecast of exogenous variables). Having true data of an-
nual inflation and our forecast, we calculated forecast errors for the first quarter
projected (t+1) and for the successive quarters (t+2,…,t+12).
⋅ We repeated the same procedure starting our forecast exercise in 2002Q2,
2002Q3 … and 2008Q1. We obtained a matrix of forecast errors from all runs for
different forecast horizons.
⋅ Then we averaged the, calculating mean error (ME), mean absolute error (MAE)
and mean absolute percentage error (MAPE).

Because of the role food price play in determining inflation in Mongolia and inability
to forecast them well with the use of our statistical approximating equation, we re-
peated the procedure described above with food prices assumed exogenous (perfect
forecast).

Figures 4 and 5 present the results of the all forecasting runs relative to true inflation
with food inflation treated as exogenous and endogenous respectively. It occurs that
food price shocks deteriorate forecasting properties of the model in 2004-2005, when
the peak of inflation is not able to be predicted by the model.

Figure 4. True inflation and inflation forecasts, food prices exogenous

20%

AINF Base
AINF 200201
AINF 200202
15%
AINF 200203
AINF 200204
AINF 200301
AINF 200302
AINF 200303
10% AINF 200304
AINF 200401
AINF 200402
AINF 200403
AINF 200404
AINF 200501
5% AINF 200502
AINF 200503
AINF 200504
AINF 200601
AINF 200602
0% AINF 200603
AINF 200604
01

02

03

04

01

02

03

04

01

02

03

04

01

02

03

04

01

02

03

04

01

02

AINF 200604
02

02

02

02

03

03

03

03

04

04

04

04

05

05

05

05

06

06

06

06

07

07
20

20

20

20

20

20

20

20

20

20

20

20

20

20

20

20

20

20

20

20

20

20

AINF 200701

-5%

- 17 -
 Figure 5. True inflation and inflation forecasts, food prices endogenous
20%

AINF Base
AINF 200201
AINF 200202
15%
AINF 200203
AINF 200204
AINF 200301
AINF 200302
AINF 200303
10% AINF 200304
AINF 200401
AINF 200402
AINF 200403
AINF 200404
AINF 200501
5% AINF 200502
AINF 200503
AINF 200504
AINF 200601
AINF 200602
0% AINF 200603
AINF 200604
01

02

03

04

01

02

03

04

01

02

03

04

01

02

03

04

01

02

03

04

01

02
AINF 200604
02

02

02

02

03

03

03

03

04

04

04

04

05

05

05

05

06

06

06

06

07

07
20

20

20

20

20

20

20

20

20

20

20

20

20

20

20

20

20

20

20

20

20

20
AINF 200701

-5%

The results of our exercise (Figure 6 and 7) suggest that in the case of food prices
treated as exogenous the forecasting errors of the endogenous part of the model are
relatively small.
⋅ Mean absolute errors (MAE) increase from approximately 0.5 pp in the first quar-
ter projected to approximately 1.5 pp in the sixth quarter projected and then stabil-
ize.
⋅ Mean absolute percentage errors (MAPE) do not exceed 20% of the value of true
annual inflation.
However, if we treat food prices as endogenous, the forecasting accuracy of the model
deteriorates significantly (MAE increases to 4.5 pp and MAPE reaches 110% of true
inflation), even with statistical properties of the equation for food price changes being
satisfactory. It is due to huge shocks to food prices, which are not possible to be pre-
dicted with a purely statistical approach.

- 18 -
 Figure 6. Mean absolute error
MAE

4.5

4.0

3.5

3.0

2.5

2.0

1.5

1.0

MAE (in pp)_food

0.5 MAE (in pp)_no food

0.0
t+1 t+2 t+3 t+4 t+5 t+6 t+7 t+8 t+9 t+10 t+11 t+12

Figure 7. Mean absolute percentage error

MAPE

120.0

100.0

80.0

60.0

40.0

MAPE (in % of inflation)_food

MAPE (in % of inflation)_no food
20.0

0.0
t+1 t+2 t+3 t+4 t+5 t+6 t+7 t+8 t+9 t+10 t+11 t+12

VI. CONCLUSIONS AND RECOMENDATIONS

The results of the model simulations and forecasting exercises presented in this paper
lead to some conclusions, which may be important for the monetary policy making in
Mongolia:

- 19 -
⋅ CPI inflation in Mongolia is driven by a large number of shocks, both internal
(fiscal expenditure shocks, food price shocks) and external (Chinese GDP shocks,
US inflation shocks, copper, gold and oil price shocks).
⋅ At the same time the effectiveness of the monetary transmission mechanism is rel-
atively weak (although stronger than previously perceived). The exchange rate
channel seems to be the most efficient channel of the monetary transmission me-
chanism in Mongolia.
⋅ Net inflation, excluding food prices and fuel prices, seems to be useful in model-
ing inflation in Mongolia and analyzing the link between monetary policy actions
and prices in the economy.
⋅ Forecasting properties of the quarterly forecasting model SIMOM are relatively
good. However, food price shocks may significantly undermine the forecasting ac-
curacy.

REFERENCES

Bank of England (1999), Economic models at the Bank of England.

Batini N., Haldane A. G. (1999), Forward-looking rules for monetary policy, in Tay-
lor J. B., “Monetary Policy Rules”, University of Chicago Press for NBER

Beechey M., Bharucha N., Cagliarini A., Gruen D., Thompson C. (2000), A small
model of the Australian macroeconomics, Economic Research Department, Reserve
Bank of Australia

Berg A., Karam P. and Laxton D. (2006), A practical model-based approach to mone-
tary policy analysis – guide, IMF WP/06/81

Berg A., Karam P., Laxton D. (2006), A practical model-based approach to monetary
policy analysis-overview, IMF WP/06/80

Canales-Kriljenko J. I., Kisinbay T., Maino R., Parrado E. (2006), Setting the opera-
tional framework for producing inflation forecasts, IMF WP/06/122

Fic T., Kolasa M., Kot A., Murawski K., Rubaszek M., Tarnicka M. (2005), ECMOD
model of the Polish Economy, Paper No 36, the National Bank of Poland

Kłos B., Kokoszczyński R., Łyziak T., Przystupa J., Wróbel E. (2005), Structural
econometric models in forecasting inflation at the National Bank of Poland, Paper No
31, National Bank of Poland

Kotlan V. (2002), Monetary policy and term spread in a macro model of a small open
economy, Working Paper Series, No. 1/2002, Czech National Bank
Łyziak T. (2002), Monetary transmission mechanism in Poland. The strength and de-
lays, Paper No 26, National Bank of Poland

- 20 -
Łyziak T. (2005), Inflation targeting and consumer inflation expectations in Poland.
A success story?, “Journal of Business Cycle Measurement and Analysis”, No 2(2),
pp. 185-212

Pagan A. (2002), What is a good macroeconomic model for central bank to use?,
Australian National University, mimeo

Pagan A. (2003), Report on modelling and forecasting at the Bank of England, Bank
of England.

- 21 -
ANNEX 1. DATA SOURCES

Quarterly real GDP

Definition: Gross domestic product (2005 reference prices).
Source: National Statistical Office, unpublished data

Output gap
Definition: The data was obtained by subtracting from a seasonally adjusted quarterly real
GDP data the potential output estimated using the Hodrick-Prescott (HP) filter.
Source: author’s estimation

Nominal exchange rate (USD/MNT)

Definition: Nominal exchange rate of US dollar against Mongolian Togrog (USD/MNT).
Source: Bank of Mongolia, Monthly Bulletin

Headline consumer price index (CPI)

Definition: Acquisitions consumer price index (December 2005 = 100).
Source: National Statistical Office, Monthly Bulletin

Food price index

Definition: Acquisitions food price index (December 2005 = 100).
Source: National Statistical Office, Monthly Bulletin

Fuel price index

Definition: Acquisitions fuel price index (December 2005 = 100).
Source: National Statistical Office, Monthly Bulletin and author’s calculation

Nominal domestic currency loan rate

Definition: Weighted average lending rate of commercial banks (loans in Togrog)
Source: Bank of Mongolia, Monthly Bulletin

Reserve money
Definition: Reserve money or sum of outside banks currency and commercial banks’ reserves.
Source: Bank of Mongolia, Monthly Bulletin

Short term central bank bill rate (monetary policy rate)

Definition: 1-week central bank bill rate
Source: Bank of Mongolia, Monthly Bulletin

Nominal foreign currency loan rate

Definition: Weighted average lending rate of commercial banks (loans in foreign currency).
Source: Bank of Mongolia, Monthly Bulletin

Foreign interest rate

Definition: USD LIBOR 3M.
Source: Bloomberg

Chinese economic growth

Definition: Annual growth of Chinese real GDP.
Source: Bloomberg

Chinese inflation
Definition: Chinese annual CPI inflation.
Source: Bloomberg

- 22 -
Exchange rate RMB/USD
Definition: Nominal exchange rate of Chinese Yuan against US dollar.
Source: Bloomberg

US inflation and CPI

Definition: Annual US inflation and US consumer price index
Source: Bloomberg

Oil price
Definition: USD price of crude oil per barrel.
Source: Energy Information Administration of US government

Copper price
Definition: USD price of crude copper per 1 ton.
Source: Bloomberg

Gold price
Definition: USD price of crude gold per 1 ounce.
Source: Bloomberg

Weight of foreign currency loans in total loans

Definition: Foreign currency loans to total loans.
Source: Balance sheet of Commercial bank

Copper weight used for the weighted average price of copper and gold
Definition: Copper weight in total trade of copper and gold.
Source: author’s calculation

Fiscal expenditure
Definition: Fiscal expenditure of the total budget account.
Source: Bank of Mongolia, Monthly Bulletin

Wage in the public sector

Definition: Average wage in the public sector (monthly).
Source: National Statistical Office, Monthly Bulletin

Weight of food in the CPI basket

Definition: Weight of food items in the CPI basket.
Source: National Statistical Office, unpublished data

Weight of fuel in the CPI basket

Definition: Weight of fuel items in the CPI basket.
Source: National Statistical Office, unpublished data

Chinese RMB weight used for the calculation of the nominal effective exchange rate
Definition: Share of trade in RMB in total trade.
Source: Bank of Mongolia, International Department, unpublished data

- 23 -
ANNEX 2. ESTIMATION RESULTS OF EQUATIONS OF THE SIMOM

(1) Domestic currency loan rate equation

Dependent Variable: D(LR)
Method: Least Squares
Sample: 2001:1 2008:1
Included observations: 29
Convergence achieved after 5 iterations
Newey-West HAC Standard Errors & Covariance (lag truncation=3)
D(LR)=C(1)*(LR(-1)+0*I(-1)+C(3)*M0_L(-1))+C(4)*D(I(-2))+C(6)
*@SEAS(2)+C(7)*@SEAS(3)+C(8)

Coefficient Std. Error t-Statistic Prob.

C(1) -0.619242 0.114901 -5.389342 0.0000

C(3) 0.141984 0.010956 12.96001 0.0000
C(4) 0.119251 0.046726 2.552113 0.0178
C(6) -0.021461 0.010774 -1.991976 0.0584
C(7) 0.013223 0.006430 2.056612 0.0512
C(8) 1.276916 0.225927 5.651895 0.0000

R-squared 0.568933 Mean dependent var -0.004793

Adjusted R-squared 0.475223 S.D. dependent var 0.023564
S.E. of regression 0.017070 Akaike info criterion -5.121021
Sum squared resid 0.006702 Schwarz criterion -4.838132
Log likelihood 80.25480

(2) Foreign currency loan rate equation

Dependent Variable: D(LR_F)
Method: Least Squares
Sample: 2001Q1 2008Q1
Included observations: 29
Convergence achieved after 8 iterations
Newey-West HAC Standard Errors & Covariance (lag truncation=3)
D(LR_F)=C(1)*(LR_F(-1)+0*I(-1)+C(3)*M0_L(-1))+C(4)*D(I(-1))+C(6)
*@SEAS(2)+C(8)

Coefficient Std. Error t-Statistic Prob.

C(1) -0.685238 0.109832 -6.238960 0.0000

C(3) 0.071372 0.006423 11.11183 0.0000
C(4) 0.117076 0.049969 2.342974 0.0278
C(6) 0.013081 0.006288 2.080092 0.0484
C(8) 0.727983 0.125707 5.791127 0.0000

R-squared 0.615310 Mean dependent var -0.004138

Adjusted R-squared 0.551195 S.D. dependent var 0.018351
S.E. of regression 0.012294 Akaike info criterion -5.803826
Sum squared resid 0.003627 Schwarz criterion -5.568086
Log likelihood 89.15548

- 24 -
(3) IS curve
Dependent Variable: GAP
Method: Least Squares
Sample: 2002:3 2008:1
Included observations: 23
Newey-West HAC Standard Errors & Covariance (lag truncation=2)
GAP=C(1)+C(3)*LR_AV_R_Q(-1)-0.15*REER_L(-4)+C(5)*CHI_G(-2)
+C(6)*(P_COP_L(-3)-P_COP_L(-4))+C(7)*(P_GOLD_L(-0)
-P_GOLD_L(-1))+C(9)*D2+0*(GAP(-1)-GAP(-2))+C(11)*(LOG(M0(
-3))-LOG(M0(-4)))

Coefficient Std. Error t-Statistic Prob.

C(1) -1.273819 0.051500 -24.73448 0.0000

C(3) -0.255613 0.151380 -1.688546 0.1107
C(5) 2.788468 0.526521 5.296021 0.0001
C(6) -0.074794 0.025057 -2.984950 0.0087
C(7) -0.258858 0.076508 -3.383428 0.0038
C(9) -0.037257 0.011445 -3.255294 0.0050
C(11) 0.136201 0.053316 2.554608 0.0212

R-squared 0.763180 Mean dependent var -0.004190

Adjusted R-squared 0.674372 S.D. dependent var 0.041506
S.E. of regression 0.023685 Akaike info criterion -4.402163
Sum squared resid 0.008976 Schwarz criterion -4.056578
Log likelihood 57.62487 F-statistic 8.593636
Durbin-Watson stat 2.414197 Prob(F-statistic) 0.000278

(4) LM curve
Dependent Variable: M0_L
Method: Least Squares
Sample: 1999Q1 2008Q1
Included observations: 37
Newey-West HAC Standard Errors & Covariance (lag truncation=3)

Variable Coefficient Std. Error t-Statistic Prob.

I -1.264446 0.633755 -1.995165 0.0549

M0_L(-1) 0.610773 0.101785 6.000609 0.0000
@SEAS(4) -0.224472 0.039543 -5.676699 0.0000
FEXP_L 0.234383 0.065194 3.595181 0.0011
LOG(GDP) 0.163402 0.065751 2.485156 0.0186
DUM0703 -0.055117 0.038285 -1.439670 0.1600

R-squared 0.975275 Mean dependent var 12.25900

Adjusted R-squared 0.971288 S.D. dependent var 0.517072
S.E. of regression 0.087617 Akaike info criterion -1.884301
Sum squared resid 0.237976 Schwarz criterion -1.623071
Log likelihood 40.85956

- 25 -
(5) Exchange rate equation
Dependent Variable: USDTG_L
Method: Least Squares
Sample: 2000:1 2008:1
Included observations: 33
Newey-West HAC Standard Errors & Covariance (lag truncation=3)
USDTG_L=C(2)*USDTG_L(-1)+C(3)*(LR-I_F)+C(4)*DEV_PPP(-1)+C(8)
*D(P_GOLD_L(-2))+C(10)*FEXP_REL(-3)+C(11)*USDTG_L(-2)
+C(13)*RMBUSD_L(-1)

Coefficient Std. Error t-Statistic Prob.

C(2) 0.496323 0.128171 3.872337 0.0007

C(3) 0.092922 0.042348 2.194251 0.0373
C(4) 0.062959 0.026729 2.355425 0.0263
C(8) 0.141484 0.048213 2.934580 0.0069
C(10) -0.039316 0.011493 -3.420735 0.0021
C(11) 0.426274 0.107713 3.957488 0.0005
C(13) -0.177042 0.068742 -2.575452 0.0160

R-squared 0.946518 Mean dependent var -7.045582

Adjusted R-squared 0.934177 S.D. dependent var 0.038963
S.E. of regression 0.009996 Akaike info criterion -6.187334
Sum squared resid 0.002598 Schwarz criterion -5.869893
Log likelihood 109.0910 Durbin-Watson stat 2.143098

(6) Phillips curve (net inflation equation)

Dependent Variable: QNINF
Method: Least Squares
Sample: 2002Q1 2008Q1
Included observations: 25
Newey-West HAC Standard Errors & Covariance (lag truncation=2)
QNINF=C(1)*QOINF+C(2)*QNINF(-1)+C(4)*GAP(-3)+C(5)
*D(IMP_CPI_D_L(-2))+C(6)*D(LOG(W_P(-2)))+C(8)*@SEAS(2)
+C(9)

Coefficient Std. Error t-Statistic Prob.

C(1) 0.083116 0.025354 3.278269 0.0042

C(2) -0.332335 0.112208 -2.961766 0.0084
C(4) 0.159537 0.046447 3.434840 0.0030
C(5) 0.271257 0.166376 1.630383 0.1204
C(6) 0.120545 0.029407 4.099136 0.0007
C(8) -0.028405 0.006071 -4.678977 0.0002
C(9) 0.013176 0.004831 2.727269 0.0138

R-squared 0.787783 Mean dependent var 0.012242

Adjusted R-squared 0.717045 S.D. dependent var 0.021766
S.E. of regression 0.011578 Akaike info criterion -5.847951
Sum squared resid 0.002413 Schwarz criterion -5.506665

- 26 -
Log likelihood 80.09938

(7) Food price dynamics

Dependent Variable: QFINF
Method: Least Squares
Sample: 2003:1 2008:1
Included observations: 21
Newey-West HAC Standard Errors & Covariance (lag truncation=2)
QFINF=C(1)*QFINF(-2)+C(2)*QNINF(-1)+C(3)*(LOG(FCPI(-1))-CPI_L(
-1))+C(7)*@SEAS(2)+C(8)*DUM0703+C(11)*DUM0801+C(12)

Coefficient Std. Error t-Statistic Prob.

C(1) -0.427864 0.099380 -4.305315 0.0007

C(2) 1.186077 0.505999 2.344028 0.0344
C(3) -0.410402 0.130924 -3.134667 0.0073
C(7) 0.108950 0.024246 4.493514 0.0005
C(8) 0.113494 0.018783 6.042253 0.0000
C(11) 0.175425 0.011735 14.94827 0.0000
C(12) -1.897643 0.602882 -3.147619 0.0071

R-squared 0.816770 Mean dependent var 0.039755

Adjusted R-squared 0.738243 S.D. dependent var 0.076972
S.E. of regression 0.039380 Akaike info criterion -3.369893
Sum squared resid 0.021711 Schwarz criterion -3.021719
Log likelihood 42.38388

(8) Fuel price dynamics

Dependent Variable: QOINF
Method: Least Squares
Sample: 2001Q2 2008Q1
Included observations: 28
Newey-West HAC Standard Errors & Covariance (lag truncation=3)
QOINF=C(1)+C(2)*D(LOG(OIL(-2)))+C(4)*D(LOG(OIL(-0)))+C(5)
*D(USDTG_L(-3))+C(7)*@SEAS(1)

Coefficient Std. Error t-Statistic Prob.

C(1) 0.037370 0.008924 4.187585 0.0004

C(2) 0.414751 0.098680 4.202975 0.0003
C(4) 0.183016 0.104120 1.757743 0.0921
C(5) -2.051152 0.779133 -2.632607 0.0149
C(7) -0.084149 0.025458 -3.305394 0.0031

R-squared 0.452516 Mean dependent var 0.046059

Adjusted R-squared 0.357301 S.D. dependent var 0.096959
S.E. of regression 0.077730 Akaike info criterion -2.110709
Sum squared resid 0.138966 Schwarz criterion -1.872815
Log likelihood 34.54992

- 27 -
(9) Real GDP equation (approximation, ad-hoc specification)
Dependent Variable: LOG(GDP)
Method: Least Squares
Sample: 2002Q1 2008Q1
Included observations: 25
LOG(GDP)=LOG(GDP(-1))+C(3)*GAP+C(4)*@SEAS(1)+C(5)
*@SEAS(2)+C(6)*@SEAS(4)

Coefficient Std. Error t-Statistic Prob.

C(3) 0.838282 0.273912 3.060405 0.0059

C(4) -0.402544 0.020863 -19.29442 0.0000
C(5) 0.437864 0.022714 19.27757 0.0000
C(6) 0.084676 0.022540 3.756676 0.0012

R-squared 0.952921 Mean dependent var 13.39495

Adjusted R-squared 0.946195 S.D. dependent var 0.237929
S.E. of regression 0.055190 Akaike info criterion -2.810430
Sum squared resid 0.063964 Schwarz criterion -2.615410
Log likelihood 39.13037

(10) Public wage equation

Dependent Variable: LOG(W_P)
Method: Least Squares
Sample: 2002:2 2008:1
Included observations: 24
LOG(W_P)=C(1)+C(2)*LOG(GDP)+C(3)*@SEAS(1)+C(4)*@SEAS(2)
+C(5)*@SEAS(3)+C(6)*LOG(W_P(-1))+C(7)*DUM_WP+C(8)
*WP_COPGOLD_L(-1)

Coefficient Std. Error t-Statistic Prob.

C(1) -6.055365 1.817289 -3.332087 0.0042

C(2) 0.517032 0.156447 3.304835 0.0045
C(3) 0.216022 0.073904 2.923022 0.0100
C(4) 0.071210 0.021050 3.382908 0.0038
C(5) -0.109433 0.028925 -3.783396 0.0016
C(6) 0.739546 0.072916 10.14244 0.0000
C(7) 0.139165 0.034176 4.072002 0.0009
C(8) 0.041178 0.022894 1.798606 0.0910

R-squared 0.994207 Mean dependent var 4.612799

Adjusted R-squared 0.991672 S.D. dependent var 0.380068
S.E. of regression 0.034684 Akaike info criterion -3.623868
Sum squared resid 0.019248 Schwarz criterion -3.231183
Log likelihood 51.48642 F-statistic 392.2543
Durbin-Watson stat 2.663098 Prob(F-statistic) 0.000000

- 28 -