Tailoring the QPM model to Azerbaijan
OGResearch
Tomas Motl
March / ? 2020
2/10
Implications for the model
• Oil effects:
− REER: money inflow, real appreciation; monetary policy decided about the
split between FX and inflation
− GDP: fiscal effects (business cycle), but also investment (potential output)
− FX Reserves
− Which of these can we represent in the model without too much additional
effort?
• Monetary policy:
− Is it different from the standard QPM model?
− Is it different before and after the crisis?
• Note: for other countries, we could consider remittances, foreign aid,
money targeting, parallel exchange rate... Anything that is relevant.
3/10
Introducing oil
• We need to introduce oil prices to the model
• Equations don’t have to sophisticated - external variables (oil prices,
food prices, foreign inflation, …) usually taken over from other models
/ forecasters and imposed over the whole forecast horizon
• We’ll again employ trend-gap decomposition and AR processes
qoilt = oilt − p∗
t
qoilt = qoilt + qoilt
∆qoilt = ρ1∆qoilt−1 + (1 − ρ1)∆qoilss + ε1
qoilt = ρ2qoilt + ε2
• Note that since oil (and food) are important inputs into production,
we can plug them into the Phillips Curve to help forecast inflation
4/10
Changes to Phillips Curve
• Oil prices are regulated in Azerbaijan, but food is an important
inflation driver:
πt = α1Eπt+1
+ (1 − α1 − α4 − α6)πt−1
+ α2(yt)
+ α3zt
+ α4(∆st + π∗
t − πt)
+α5 · qfoodt
+α6 · (∆foodt + π∗
t − πt)
+ επ
t
5/10
Monetary policy
• Clear preference for FX rate stability over inflation stability
• But also sometimes adjustments – clearly not a strict fixed FX rate, so
we cannot use the simple equation:
st = st−1 + εs
t
• We need a rule that shows clear preference for FX smoothing
• The rule should also allow the FX to follow trends (REER movements)
6/10
Exchange rate modeling
• Exchange rate rule - we replace ”natural” UIP with a policy rule:
st = κ1 ∗
(
(st−1 + ∆star
t − κ2zt
)
+ (1 − κ1)
(
Et[st+1] + (i∗
t + premt − it)/4−κ3oilt
)
∆star
t = ∆zt + πtar
t − π∗
t
• Parameter κ1 = 0.85 controls how much the FX is flexible vs controlled
• We weaken the FX rate response to shocks
• Also, the external sector is not just ”*”, we have US, RU, Eurozone
7/10
Changes to trend equations
• IS curve:
yt = β1yt+1 + β2yt−1
− β3rt + β4zt
+ β5y∗
t
+β6oilt
+ εy
t
• REER tnd:
∆zt = ρz
∆zt−1 + (1 − ρz
) · zss + −(∆oilt − ∆oilss) + εz
t
• Output potential:
∆yt = ρy
∆•neyt−1 + (1 − ρy
) · yss + +(∆oilt − ∆oilss) + εy
t
8/10
Effect of changing kappa
9/10
@CMOP Infrastructure
• Start IRIS
• initialize CMOP:
c = cmop(’./az_model’,’az202004’,’az_’);
• c.readmodel();
• c.observeddata();
• c.analyzemodel();
• c.filterhistory(’scenario’)
• c.forecast(’scenario’) or c.forecast(’base’,’alternative’)
• Scenarios have to be defined in ”az_round_options”
10/10
@CMOP Infrastructure cont.
• Scenarios have to be defined in ”az_round_options”
• Each scenario has a CSV file with tunes
• The tunes CSV is the primary place where you should work