A CLOSER LOOK AT EXCHANGE RATE INDUCED INFLATION IN ETHIOPIA

Abstract

In this paper, we examine the rate of exchange rate pass-through (ERPT) to domestic prices with a particular focus on the Ethiopian economy. We have employed a Structural Vector Autoregressive (SVAR) model where identification is achieved based on a combination of short- and long-run restrictions. The long-run identifying restrictions are derived from a simple small open economy macro model. The results from the impulse response analysis point to a pass-through rate that starts out small but steadily increases in subsequent quarters, becoming close to complete in about two years. According to the variance decomposition results, however, other macroeconomic shocks, including shocks to aggregate demand, output supply as well as shocks to foreign financial flows and trade balance appear to play an important role in explaining the variation in inflation in Ethiopia.

Full text

A CLOSER LOOK AT EXCHANGE RATE
INDUCED INFLATION IN ETHIOPIA
Tseday J. Mekasha*Kiflu G. Molla†
Abstract
In this paper, we examine the rate of exchange rate pass-through (ERPT)
to domestic prices with a particular focus on the Ethiopian economy. We
have employed a Structural Vector Autoregressive (SVAR) model where iden-
tification is achieved based on a combination of short- and long-run restric-
tions. The long-run identifying restrictions are derived from a simple small
open economy macro model. The results from the impulse response analy-
sis point to a pass-through rate that starts out small but steadily increases
in subsequent quarters, becoming close to complete in about two years. Ac-
cording to the variance decomposition results, however, other macroeconomic
shocks, including shocks to aggregate demand, output supply as well as
shocks to foreign financial flows and trade balance appear to play an im-
portant role in explaining the variation in inflation in Ethiopia.
Keywords: Exchange Rate, Pass-through, Inflation, SVAR, Ethiopia
JEL Classification: F14, F31, F23
*Copenhagen University, Department of Economics. Copenhagen, Denmark; tseday.
mekasha@econ.ku.dk; Tel +45 35 32 37 83
†Stockholm University, Department of Economics. SE-106 91 Stockholm, Sweden; kiflu.
molla@ne.su.se; Tel +46 8 16 32 15

1 Introduction
The degree, speed and duration at which the domestic price level responds
to changes in the exchange rate are crucial factors in understanding inflation
dynamics and determining the effectiveness of monetary policy. In particu-
lar, whether the price level responds proportionately or less proportionately to
changes in the exchange rate and the speed at which these occur is among the
crucial issues in Exchange Rate Pass-Through (ERPT) analysis.1For instance,
if ERPT happens to be quick and complete, any induced nominal deprecia-
tion (devaluation) will immediately be matched by a proportional rise in prices
which will cut short any potential change in the real value of the currency.
In such cases, monetary policy is less likely to be effective in changing the
real exchange rate and enhancing the competitiveness, even in the short run.
On the other hand, in an environment where the rate of pass-through is low
or incomplete, depreciations (devaluations) of the nominal exchange rate are
highly likely to lead to a depreciation of the real exchange rate and hence help
prevent a loss of competitiveness. Therefore, it is important to have a quantita-
tive estimate of ERPT to predict if nominal depreciations (devaluations) lead to
increased domestic inflation or help improve (or at least maintain) the export
competitiveness of a country.
For countries in sub-Saharan Africa (SSA) in general and Ethiopia in partic-
ular, preventing a loss of export competitiveness is a pressing policy challenge
and the more so when faced with inflationary environments. In this regard, the
Ethiopian economy is a case in point where, in recent years, observing double
digit inflation has become the norm rather than the exception.2In such in-
flationary instances, policy makers often resort to currency devaluation mea-
sures so as to improve/maintain the export competitiveness. For example,
following the recent inflationary pressures, National Bank of Ethiopia (NBE)
has taken various devaluation measures, the major one being on September
1The degree of pass-through (complete vs incomplete) to the local currency price of imports
by and large depends on the pricing behavior of foreign exporting firms. If foreign firms follow a
Producer Currency Pricing (PCP) strategy, then the exchange rate change will be fully reflected
in the local currency price of imports (a case of complete pass-through). On the other hand,
firms might pursue a Local Currency Pricing (LCP) strategy where they absorb all (a large part)
of the exchange rate change into their markups. In this case, assuming price rigidity, the local
currency price of imports will not change with movements in the exchange rate (in case of low
or incomplete pass-through). Another popular explanation often used in the literature to
explain the issue of incomplete pass-through, without assuming price rigidity, is the Pricing
To Market (PTM) strategy of foreign firms. That is, firms adjust the price of exports (in their
home currency) by the same proportion as the change in the exchange rate, thus exercising
some degree of price discrimination across destinations. In this case, the local currency price
of imports will not be affected by the change in the exchange rate implying incomplete or low
ERPT.
2The inflation rate reached a peak of 40 percent in quarter three of 2011, remained above
20 percent for most of 2012 before it became a single digit in the second quarter of 2013 (see
Figure C18, see also Geiger and Goh (2012)).
1

2010 when the Ethiopian currency (Birr) was devalued by 20 percent.3
Moreover, both the World Bank and IMF are still calling for a further de-
valuation of the Birr arguing that the overvalued exchange rate is the main
factor dragging the country’s export competitiveness (see IMF (2014), World
Bank (2014)). Although these devaluation measures seem unavoidable in the
face of alarming inflation rates, excess demand for foreign currency and a de-
terioration of the trade balance, it is not clear whether such measures will add
up more to the inflationary pressure rather than helping improve the export
competitiveness.4This is also a contentious issue for other SSA countries and
it begs an empirical exploration.
Against this background, the main objective in this paper is to empirically
analyze the extent of exchange rate induced inflation in Ethiopia. In particular,
this paper makes an effort to answer important questions including how big
is the rate of ERPT, how quickly it leads to inflation and for how long the
effect will last? In addition, for each consecutive year after the exchange rate
shock, an attempt is made to estimate the percentage variation in prices that
is attributable to an exchange rate shock and shocks to other macroeconomic
variables. To get a further insight into the results, the main analysis is further
complemented with disaggregated level evidence where we have estimated the
degree of pass-through for food and non-food prices.
To achieve the aforesaid objectives, the paper employs a Structural Vec-
tor Auto Regressive (SVAR) approach where identification is achieved using a
combination of long-run and short-run restrictions. The long-run restrictions
are derived from a simple small open economy macro model in the spirit of
Clarida and Gali (1994). As discussed below, this is unlike the common prac-
tice in the literature where studies rely on either a single equation method or
a SVAR approach based on short-run restrictions. As presented in Section 3,
combining long-run and short-run restrictions has the merit of addressing the
apparent identification challenge in the ERPT analysis, which arises due to the
simultaneous relationship between price and exchange rate. To the best of our
knowledge, this paper is the first to use a combination of long-run and short-
run restrictions to overcome the identification challenge in the ERPT analysis.
This is one major contribution of the current paper.
3Figure C17 shows the trend in CPI, nominal and real exchange rates. As can be seen from
this figure, nominal depreciations were often associated with increases in the consumer price
index and, as a result, the real exchange rate has been appreciating. This has been particu-
larly the case from around 2002 onwards. In the period before 2002, nominal depreciations
were not accompanied by an increase in the CPI, hence resulted in real depreciations. Sound
macroeconomic policies followed during those periods and/or the fact that the economy was
not booming, as it does in recent years, leaving some resources unemployed can be potential
explanations to this heterogeneity across periods.
4For example, in its exchange rate assessment for the 2013/14 fiscal year, the fund indi-
cated an REER over-valuation of the Birr in the range of 10 to 13 percent. It is further reported
that policy makers in Ethiopia resist IMF’s call for further devaluations due to concerns for
potential feedback effects on inflation IMF (2014).
2

Although the issue of ERPT is one of the extensively researched areas in
international macroeconomics, the existing literature mainly focuses on de-
veloped countries and the evidence in the case of developing countries, par-
ticularly those in SSA, is relatively limited. Among the few existing studies,
using a number of developed and developing countries including some SSA
countries, Choudhri and Hakura (2006) find a positive relationship between
the rate of pass-through and the average level of inflation. This is in line
with the hypothesis suggested by Taylor (2000) that pass-through is low in
low inflation environments. Similarly, for SSA countries, Razafimahefa (2012)
investigates an exchange rate pass-through and its determinants for all SSA
countries and finds an average ERPT rate of about 40 percent and indicates
a considerable heterogeneity in ERPT rates across SSA countries. Apart from
these, most of the remaining few papers on the region are country-level stud-
ies. These include Younger (1992) and Frimpong and Adam (2010) for Ghana,
Mwase (2006) for Tanzania and Melesse (2014) for Ethiopia. The recent paper
by Melesse (2014) studies sectoral consumer price inflation in Ethiopia using
the SVAR approach.
As indicated above, the current paper uses a different identification strat-
egy than the above papers. Specifically, these papers use differing method-
ologies, ranging from a single equation method to a SVAR approach mainly
using short-run restrictions (see also Aron et al. (2014)). Unlike single equa-
tion models, which have been quite common in the ERPT literature, SVAR
modeling has the merit of allowing for endogenous interaction between the
exchange rate and other macroeconomic variables including prices. Although
single equation models can also be informative, their treatment of movements
in the exchange rate as exogenous is problematic. This implicit assumption
of single equation models ignores the possible reverse causation that runs
from prices to the exchange rate or the fact that prices and exchange rates
potentially respond to the same shocks.
For example, as is also discussed in Ito and Sato (2007), in the case of
floating exchange rate regimes, an increase in the domestic price level is highly
likely to lead to nominal depreciations. On the other hand, in the case of fixed,
adjustable pegged or managed floating exchange rate regimes, devaluation
measures are in most cases policy responses to domestic inflation and/or to
the underlying fundamentals of the economy (see also Hossain (2005), Sham-
baugh (2008) and Younger (1992)). In view of this, it is important to treat both
exchange rate and prices as endogenous variables and this warrants the use
of the system of equations approach.
Although the SVAR approach is preferable to the single equation method
for the reason discussed above, identification of the structural shocks is still
a major concern. Papers that employ the SVAR approach in analyzing ERPT
(among others, see Ito and Sato (2008), McCarthy (2000,2007) and the papers
cited above) identify the structural shocks using restrictions on the contempo-
raneous reactions of the endogenous variables to the structural shocks. Fol-
3

lowing the seminal work by Sims (1986), these papers impose restrictions on
the short-run effects of the structural shocks so as to get a recursive ordering
of the variables in the system, that is, by applying a Choleskey decomposi-
tion to identify the structural shocks. Although this approach has generally
been useful, its application in the ERPT analysis is commonly criticized for
imposing short-run restrictions that do not have any strong theoretical sup-
port. Given that monetary policy, prices and exchange rates do potentially
affect each other, the recursive ordering of the variables in the system implied
by the short-run restrictions is hard to justify theoretically.
Cognizant of this problem, this paper mainly relies on theory guided long-
run restrictions (neutrality properties) to recover the structural shocks follow-
ing the Blanchard and Quah (1989) (BQ) approach. Specifically, unlike the
existing literature which is entirely dependent on short-run restrictions, in
this paper identification is achieved based on a combination of short-run and
long-run restrictions. Guided by a simple small open economy model, we put
restrictions on the long-run effects of the structural shocks to the endogenous
variables in the system and we combine this with two additional restrictions
on the contemporaneous relationship between the variables in the system.
This is particularly important in the case of developing countries where it is
relatively difficult to determine the short-run relationship between macro vari-
ables. Apart from its advantage in that it does not entirely rely on theoretically
controversial restrictions imposed on the contemporaneous relationship be-
tween the variables in the model, this approach, as noted earlier, has its merit
in dealing with the simultaneity between domestic prices and movements in
the nominal exchange rate. Further details of this methodology are discussed
in Section 3 of the paper.
In order to derive economically plausible restrictions on the long-run rela-
tionships between the variables in the system, a modified version of the small
open macroeconomic model of Clarida and Gali (1994) is used as a guide-
line. A variant of this model has recently been used by Shambaugh (2008)
and Barhoumi (2009) in the context of the ERPT analysis. Unlike these pa-
pers, however, the model applied in the current paper is modified to include,
among others, the role of external financial flows and shocks to the trade
balance.5These extensions are particularly relevant for developing countries
and the more so for countries in SSA where these factors play a major role
in determining movements of both exchange rates and prices. In the case
of the Ethiopian economy, for example, running a current account deficit is
the norm and external financial flows (development aid and remittance) play
a major role in financing this deficit. For instance, in 2013/14, Ethiopia’s
current account deficit reached 7.1 percent of GDP and earnings from remit-
tances that reach 7 percent of GDP and official transfers of 1.3 Billion USD in
2013/14 help offset the current account deficit (see IMF (2014)). Therefore, in
addition to our identification strategy used in this paper, an explicit inclusion
5We also differ from these papers in how we measure the rate of pass-through.
4

of shocks to the trade balance and external financial flows in an open economy
model is another contribution of the paper.6
Just to give a preview of the results, the response of CPI to a one standard
deviation nominal exchange rate shock is found to be positive and statisti-
cally significant starting from the second quarter and onwards. The response
cumulates over subsequent quarters and leaves the domestic price at a per-
manently higher level. In particular, even if the corresponding estimate of the
rate of exchange rate pass-through to aggregate CPI starts out low, it quickly
increases to more than 80 percent in about four quarters and becomes almost
complete in about two years after the shock. In terms of the relative impor-
tance of shocks, only about 7 percent of the variation in inflation in three
to four years is explained by shocks to the nominal exchange rate. Other
macroeconomic shocks including shocks to BOP, aggregate supply and aggre-
gate demand are found to explain the lion’s share of the variation in inflation,
particularly in longer horizons.
The rest of the paper is structured as follows: While Section 2 presents
the theoretical model, Section 3 discusses the econometric methodology and
identification strategies. Section 4 presents the data and relevant tests. The
estimation results and related discussions are presented in Section 5. Finally,
Section 6 gives the concluding remarks.
2 Theory
This section presents the small open economy model which is used as a
guideline for deriving the long-run identifying restrictions that we apply in
the empirical analysis. In particular, the theoretical model mainly draws on
a stochastic version of the two-country rational expectations open macroe-
conomic model presented in Clarida and Gali (1994). This model was origi-
nally developed by Obstfeld (1985) and has been extensively used to explain
sources of real exchange rate fluctuations. In the context of exchange rate
pass-through, however, the application of this model has been quite limited.
Only recently Shambaugh (2008) and Barhoumi (2009) have used this model
to explicitly study exchange rate pass-through.
As indicated earlier, we modify this model to make it more compatible with
the underlying macroeconomic realities of developing countries in general and
Ethiopia in particular. Specifically, we have added shocks to the trade balance
and external financial flows to this otherwise standard rational expectation
open macroeconomic model. The inclusion of trade balance shocks is impor-
tant as the widening of the current account deficit puts pressure on the level of
6Even if there are extensions of open economy models that include shocks to the oil price,
the commodity price and external flows, we have not come across any paper that apply such
extensions in the context of exchange rate pass-through applications.
5

the exchange rate that policy makers want to maintain. For instance, accord-
ing to IMF (2014), a REER depreciation of about 10 percent is needed in order
to close the current account gap in Ethiopia. Similarly, given that Ethiopia
is highly dependent on external assistance and earnings from remittances,
shocks to external financial flows are also crucial in determining movements
in the exchange rate and other macro variables in the economy. The above
IMF report, for example, points out that the 3.7 Billion USD average remit-
tance inflow during the three-year period 2011/12-2013/14 was higher than
the earnings from exports of goods in the same period. This, together with of-
ficial transfers that reached 1.3 Billion USD in 2013/14, largely supports the
country’s current account deficit. Shocks to these resource inflows will thus
undoubtedly influence the price level and the exchange rate that prevail in the
economy. For this reason, the aforesaid extension is important in the context
of the Ethiopian economy. On top of the above extensions, we have also taken
into account the limited capital mobility (closed capital account) nature of the
Ethiopian economy. For a similar extension in the case of SSA countries with
limited capital mobility, see the paper by Sissoko and Dibooglu (2006). These
authors apply a similar extension to examine the sources of macroeconomic
fluctuations with a particular focus on the exchange rate system in SSA coun-
tries.
In what follows, we first list the necessary equations that comprise the sys-
tem and move on to solving the long-run equilibrium of the model. Thus, we
start by describing the aggregate demand (IS-equation) given in Equation 2.1
below.7In this equation, all variables, except the real interest rate, are ex-
pressed in their log forms. Apart from explicitly showing the autonomous
components of net exports as well as the components of consumption, invest-
ment and government expenditure that are financed by foreign financial flows,
Equation 2.1 represents the standard IS equation. Unless otherwise stated,
all parameters used in this model are assumed to be positive.
yd
t=γ[dt+xt+ft−ρ[it−E(pt+1−pt)] +η(st−pt)] (2.1)
While xtrepresents autonomous net export, ftstands for foreign finan-
cial flows (foreign aid and remittances) used to finance consumption, invest-
ment and government expenditure. On the other hand, dtcaptures an au-
tonomous component of aggregate demand financed through domestic (inter-
nal) resources.
We have expressed the real interest rate in nominal terms using the relation
Rt=it−Πe,t, where itand Πe,tdenote the nominal interest rate and expected
7In Appendix A, we list the different equations that form aggregate demand in order to
show how we have introduced the different shocks. While the aggregate demand equation we
have in the appendix is expressed in levels of the variables, here we have expressed the same
equation using the logarithms of the variables for the sake of convenience in the empirical
model estimation and interpretation. In short, Equation 2.1 is not a log transformation of
Equation A8.
6

inflation, respectively. We have also used the following approximation to define
expected inflation (Πe,t). That is, Πe,t=E∆Pt+1
Pt≈Eln(1+∆Pt+1
Pt)=Eln(Pt+1
Pt)=E(pt+1−pt)
and hence, the real interest rate can be given by it−E(pt+1−pt). Finally, qt=st−pt
is the log of the real exchange rate where the foreign price level is normalized to
one. The parameters γ,ρand η, respectively, capture the Keynesian multiplier,
the response of investment to the real interest rate and the response of net
export to the real exchange rate.
The autonomous (exogenous) part of aggregate demand financed through
domestic resources (dt), autonomous net exports (xt) and foreign financial flows
(ft) is assumed to follow a random walk process as given below (see Equations
2.2-2.4):
dt=dt−1+µd
t(2.2)
xt=x0,t−1+µx
t(2.3)
ft=ft−1+µf
t(2.4)
where µd
tis the aggregate demand shock which can be taken as a fiscal ex-
pansion or shifts in consumption or investment functions. On the other hand,
µx
tand µf
tare assumed to capture shocks to net exports (trade balance) and
foreign financial flows, respectively. In a standard IS equation representation,
µx
tis entered as an aggregate demand shock and is assumed to capture exoge-
nous increases in exports or decreases in imports of the country concerned
and represents a rightward shift in the IS curve.
Turning to the supply-side of the goods market, the output supply equation
in this small open economy model is given by Equation 2.5. As in Clarida and
Gali (1994), the output supply is assumed to be determined by the productive
capacity of the economy ( ˆ
yt). We have also assumed that the productive capac-
ity of the economy follows a random walk process as given in Equation 2.6.
ys
t=ˆ
yt(2.5)
ˆ
yt=ˆ
yt−1+µs
t(2.6)
where µs
tdenotes the aggregate supply shock and it can be considered as a
technology or a labor supply shock.
The market clearing condition in the goods market yd
t=ys
t=ytestablishes
an equilibrium in the goods market and we can thus rewrite the IS equation
as:
7

yt=γ[dt+xt+ft−ρ[it−Et(pt+1−pt)] +η(st−pt)] (2.7)
In the above IS equation, output (yt) is positively related to the real exchange
rate (st−pt)and shocks to domestic demand (dt), net exports (xt) as well as
foreign financial flows ( ft) while it is negatively related to the real interest rate.
Turning to the money market, the demand for money (the real money bal-
ance) is given by the standard money demand equation:
md
t−pt=yt−λit(2.8)
where λis the response of demand for real money balances to the nominal
interest rate.
Money supply (ms
t), on the other hand, is assumed to be determined by the
National Bank of Ethiopia and, for simplicity, it is assumed to follow a random
walk process given in Equation 2.9:
ms
t=ms
t−1+µm
t(2.9)
The market clearing condition in the money market, md
t=ms
t, is given by the
standard LM condition indicated in Equation 2.10.
mt=pt+yt−λit(2.10)
Finally, we introduce a separate equation for the nominal exchange rate,
which is given as a function of the domestic money supply (mt)and a dis-
turbance term (et). The disturbance term captures other determinants of the
nominal exchange rate like foreign prices (foreign money supply) and other
exogenous disturbances to the nominal exchange rate. For a similar way of
defining the nominal exchange rate, see also Wehinger (2000). This defini-
tion of the nominal exchange rate is partly based on the monetary approach
to exchange rate determination which establishes that increases in domestic
money supply lead to an equiproportionate depreciation of the home currency
(see Dornbusch (1980)).8Similarly, Bacchetta and van Wincoop (2005) also
derive the equilibrium nominal exchange rate as the ratio of home to foreign
money supplies (St=Mt/M∗
t). Thus, assuming that the disturbance to the nom-
inal exchange rate (et)follows a random walk process given by Equation 2.12,
the nominal exchange rate equation is given by:
st=mt+et(2.11)
8According to the monetary approach to the exchange rate determination, the nominal
exchange rate is given by st=mt−m∗
t−a(yt−y∗
t)+b(it−i∗
t).
8

et=et−1+µe
t(2.12)
Long-run Equilibrium
As can be noted from above, the standard Uncovered Interest Rate Parity
(UIP) equation is not used to determine the interest rate. Given the underde-
velopment of the capital market in the Ethiopian economy or that the country
has a closed capital account, UIP, which relies on the existence of perfect cap-
ital mobility, is unlikely to hold. Thus, before solving the long-run equilibrium
solutions of this aggregate demand-aggregate supply model, we eliminate the
interest rate variable from the goods market equilibrium (IS) condition using
the relation described in the money market equilibrium (LM) condition. That
is:
yt=γdt+xt+ft−ρ1
λ(pt+yt−mt)−E[pt+1−pt]+η(st−pt)(2.13)
Using the definition of the nominal exchange rate given in Equation 2.11,
we can re-write Equation 2.13 as:
yt=γdt+xt+ft−ρ1
λ(pt+yt−mt)−E[pt+1−pt]+η(mt+et−pt)(2.14)
Re-arranging this, we get a first-order difference equation in the price level
(Equation 2.15) which is used to solve for the long-run equilibrium price.
pt=λ
ρ+ρλ +ηλ "dt+xt+ft−λ+γρ
γλ yt+ρ+ηλ
λmt−ρEpt+1+ηet#(2.15)
Solving this first-order difference equation in price (see Appendix Appendix
Bfor the derivation), the long-run equilibrium price level is given by:
pt=λ
ρ+ηλ[dt+xt+ft]−λ+γρ
ρ+ηλ yt+mt+λη
ρ+ηλet(2.16)
The above equation can also be used to define the demand for equilibrium
real money balances which is given by:
mt−pt=−λ
ρ+ηλ[dt+xt+ft]+λ+γρ
ρ+ηλ yt−λη
ρ+ηλet(2.17)
9

The equilibrium real exchange rate is derived as a rate that, given the ex-
isting patterns of trade and financial flows, is compatible with a financeable
trade deficit and it is given by Equation 2.18. That is, in the long-run, a coun-
try’s current account is said to be sustainable if the sum of the trade deficit
and foreign financial inflows becomes zero.
st−pt=θ
ηyt−1
η(xt+ft)(2.18)
where θand ηin Equation 2.18, respectively, capture the propensity to
import out of disposable income and the response of net exports to the real
exchange rate.
To show the long-run impact of the five structural shocks on five endoge-
nous variables in the system, below we express the long-run solution to the
model using the first difference of the endogenous variables:
∆yt=µs
t(2.19)
∆(st−pt)
| {z }
∆qt
=θ
ηµs
t−1
η(µx
t+µf
t)(2.20)
∆(mt−pt)=−λ
ρ+ηλ[µd
t+µx
t+µf
t]+λ+γρ
ρ+ηλ µs
t−λη
ρ+ηλµe
t(2.21)
∆pt=λ
ρ+ηλ[µd
t+µx
t+µf
t]−λ+γρ
ρ+ηλ µs
t+µm
t+λη
ρ+ηλµe
t(2.22)
∆st=λ
ρ+ηλ 




µd
t−µx
t+µf
t
η





+ θ
η−λ+γρ
ρ+ηλ!µs
t+µm
t+λη
ρ+ηλµe
t(2.23)
Equation 2.19 to Equation 2.23 define the long-run restrictions that we use
to identify the structural shocks in the system. In particular, assuming that
prices are flexible in the long-run, in this model, output is considered to be
supply determined. That is, demand and nominal shocks do not affect the
level of output in the long-run. Thus, only supply shocks are assumed to have
a long-run impact on the level of output.9
On the other hand, the real exchange rate is determined by shocks to output
supply as well as shocks to the balance of payments (µz
t=µx
t+µf
t).10 Moreover,
9Blanchard and Quah (1989) also make a similar simplifying assumption arguing that even
if demand shocks do also affect long-run output, their effect is small relative to that of supply
disturbances.
10A shock to BOP will thus be given by any long run changes in the real exchange rate that
10

in the long-run, the demand for real money balances is affected by all shocks
except shocks to money supply. In this model, shocks to the demand for real
money balances are captured by shocks to aggregate demand. Furthermore,
in the long-run, the domestic price level and the nominal exchange rate are
affected by all shocks in the system. While shocks to domestic prices are
captured by shocks to money supply, shocks to the nominal exchange rate
capture changes in the nominal exchange rate that are not caused by changes
in the other endogenous variables in the system.
3 Empirical Model and Identification
As pointed out in Section 1, the structural shocks in this paper are identified
using a combination of long-run and short-run restrictions. The long-run re-
strictions are done in the spirit of the Blanchard and Quah (1989) approach.
Accordingly, guided by the theory model presented in the previous section,
we put restrictions on the long-run effects of the structural shocks on the
endogenous variables in the system. We combine this with a short-run re-
striction where we put two restrictions on the contemporaneous relationship
between some of the variables in the system. Identification using a combina-
tion of short-run and long-run restrictions is first introduced by Gali (1992).
For a similar approach, see also Alexius and Post (2008). This approach is
important to deal with the simultaneity between domestic prices and move-
ments in the nominal exchange rate. Another advantage of this method is
that we do not need to entirely rely on restrictions imposed on the short-run
(contemporaneous) relationship between the variables in the model.
In this section we thus specify a VAR model with five variables including
output supply (yt), real exchange rate (qt), demand for real money balances
(mt−pt), consumer prices (CPIt) and the nominal exchange rate (st). The cor-
responding five structural shocks to be identified are: supply shocks, shocks
to the balance of payments (that is, shocks to the trade balance and financial
flows), shocks to aggregate demand, shocks to money supply as well as shocks
to the nominal exchange rate.
As can be seen from the previous section, supply shocks are found in
changes in output whereas shocks to the balance of payments are found in
changes in the real exchange rate that are not caused by changes in output
supply. Moreover, shocks to aggregate demand are given by changes in the
demand for real money balances that are not caused by changes in output
supply and the real exchange rate. Similarly, shocks to money supply are
given by changes in domestic prices that are not caused by changes in output
is not caused by changes in output supply. It should be noted that that in this model the long
run real exchange rate is not constant. This in turn implies that, in the long-run, stand pt
would not change by the same amount and the real exchange rate equation can be considered
as a tool enforcing this constraint.
11

supply, demand for real balances or the real exchange rate. Finally, shocks to
the nominal exchange rate are given by changes in the nominal exchange rate
that are not caused by changes in any of the four endogenous variables in the
system.
Let µtbe a vector that defines these structural shocks and ∆Ytbe a vector
of the first difference of the endogenous variables in the VAR model:
∆Yt=(∆yt,∆qt,∆(mt−pt),∆st,∆pt)0
where ∆ytis output supply which is measured by the gross domestic
product in constant prices, ∆qtis the real exchange rate, ∆(mt−pt) is real
money balances, ∆ptis the domestic price level as captured by the consumer
price index or, alternatively, food and non-food prices and ∆stis the nominal
exchange rate.11 All variables are expressed in logs. Since all variables in the
model are found to be non-stationary (as shown in Section 4), we have used
these variables in their first differences to define the VAR model.
To start with the reduced form Vector Autoregressive (VAR) model:
∆Yt=B1∆Yt−1+B2∆Yt−2+. . . +Bp∆Yt−p+utOr B(L)∆Yt=ut(3.1)
where Biare 5×5matrix of coefficients, B(L)=(Ip−B1L−B2L2−. . . −BpLp),Lis
the lag order operator and i=1,2, . . . prepresents the lag order. Equation 3.1
and the corresponding variance-covariance matrix (Σu) of the reduced form
residuals (ut) can be estimated using available information on the endogenous
variables in the system. Let the variance-covariance matrix of the reduced
form residuals be given by:
E(utu0
t)= Σu(3.2)
Assuming that the matrix polynomial Bihas all its roots inside the unit
circle and is invertible, the above reduced form VAR(p) process will have the
following infinite order moving average (Wold) representation:
∆Yt=ut+ Φ1ut−1+ Φ2ut−2+. . . =
∞
X
j=0
Φjut−j= Φ(L)ut(3.3)
where Φ0=I5,Φs=Ps
j=1Φs−jBjfor s=1,2, . . . and Φ(L)=B(L)−1
11An increase in stindicates a depreciation of the Ethiopian birr and vice versa.
12

In its structural moving average form, the above VAR model can be ex-
pressed as:12
∆Yt=C0µt+C1µt−1+C2µt−2+. . . =
∞
X
i=0
Ciµt−i=C(L)µt(3.4)
where C(L)=C0+C1L1+C2L2+. . ..
Using Equations 3.3 and 3.4 and noting that Φ0=I5, we have that:
ut=C(0)µtfor L=0(3.5)
C(L)= Φ(L)C(0) for L=0,1,2. . . (3.6)
Using the relationship between utand µtgiven above (that is, using Equa-
tion 3.2 in Equation 3.5) and the assumption that structural innovations sum-
marized in µtare mutually orthogonal and have a unit variance, we get:
Σu=C(0)E(µtµ0
t)C(0)0=C(0)C(0)0(3.7)
where C(0) is a matrix of contemporaneous effects of the structural shocks
and it models the short-run relationship between the five variables in the sys-
tem. An identification based on short-run restrictions requires putting zero
restrictions on the C(0) matrix. On the other hand, an identification of the
structural shocks using long-run restrictions requires putting restrictions on
the cumulative impulse responses which are given as: C(1)=C0+C1+C2+C3+. . ..
As we have indicated in the introduction, in this paper we use a combination
of short-run and long-run restrictions to achieve identification.
Likewise, denoting the cumulative effect of the reduced form residuals as
Φ(1) = Φ0+ Φ1+ Φ2+ Φ3+. . ., we can define the matrix of cumulative impulse
responses (C(1)) in terms of Φ(1) and C(0). To do this, we use repeated substi-
tution into Equation 3.6 to arrive at Ci= ΦiC0; and summing both sides of this
equation over igives us the key equation for identification:13
C(1)=Φ(1)C(0) (3.8)
The aim is to identify the contemporaneous relationship (C(0)), the struc-
tural system dynamics as defined in (C(1)) and the time-series of structural
12Inverting C(L), in Equation 3.4, we can also write the structural vector autoregressive
(SVAR) representation of the model as A(L)∆Yt=µt.
13For L=0,C0=C0,for L=1,C1= Φ1C0,for L=2,C2= Φ2C0, . . . Ci= ΦiC0.
13

shocks (µt). With these, we can estimate the structural impulse responses
functions (SIRFs) and compute the structural forecast error variance decom-
position (SFEVD). What we need here is therefore to estimate the matrix C(0)
which, as indicated in Equation 3.7, is given by C(0)C(0)0= Σu. Using C(0) to-
gether with Φ(1), which can be computed from the reduced form VAR model,
we can get the cumulative effects of the structural shocks using Equation 3.8.
Since the structural shocks are not observed or one cannot directly recover
the parameters of the structural moving average model in Equation 3.4, the
reduced form representation is instead used to recover the structural shocks.
Specifically, the structural shocks are obtained by transforming the reduced
form residuals using Equations 3.5 and 3.7. However, since the reduced form
model is itself under-identified, we need to impose restrictions in the sys-
tem. That is, the variance covariance matrix of the reduced form residuals in
Equation 3.7 represents 25 equations in 25 unknowns, where, given that Σu
is symmetric, 10 of the equations are redundant. We are therefore left with
15 equations to identify the 25 parameters, which implies that we need 10
additional restrictions to arrive at a just identified system.
To achieve an identification, one can put short-run, long-run or a combina-
tion of short-run and long-run restrictions on the reactions of the endogenous
variables to the structural shocks. Identification based on short-run (as in
Sims (1986)) and long-run restrictions (as in Blanchard and Quah (1989)) re-
quires putting the remaining ten restrictions on the C(0) and C(1)matrices,
respectively. As indicated above, this paper uses a combination of short-run
and long-run restrictions and the specific restrictions we are imposing on the
system are outlined below.
Using the predictions derived from the theory model in the previous section
(see Equations 2.19 to 2.23), we obtained eight long-run restrictions. Adding
two short-run restrictions, which are specified below, we get the 10 restrictions
needed for a just identified system. This amounts to putting zero restrictions
on the C(1)and C(0) matrices shown in Equation 3.8. The long-run restrictions
are:
1. Only supply shocks (µs
t) have long-run effects on output supply;14
2. In the long run, the real exchange rate is determined by shocks to the
balance of payments (µz
t=µx
t+µf
t) as well as shocks to the output supply
(µs
t);
3. Demand for real money balances, in the long-run, is affected by all
shocks in the system (µs
t, µz
t,µd
tand µe
t) other than shocks to money supply
(µm
t).
14One could reasonably argue about adding shocks to oil and/or commodity prices (captur -
ing terms of trade shocks) as this could potentially affect output in the long-run. To limit the
size of the system of equations we are dealing with, we instead add these variables as purely
exogenous shocks to the system in the estimation of the structural VAR model.
14

4. In the long-run, the domestic price level is affected by all shocks in the
system; specifically, it is affected by shocks to output supply, aggregate
demand, the balance of payments as well as shocks to money supply and
the nominal exchange rate;
5. Finally, in the long-run, the nominal exchange rate is also affected by all
shocks in the system.
In addition to the above long-run restrictions, we have also introduced two
additional restrictions on the short-run (contemporaneous) relationship be-
tween the variables. Specifically, we assume that shocks to the nominal ex-
change rate will not have short-run (contemporaneous) effects on domestic
prices. This is consistent with the idea that prices adjust sluggishly to changes
in the nominal exchange rate. Similarly, money supply shocks will not have
short-run (contemporaneous) effects on the nominal exchange rate. Given
Ethiopia’s managed floating exchange rate regime, it is highly likely that the
nominal exchange rate will remain the same within the quarter-of the money
supply shock. We impose these two additional short-run restrictions on the
C(0) matrix and we arrive at a just identified system.
Taking the above long-run restriction into account, Equation 3.4, the long-
run response of the endogenous variables to the structural shocks, with the
expected signs predicted from the theory model, can be summarized in the
following matrix:






















∆yt
∆qt
∆(mt−pt)
∆st
∆pt






















=
C(1)
z }| {






















+0000
+−000
+−−−0
+−+++
−++++












































µs
t
µz
t
µd
t
µe
t
µm
t






















Similarly, given the above restrictions on the C(1)and C(0) matrices, the
relation indicated in Equation 3.8 can be summarized in the following matrix.
C(1)
z }| {






















10000
UR UR 000
UR UR UR UR 0
UR UR UR UR 1
UR UR UR UR 1






















=Φ(1)
C(0)
z }| {






















UR UR UR UR UR
UR UR UR UR UR
UR UR UR UR UR
UR UR UR UR 0
UR UR UR 0UR






















15

where UR refers to unrestricted parameters.
In addition to the ten zero restrictions indicated above, we also have three
identified parameters (indicated by ”1”s in the above matrices) giving us an
overidentified system. However, since the estimation procedure that we are
using requires the system to be just identified, we have not used these iden-
tified parameters for identification. Another point worth noting here is that in
our identification scheme presented above, the respective ordering of price and
exchange rate variables does not change the empirical results. This is unlike
the identification schemes based entirely on short-run or long-run restrictions
and can deal with the simultaneous relationship between price and exchange
rate in a better way.
4 Data and Pre-estimation Tests
The data used in this paper is quarterly data covering the period 1993q1-
2014q4.15 The main data sources are IMF International Financial Statistics
(IFS), World Bank’s World Development Indicators data base and national
data sources including National Bank of Ethiopia and the Central Statistics
Office of Ethiopia. Since data on the nominal effective exchange rate is not
available for the period after 2010q4, we had to use the bilateral nominal
exchange rate which is available until 2014q4. However, we have also checked
the results using the multilateral exchange rate data and the results are not
qualitatively different.
Before we do a formal test on the time series properties of the variables,
we have started with a preliminary assessment through a graphical inspec-
tion of the level and difference of each of the series presented in Appendix
C.16 It is clear from these figures that, except for the real exchange rate, the
levels of all variables (CPI, nominal exchange rate, GDP, real money balances)
appear to have a clear upward trend. The first difference of all variables, on
the other hand, is close to a stationary series. The plots based on the first
difference of the variables do also reveal the presence of some outliers in the
aforementioned series that we take into account in the main analysis.17
Given the subjectivity inherent in the visual inspection, we move on to a
formal testing of the behavior of each series so as to determine whether the
15The years before 1993 are not included as Ethiopia had a socialist regime that practiced
heavy price control and followed a fixed exchange rate regime.
16As is also indicated in Juselius (2006), it is “much easier to detect an outlier observation
in the differences of a non-stationary variable than in the levels.”
17Food and Non-food price indices are also found to be integrated of order one even after
accounting for structural breaks. Trends in the level and first difference of these variables are
given in Figure C16.
16

variables in question are stationary or not. This is done using standard unit
root tests including Augmented Dickey Fuller (ADF), Phillip Perron (PP) and
Kwiatkowski, Phillips, Schmidt and Shin (KPSS) tests. The results from these
tests are reported in Table C4 in the appendix.
As can be seen from Table C4, both the ADF and PP tests indicate that all
variables are non-stationary at levels. Confirming these results, the estimated
test statistics from KPSS indicate that the null of stationarity is rejected in all
cases. As can be seen from Table C5 in the appendix, the first difference of all
variables, according to all the three tests, appear to be stationary, suggesting
that all variables are integrated of order one, I(1).
The main criticism of the above unit root tests is that these tests may fail to
reject the unit root in the presence of a structural break. In particular, if there
is a structural break in the series, the above tests can wrongly suggest unit
root while the series might actually be stationary around a structural break.
To address this concern, we check the time series properties of the variables
by taking the presence of a structural break into account. This is done using
the Clemente et al. (1998) method which endogenously determines the break
dates from the data.
The Clemente et al. (1998) test gives two models: Additive outlier (AO) and
Innovative outlier (IO) models which, respectively, capture sudden and gradual
changes in the mean of the series. As can be seen from the results reported in
Table C6, the null hypothesis of a unit root in each series cannot be rejected
in all cases, even after taking into account the presence of structural breaks
in the series (the critical value of this test associated with a 5 percent level of
significance is equal to -5.49). Therefore, based on the above test results, we
continue our analysis treating all the variables as having a unit root process.18
5 Results and Discussion
In this section, we present the results from the estimation of the structural
VAR model with five variables: Real GDP, real exchange rate, demand for real
money balances, nominal exchange rate and consumer price index or, alter-
natively, its disaggregated components. To control for seasonal variations,
account for outlier observations and structural breaks, we have respectively
included quarterly dummies, impulse dummies as well as a shift dummy in
the structural VAR estimation.19 We have also included world prices of oil
and coffee as purely exogenous variables potentially affecting the endogenous
18We have also checked if there is any co-integration between the variables in the sys-
tem using a Johansen maximum likelihood test for co-integration. The results show one
co-integrating relationship among the variables in question (see Table C7) for the results.
19As can be seen from trends in GDP (Figure C10) and CPI (Figure C12), there is a clear
structural break (around 2003q1) and we have captured this using a shift dummy.
17

variables in the system. Given that Ethiopia is an oil importing country and
heavily dependent on commodity (mainly coffee) exports, these two variables,
capturing terms of trade shocks, are important sources of macroeconomic
shocks.
For the structural VAR estimation, a lag length is selected using the stan-
dard lag selection methods including LR, FPE, AIC, HQIC and SBIC. Except
the LR method which suggests a lag length of four, all the other methods
suggest a lag length of two. Thus, we choose a lag length of two as this is
suggested by most of the tests and as it is also found to be sufficient to clear
the autocorrelation in the disturbances. We have also run tests for autocorre-
lation, normality and VAR stability (the results are attached in the appendix,
see Table C8,Table C9 and Figure C15).
5.1 Response of CPI to Macroeconomic Shocks
One of the main tools in the structural VAR analysis is the impulse response
function which shows the contemporaneous and an over time response of
endogenous variables in the system to one standard deviation increase in the
structural shocks. Since the long-run constraints (restrictions) we have used
for the SVAR estimation are motivated by economic theory, it is possible to
attach a causal interpretation to the resulting Structural Impulse Response
Functions (SIRFs). In each case, the SIRFs are plotted with the 0.16 and
0.84 percentile (confidence) bands. Following recommendations by Sims and
Zha (1999) (see also Estima (2010)), the confidence bands used to test the
statistical significance of the impulse responses are computed using a Monte
Carlo integrations method with 10000 draws.
As indicated above, we estimate the SVAR model in the log first difference of
the variables and hence the impulse response functions indicate the responses
of the growth rates of the variables rather than their levels. Therefore, we base
our analysis on the cumulated impulse response functions which can be in-
terpreted as responses of the logs of the variables in question. In this case,
the estimated impulse response functions will have a percentage interpreta-
tion; where 100 ×, the cumulative impulse response estimates indicate the
percentage change in the endogenous (responding variable) following a one
standard deviation in the shock variable.
As pointed out in the introduction, the main identification challenge in em-
pirically estimating the rate of exchange rate pass-through to domestic prices
is deciding the ordering of the variables in the VAR; specifically the respective
ordering of prices and the nominal exchange rate. In this paper, using an iden-
tification strategy that is based on a combination of short-run and long-run
restrictions, we have managed to make the ordering of these two variables ir-
relevant for the results. Therefore, it is important to emphasize that although
the results reported below are obtained based on an ordering of the variables
18

that puts the nominal exchange rate last in the system, the results do not
change even if we instead order price last in the system.
Response of CPI to other Macroeconomic Shocks in the System
Before turning our focus to the central theme of this paper, that is, exchange
rate induced inflation, we start by investigating the impact of the structural
shocks on the endogenous variables in the system. In doing so our main focus
is the response of domestic prices to other macroeconomic shocks, apart from
the nominal exchange rate, so as to assess the roles these other shocks play
in determining the domestic price level in Ethiopia. In the process, we also
assess if the impulse response results are in line with the predictions of the
theory model presented in this paper and, more generally, of other standard
small open economy macro models.
To start with the supply shock, as can be seen from Figure 1, in the left-
hand panel, the aggregate supply shock, as theoretically expected, has a price
dampening effect. In particular, a one standard deviation positive shock to
aggregate supply appears to have a negative and statistically significant effect
on the domestic price level that converges to -1.9 percent in about nine quar-
ters after the shock. On the other hand, all the remaining variables in the
system are found to respond positively to the supply shock. The increase in
the demand for money following a supply shock is in line with standard eco-
nomic theory (see Equation 2.8). Moreover, the increase (depreciation) of the
nominal exchange rate following a supply shock can be justified under cer-
tain conditions. In particular, as can be seen from Equation 2.23, a positive
supply shock can induce a nominal depreciation if the elasticity of imports to
disposable income (θ) is high. Similarly, the real exchange rate is also found to
depreciate following a positive supply shock. In view of the observed nominal
depreciation coupled with the decline in the domestic price following a supply
shock, the real depreciation is expected.
Moving to shocks to the balance of payments (BOP), which capture shocks
to the trade balance and external financial flows, as can be seen from the
right-hand panel of Figure 1, the BOP shock has a positive and statistically
significant impact on CPI, both on impact and in subsequent quarters. A one
standard deviation BOP shock is found to have a persistent positive impact
on CPI that starts out below 2 percent, on impact and converges to around
3 percent on the eighth quarter after the shock. Since BOP shocks are in a
way capturing aggregate demand shocks that are financed through external
financial flows, their positive impact on the domestic price is in line with a
priori expectation. Apart from affecting domestic prices, BOP shocks are also
found to lead to a decrease in (appreciation of) the nominal and real exchange
rates and to a decrease in the demand for money balances as is also predicted
by the theory. The lower money demand in response to a positive shock to BOP
can plausibly be explained by agents fear of expected inflation. On the other
19

hand, a positive shock to BOP is not found to have a statistically significant
effect on supply.
Similarly, aggregate demand shocks are also found to have a positive im-
pact on the domestic price level as expected. In particular, following a one
standard deviation shock to aggregate demand, the increase in the domestic
price level converges to 1.7 percent in about seven quarters after the shock.
Regarding the impact on other endogenous variables, while a positive shock
to demand does not have a permanent effect on output and the real exchange
rate, the nominal exchange rate is found to respond positively both on impact
and subsequent quarters and all estimates are statistically significant. Given
that both the domestic price level and the nominal exchange rate respond or
adjust to a positive shock to demand, the lack of impact on the real exchange
rate is expected. On the other hand, contrary to the expectation, the impact
of the demand shock on real money demand is not statistically significant in
all horizons.
Finally, the estimated effect of shocks to money supply, which captures the
variation in the price level that is not explained by the above three shocks,
is found to have a positive impact on domestic price as can be seen from
the cumulative impulse responses depicted in Figure 2. In particular, the
effect of a money supply shock on the domestic price level starts out high, at
0.45 percent, on impact, declines in subsequent quarters and converges to
about 0.43 percent nine quarters after the shock. Looking at the impact of the
money supply shock on other variables in the system, shocks to money supply
do not have any impact on output. In the case of the real exchange rate,
however, monetary shocks induce an appreciation of the real exchange rate
until the first quarter after the shock. This is mainly due to the restrictions
we put on the model. In particular, since the nominal exchange rate in our
model is restricted not to respond to a monetary shock in the short run, it is
inevitable for the real exchange rate to fall (appreciate) for a short while, until
the nominal exchange rate adjusts and offsets the rise in the domestic price
induced by money supply shocks.
Overall, as can be noted above, the estimated impulse response functions
in almost all cases make sensible predictions or concur with theoretical pre-
dictions giving credence to the estimated model. In particular, the direction of
the response of the domestic price level to supply, demand, BOP and money
supply shocks is all in line with both the theory model presented in Section 2
and/or is broadly consistent with other open economy macro models. More-
over, in all cases, the impact of the shocks on the domestic price level is statis-
tically significant as can be seen from the confidence bands of the respective
impulse responses in Figure 1 and Figure 2. Below we turn our attention to
answering the main question of this paper, that is, to investigate the extent
of exchange rate induced inflation, or the degree and speed of exchange rate
pass-through to domestic prices, in Ethiopia.
20

Figure 1: Cumulative Impulse Responses to Supply and BOP Shocks
Supply Shock
Output
0 5 10 15 20
0.002
0.004
0.006
0.008
0.010
0.012
0.014
Real ExRate
0 5 10 15 20
0.005
0.010
0.015
0.020
0.025
0.030
0.035
0.040
0.045
Real Money Bal.
0 5 10 15 20
0.00
0.01
0.02
0.03
0.04
0.05
0.06
CPI
0 5 10 15 20
-0.035
-0.030
-0.025
-0.020
-0.015
-0.010
-0.005
0.000
Nom. ExRate
0 5 10 15 20
-0.0025
0.0000
0.0025
0.0050
0.0075
0.0100
0.0125
0.0150
0.0175
0.0200
BOP Shock
Output
0 5 10 15 20
-0.0015
-0.0010
-0.0005
0.0000
0.0005
0.0010
0.0015
0.0020
Real ExRate
0 5 10 15 20
-0.050
-0.045
-0.040
-0.035
-0.030
-0.025
-0.020
-0.015
-0.010
Real Money Bal.
0 5 10 15 20
-0.055
-0.050
-0.045
-0.040
-0.035
-0.030
-0.025
-0.020
-0.015
-0.010
CPI
0 5 10 15 20
0.010
0.015
0.020
0.025
0.030
0.035
0.040
0.045
Nom. ExRate
0 5 10 15 20
-0.0175
-0.0150
-0.0125
-0.0100
-0.0075
-0.0050
-0.0025
0.0000
0.0025
Evidence on Exchange Rate Induced Inflation
Figure 4 summarizes the cumulative structural impulse responses of the do-
mestic price level to the different macroeconomic shocks in the system. Our
focus in this section is on the impact of shocks to the nominal exchange rate
on the domestic price level. The response of CPI following a shock (a depreci-
ation) in the nominal exchange rate can be seen from the second plot on the
right-hand side of Figure 4.
On impact, the structural impulse response of CPI following a nominal ex-
change rate shock appears to be zero. This is expected as we have restricted
the contemporaneous impact of a shock to the nominal exchange rate on CPI
to be zero. In the first quarter after the nominal exchange rate shock, the im-
pulse response of CPI becomes positive, albeit statistically insignificant. From
the second quarter onwards, however, the estimated impulse response of CPI
to shocks to the nominal exchange rate appears to be positive and statistically
significant.
The estimated response of the domestic price to a nominal exchange rate
shock cumulates over periods leaving it at a permanently higher level. Specif-
ically, as can be noted from Figure 4, a one standard deviation shock to the
nominal exchange rate is associated with a 0.60 percent and a 1.0 percent
increase in the price level in the second and fifth quarters after the exchange
21

Figure 2: Cumulative Impulse Responses to Aggregate Demand and Money Supply
Shocks
Demand Shock
Output
0 5 10 15 20
-0.0030
-0.0025
-0.0020
-0.0015
-0.0010
-0.0005
0.0000
0.0005
Real ExRate
0 5 10 15 20
-0.0100
-0.0075
-0.0050
-0.0025
0.0000
0.0025
0.0050
Real Money Bal.
0 5 10 15 20
-0.020
-0.015
-0.010
-0.005
0.000
0.005
0.010
0.015
0.020
CPI
0 5 10 15 20
0.0050
0.0075
0.0100
0.0125
0.0150
0.0175
0.0200
0.0225
0.0250
Nom. ExRate
0 5 10 15 20
0.0025
0.0050
0.0075
0.0100
0.0125
0.0150
0.0175
0.0200
0.0225
0.0250
Money SS Shock
Output
0 5 10 15 20
-0.0020
-0.0015
-0.0010
-0.0005
0.0000
0.0005
0.0010
0.0015
Real ExRate
0 5 10 15 20
-0.010
-0.008
-0.006
-0.004
-0.002
0.000
0.002
0.004
Real Money Bal.
0 5 10 15 20
-0.010
-0.008
-0.006
-0.004
-0.002
0.000
0.002
0.004
CPI
0 5 10 15 20
0.000
0.001
0.002
0.003
0.004
0.005
0.006
0.007
0.008
0.009
Nom. ExRate
0 5 10 15 20
-0.002
0.000
0.002
0.004
0.006
0.008
0.010
0.012
rate shock, respectively. The price response converges to 1.3 percent in about
three years after the exchange rate shock.20
To give the above results a more meaningful interpretation, below we have
calculated the rate of exchange rate pass-through (ERPT) to the domestic price
level as the percentage change in the price level that is due to a one percent
change in the nominal exchange rate. To this end, we follow the standard
practice in the literature and divide the cumulative response of CPI to a one
standard deviation exchange rate shock by the respective cumulative response
of the nominal exchange rate to the same shock. The estimated rates of ERPT
that we have calculated in this way for each quarter after the exchange rate
shock along with its confidence bound are depicted in Figure 5.
As can be noted from Figure 5, ERPT to domestic prices in Ethiopia starts
low (about 17 percent in the first quarter after the exchange rate shock) and
grows quickly to 53 percent in the second quarter after the shock. About 60
percent of the price change already takes place in the second quarter and in
about two years (eight quarters) after the exchange rate shock, the rate of
ERPT reaches 90 percent and it later converges to 97 percent. Although ERPT
is known to be larger for developing countries in general, the estimate that
20For the sake of comparison, we have also estimated the cumulative response of CPI fol-
lowing a nominal exchange rate shock using short run restriction applying Cholesky decom-
position. The result is qualitatively similar and it is reported in Figure D19.
22

Figure 3: Cumulative Impulse Responses to Nominal Exchange Rate Shocks
Nom. ExRate Shock
Output
0 5 10 15 20
-0.0015
-0.0010
-0.0005
0.0000
0.0005
0.0010
0.0015
Real ExRate
0 5 10 15 20
-0.002
0.000
0.002
0.004
0.006
0.008
0.010
0.012
Real Money Bal.
0 5 10 15 20
-0.035
-0.030
-0.025
-0.020
-0.015
-0.010
-0.005
0.000
0.005
CPI
0 5 10 15 20
-0.005
0.000
0.005
0.010
0.015
0.020
0.025
Nom. ExRate
0 5 10 15 20
0.0025
0.0050
0.0075
0.0100
0.0125
0.0150
0.0175
0.0200
0.0225
Figure 4: Cumulative Response of CPI to Nominal Exchange Rate and other Shocks
Response of CPI to:
Supply Shock
0 5 10 15 20
-0.035
-0.030
-0.025
-0.020
-0.015
-0.010
-0.005
0.000
BOP Shock
0 5 10 15 20
0.010
0.015
0.020
0.025
0.030
0.035
0.040
0.045
Demand Shock
0 5 10 15 20
0.0050
0.0075
0.0100
0.0125
0.0150
0.0175
0.0200
0.0225
0.0250
Money SS Shock
0 5 10 15 20
0.000
0.001
0.002
0.003
0.004
0.005
0.006
0.007
0.008
0.009
Nom. ExRate Shock
0 5 10 15 20
-0.005
0.000
0.005
0.010
0.015
0.020
0.025
23

Figure 5: Estimated Rates Exchange Rate Pass-Through
Exchange Rate Pass-Through to CPI
Quarters
Rate of Pass-Through
0 5 10 15 20
-0.25
0.00
0.25
0.50
0.75
1.00
we find here for the Ethiopian economy can be considered to be even larger,
the more so for ERPT to CPI. Different factors, including the persistence of
the exchange rate in Ethiopia, the country’s heavy reliance on imports as well
as the inflationary environment in the country, can potentially explain the
high exchange rate pass-through to the domestic price observed for Ethiopia.
Together with the potentially low importance of the Ethiopian economy to for-
eign exporters, the high rate of pass-through estimated in this paper is in line
with the predictions of producer currency pricing models (among others, see
McCarthy (2007) and Taylor (2000)).
In sum, the above evidence from the impulse response shows that exchange
rate pass-through to consumer prices in Ethiopia, although it starts out at a
lower level, is found to increase in subsequent quarters and becomes almost
complete two years after the exchange rate shock; leaving the domestic price
at a permanently higher level.
Apart from shocks to the nominal exchange rate, we have also seen that
the domestic price level responds positively to shocks to BOP, aggregate de-
mand and money supply (see Figure 4). In particular, while the domestic price
level responds negatively to aggregate supply shocks, the price level responds
positively to BOP, demand and money supply shocks. This shows that other
macroeconomic shocks are also important in defining movements of CPI in
Ethiopia.
To get a better insight into how the different macroeconomic shocks con-
tribute to variations in domestic inflation (price level) in Ethiopia, below we
present results on the variance decomposition of CPI.
24

Table 1: Decomposition of Variance for Consumer Price Index
Supply BOP Demand Money SS Nom. ExRate
Horizon Shock Shock Shock Shock Shock
1 7.3 59.2 21.9 5.2 0.0
[ 0.8,22.7] [37.4,77.7] [ 8.7,38.6] [ 1.0,13.0] [ 0.0, 0.0]
2 10.1 59.2 19.2 4.6 0.3
[ 1.4,27.7] [37.8,77.2] [ 7.7,34.6] [ 1.1,11.1] [ 0.0, 1.1]
3 10.0 60.1 18.4 2.8 1.8
[ 1.4,28.4] [39.0,77.4] [ 7.3,33.9] [ 0.8, 7.2] [ 0.4, 4.5]
4 10.4 59.9 17.3 2.0 3.0
[ 1.4,29.8] [39.0,77.2] [ 6.6,32.7] [ 0.6, 5.4] [ 0.6, 7.2]
8 11.5 55.1 16.7 1.5 6.1
[ 1.6,33.8] [34.8,73.1] [ 4.9,33.4] [ 0.4, 4.2] [ 1.2,13.8]
12 12.2 52.6 16.3 1.4 7.3
[ 1.6,35.7] [32.3,71.0] [ 4.2,34.0] [ 0.4, 4.0] [ 1.4,16.4]
16 12.6 51.4 16.0 1.3 7.9
[ 1.7,36.9] [30.8,70.2] [ 3.8,34.4] [ 0.3, 3.9] [ 1.5,17.6]
5.2 Decomposing the Variation in Consumer Price
A large exchange rate pass-through estimate does not necessarily mean that
depreciation (devaluation) of the nominal exchange rate is the only or the most
important source of variation in inflation in Ethiopia. Despite observing a very
large exchange rate pass-through, the exchange rate shock might have a small
contribution in explaining the variation in inflation if the size of the shock is
small. Determining the percentage variation in the domestic price level that
can be attributed to shocks to the nominal exchange rate as well as shocks
to other macroeconomic variables in the system is therefore the next relevant
exercise. Table 1 presents the structural forecast error variance decomposi-
tion (SFEVD) which shows the relative importance of each structural shock in
explaining variations in growth rates of consumer prices in Ethiopia.21
To start with our variable of interest, as can be seen from Table 1, in the
first quarter, shocks to the nominal exchange rate are found to play no role in
explaining the variance in inflation. This is expected given the identification
restriction we imposed on the contemporaneous relationship between these
two variables. However, even in the second and third quarters, shocks to the
21Since we have estimated the model in log differences, it should be noted that we are de-
composing the variance in the growth rates of the endogenous variables, and not the variance
in the levels of the variables. Moreover, since we have reported the median SFEVDs, the
decompositions attributable to the different shocks do not always add up to 100%.
25

nominal exchange rate tend to explain only 0.3 and 1.8 percent of the variation
in inflation, respectively. The percentage share of the variance in inflation that
is attributable to changes in the nominal exchange rate appears to increase
in subsequent quarters. In particular, after 12 quarters (three years) and 16
quarters (four years), 7.3 and 7.9 percent of the variation in the growth rate of
consumer prices in Ethiopia are explained by shocks to the nominal exchange
rate, respectively. From the fourth quarter onwards, shocks to the nominal
exchange rate appear to explain a larger percentage of the variation in inflation
than do money supply shocks.
As can be seen above, despite the high degree of pass-through, the nominal
exchange rate shock appears to only determine a smaller percentage of the
variation in inflation. This can partly be explained by the fact that most of the
observed changes in the nominal exchange rate (NER) of the Ethiopian birr
are very small depreciations of the birr; although large, devaluations are rare.
Another reason can be due to the fact that NER shocks are identified in the
system through changes in NER that are not explained by other shocks in the
system.
Nominal exchange rate shocks do play an important role in determining
the variance in the growth rate of demand for real money balances as well as
the nominal exchange rate itself. Over the horizons, about 37 to 23 percent
of the variance in the growth rate of demand for real money balances is at-
tributable to nominal exchange rate shocks. Given that the interest rate in
the Ethiopian economy is only partially liberalized and that it is not regularly
adjusted, expected inflation, exchange rates and foreign financial inflows play
an important role in explaining the variation in the demand for money (see
also Loening et al. (2009)).
The variation in inflation that is due to own shock (money supply shock) is
found to be relatively small, especially at longer steps. In particular, money
supply shocks do only contribute to 5.2 and 4.6 percent of the variation in
inflation in the first and second quarters, respectively. The contribution of
money supply shocks tends to decrease over the horizons and in the fourth
year, only 1.3 percent of the variation in inflation in Ethiopia is due to inno-
vations of money supply. The relatively smaller contribution of money supply
shocks in determining the variance in inflation can be expected, given that
money supply shocks are identified through changes in the price level that
are not caused by the other shocks to the system. In general, money supply
shocks do not appear to play any significant role in determining the variation
in any of the endogenous variables in the system.
From Table 1, it is also clear that, at shorter steps, most of the one-step
forecast error variance in domestic inflation is explained by shocks to BOP,
which are identified through changes in the real exchange rate that are not
explained by output supply shocks. In particular, BOP shocks explain 59 and
60 percent of the variation in CPI in the first and fourth quarters, respectively.
26

Afterwards, the contribution of innovations in the BOP in explaining the vari-
ation in inflation appears to decrease slightly and yet BOP shocks remain the
prime driver of variation in inflation. After a three-year period, 53 percent of
the variation in inflation are attributable to BOP shocks. Although this seems
to be anticipated, since BOP shocks are part of aggregate demand shocks that
are associated with the trade balance and foreign financial flows, both of which
are important in the Ethiopian economy, this result should not come as a sur-
prise. BOP shocks also play an important role in explaining variations in the
growth rates of the real exchange rate (47 to 72 percent), demand for money
(42 to 40) and the nominal exchange rate (2.4 to 8.2 percent) (see Appendix
D).
Similarly, aggregate demand shocks, identified through changes in the de-
mand for real money balances that are not explained by supply or BOP shocks,
are also found to play an important role in determining the variation in in-
flation. Specifically, shocks to aggregate demand contribute to 21 and 17.3
percent of the variation in inflation in the first and fourth quarters, respec-
tively. This makes the aggregate demand shock the second most important
shock, deriving variations in CPI in the short run. Over the horizons, the
contribution of shocks to aggregate demand declines only slightly where, even
after three years (12 quarters), 16.3 percent of the variation in inflation are
attributable to aggregate demand shocks. The important role of aggregate de-
mand shocks in explaining variations in inflation is expected, given that these
shocks are mainly meant to capture expansionary fiscal policies. Aggregate
demand shocks are also found to explain a sizable share of the variation in
the growth rate of the nominal exchange rate, where about 50 to 35 percent of
the variance in the growth rate of the nominal exchange rate are found to be
attributable to aggregate demand shocks.
Moreover, shocks to output supply also appear to have a major role in de-
termining the variation in consumer prices, especially at longer steps. As can
be seen from Table 1, even if the importance of shocks to output supply is
relatively small in the short-run, its contribution increases over time. In par-
ticular, in the second year and afterwards, shocks to output supply become
the third most important factor contributing to about 12 and 13 percent of the
variation in inflation in the third and fourth years, respectively.
Overall, even if BOP shocks, shocks to demand and aggregate supply are
found to be the major deriving forces determining variations in the growth
rate of domestic prices in Ethiopia, the importance of shocks to the nominal
exchange rate cannot be ignored. In the third year and afterwards, more than
7 percent of the variation in inflation are explained by shocks to the nominal
exchange rate. This, together with the strong cumulative impulse response of
CPI following shocks to the nominal exchange rate, has its own implication
for the effectiveness of monetary policy. That is, impacts of changes in
the nominal exchange rate (devaluation measures) on real variables will be
lower than anticipated. As can be seen from Table D12, in the first and
27

second quarters, 11.3 to 4.3 percent of the variance in the growth rate of
the real exchange rate are attributable to nominal exchange rate shocks.
In view of this, the devaluation measures aimed at improving (maintaining)
the country’s export competitiveness may not be able to achieve the desired
targets, since such measures will be partly matched by a rise in the domestic
price level.
We have presented the variance decompositions of the other endogenous
variables in Appendix D. As is theoretically expected, the variance in the
growth rate of output supply is mainly determined by its own shock and
the more so at the longer horizon where the other shocks do not have any
meaningful contribution. Specifically, over the long horizon, 98 percent of the
variation in output supply are attributable to a shock to output supply itself.
Over shorter horizons, on the other hand, shocks to BOP as well as shocks to
the nominal exchange rate and money supply play some role in determining
variations in the growth rate of output (see Table D11). While most (71.5 to 47
percent) of the variation in the growth rate of the real exchange rate is due to
own (BOP) shocks, 23 to 26 percent of the variation are attributable to output
supply shocks. Money supply and nominal exchange rate shocks, on the other
hand, appear to have some role in the short run (see Table D12). As indicated
above, shocks to BOP (41 to 43 percent) and the nominal exchange rate (23
to 37 percent) have important contributions towards explaining the variance
in the growth rate of demand for real money balances. At longer steps, sup-
ply shocks are also important in explaining the variance in the growth rate of
demand for real money balances; after the second year, more than 23 percent
of the variation are attributable to output supply shocks (see Table D13). The
variance in the growth rate of the nominal exchange rate can be attributed to
the different shocks in the system, no one shock is responsible for the majority
of the variance in the growth rate of the nominal exchange rate. However, over
the horizons, aggregate demand shocks (35 percent to 50 percent) and nomi-
nal exchange rate shocks (24 percent 25 percent) appear to have an important
role (see Table D14).
5.3 Impulse Response of Disaggregated CPI: Food Price and
Non-Food Price Indices
In this section, we look at the response of disaggregated price indices to
changes in the nominal exchange rate. Specifically, we estimate the structural
impulse responses of food and non-food price indices to the nominal exchange
rate as well as other shocks in the system. To this end, we employ the same
model as above except that now we had to limit the sample period between
from 1997q4 to 2014q4 because of lack of data on the food and non-food price
indices for the period prior to 1997.
28

As can be seen from Table C10, food items constitute the larger share of the
consumer basket. As can also be seen from Figure C16, which shows trends
in aggregate, food and non-food price indices over the sample period, trends
in the food price index closely mimic trends in the aggregate price index. For
these reasons, we expect food prices to respond to nominal exchange rate
shocks in a similar way as aggregate consumer prices and this appears to be
the case as can be seen from the impulse response estimates presented in
Figure 6.
Figure 6: Response of Food Price to a Nominal Exchange Rate Shocks:
Response of Food PI to:
Supply Shock
0 5 10 15 20
-0.08
-0.07
-0.06
-0.05
-0.04
-0.03
-0.02
-0.01
0.00
0.01
BOP Shock
0 5 10 15 20
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
Demand Shock
0 5 10 15 20
0.000
0.005
0.010
0.015
0.020
0.025
0.030
0.035
Money SS Shock
0 5 10 15 20
0.004
0.006
0.008
0.010
0.012
0.014
0.016
0.018
Nom. ExRate Shock
0 5 10 15 20
-0.005
0.000
0.005
0.010
0.015
0.020
0.025
0.030
As can be seen from the cumulative structural impulse response estimates
depicted in Figure 6, the response of food prices to changes in the nominal
exchange rate is positive and statistically significant in the second and sub-
sequent quarters. As in the case of the aggregate price index, a one standard
deviation shock to the nominal exchange rate leaves food prices at a perma-
nently higher level. For each quarter after the exchange rate shock, Figure 7
plots the corresponding exchange rate pass-through rates to food prices which
indicate that a pass-through to the food price in Ethiopia becomes almost com-
plete in about five to six quarters after the exchange rate shock. The observed
high rate of ERPT to food prices cannot only be explained by the relatively high
import content of the food price index, but also by the fact that the production
and distribution of food items involve imported items.
Similarly, the cumulative structural impulse response of non-food prices
29

Figure 7: Estimated Rates Exchange Rate Pass-Through to Food Price Index
Exchange Rate Pass-Through to Food Price Index
Quarters
Rate of Pass-Through
0 5 10 15 20
-0.50
-0.25
0.00
0.25
0.50
0.75
1.00
1.25
to a one standard deviation nominal exchange rate shock is also positive and
statistically significant, again starting from the second quarter onwards. Com-
pared to the above two cases, the impulse response of non-food prices appears
to be somewhat lower (see Figure 8). As can also be seen from Figure 9, the
estimated exchange rate pass-through rate to non-food prices is found to be
relatively lower and it converges to about 47 percent in about 12 quarters after
the exchange rate shock. Given that non-food items can largely be imported,
the smaller ERPT observed in the case of the non-food price index may seem
strange. However, a closer look at the composition of the non-food price in-
dex indicates that it is composed of non-tradable items (like services) which
are less likely to be affected by changes in the exchange rate. Moreover, the
fact that the non-food price index also includes items whose prices are highly
regulated by the government (e.g. water, electricity and communication) can
be another reason for the observed smaller rate of ERPT to the non-food price
index.
Now we move to the decomposition of variances for growth rates of food
and non-food price indices. As we can see from Table 2, in the first quar-
ter, the lion’s share of the variation in food inflation is explained by shocks
to the balance of payments (BOP) which determine about 60 to 62 percent
of the variance. However, while the contribution of BOP shocks declines in
subsequent quarters, shocks to aggregate supply gained importance at longer
horizons. In three to four years, about 18 percent of the variation in food
inflation are attributable to shocks to output supply. While shocks to aggre-
gate demand contribute to about 15 (on impact) and 7 percent (in about three
30

Figure 8: Response of Non-Food Price to Shocks from the Nominal Exchange Rate:
Response of Non-Food PI to:
Supply Shock
0 5 10 15 20
-0.025
-0.020
-0.015
-0.010
-0.005
0.000
0.005
0.010
BOP Shock
0 5 10 15 20
-0.015
-0.010
-0.005
0.000
0.005
0.010
0.015
Demand Shock
0 5 10 15 20
-0.005
0.000
0.005
0.010
0.015
0.020
0.025
Money SS Shock
0 5 10 15 20
0.006
0.008
0.010
0.012
0.014
0.016
0.018
0.020
Nom. ExRate Shock
0 5 10 15 20
-0.005
0.000
0.005
0.010
0.015
0.020
0.025
Figure 9: Estimated Rates Exchange Rate Pass-Through to Non-Food Price Index
Exchange Rate Pass-Through to Non-Food Price Index
Quarters
Rate of Pass-Through
0 5 10 15 20
-0.3
-0.2
-0.1
0.0
0.1
0.2
0.3
0.4
0.5
0.6
31

Table 2: Decomposition of Variance for Food Price Index
Supply BOP Demand Money SS Nom. ExRate
Horizon Shock Shock Shock Shock Shock
1 8.4 59.5 15.0 9.5 0.0
[ 0.8,28.5] [35.1,77.3] [ 4.9,31.7] [ 3.3,19.2] [ 0.0, 0.0]
2 14.6 59.4 13.4 5.1 0.3
[ 2.3,39.2] [33.7,78.9] [ 4.3,28.3] [ 1.6,11.3] [ 0.0, 1.1]
3 16.2 59.6 12.2 3.7 1.2
[ 2.5,42.7] [33.9,78.8] [ 4.0,26.0] [ 1.2, 8.1] [ 0.3, 3.2]
4 16.3 60.5 10.7 3.2 1.9
[ 2.7,44.0] [35.1,79.4] [ 3.5,23.3] [ 1.1, 6.7] [ 0.4, 4.7]
8 17.0 61.7 7.7 2.3 3.2
[ 2.9,47.5] [35.4,79.9] [ 2.2,18.8] [ 0.8, 5.1] [ 0.7, 7.8]
12 17.6 61.7 6.5 2.0 3.5
[ 2.8,49.5] [34.9,80.4] [ 1.6,17.3] [ 0.7, 4.6] [ 0.7, 8.8]
16 17.9 61.7 6.0 1.8 3.6
[ 2.8,50.9] [34.5,80.8] [ 1.3,16.6] [ 0.6, 4.3] [ 0.7, 9.2]
years), shocks to the nominal exchange rate and money supply account for a
smaller percentage of the variation in the growth rate of food prices. Although
money supply shocks explain 9.5 and 5.1 percent of the variance in food infla-
tion in the first and second quarters, its contribution declines in subsequent
quarters reaching only 2 percent in the third year. After the first quarter, nom-
inal exchange rate shocks have a small but stable contribution to the variation
in food prices. In the third year, shocks to the nominal exchange rate account
for about 3.5 percent of the variation in food inflation.
In the case of non-food prices, however, shocks to money supply appear
to explain a larger share of the variation in non-food inflation, 53 percent on
impact to 26 percent in the third year. Aggregate demand (17 to 19 percent)
and aggregate supply shocks (13 to 12 percent) also play an important role in
determining the variation in the growth rate of non-food prices (see Table 3).
Even if the nominal exchange rate plays a smaller role at shorter horizons, its
contribution has gradually increased, determining more than 12 percent of
the variation in non-food price inflation after 3 years.
32

Table 3: Decomposition of Variance for Non-Food Price Index
Supply BOP Demand Money SS Nom. ExRate
Horizon Shock Shock Shock Shock Shock
1 13.4 5.0 16.9 52.6 0.0
[ 2.0,34.1] [ 0.5,19.1] [ 4.3,42.1] [23.6,76.4] [ 0.0, 0.0]
2 12.2 4.5 18.4 51.1 0.5
[ 1.9,33.6] [ 0.8,17.0] [ 4.3,43.9] [23.6,75.5] [ 0.0, 2.2]
3 11.8 4.7 20.6 45.3 2.4
[ 2.4,34.0] [ 1.1,16.3] [ 5.2,46.1] [20.4,69.8] [ 0.3, 8.1]
4 11.7 5.1 21.7 40.2 4.6
[ 2.7,34.2] [ 1.2,16.4] [ 5.5,47.5] [17.3,64.4] [ 0.6,14.5]
8 11.6 5.5 20.8 30.4 10.2
[ 2.9,35.4] [ 1.3,17.9] [ 4.8,46.3] [10.7,55.1] [ 1.4,28.5]
12 11.6 5.6 19.7 26.3 12.4
[ 2.7,37.7] [ 1.2,19.6] [ 4.0,45.8] [ 8.2,51.3] [ 1.7,33.9]
16 11.5 5.7 18.8 24.2 13.4
[ 2.4,39.8] [ 1.2,20.5] [ 3.5,45.7] [ 6.6,49.5] [ 1.8,36.4]
6 Conclusion
In this paper, we examine the rate of exchange rate pass-through to domes-
tic inflation with a particular focus on the Ethiopian economy. Despite the
vast literature on exchange rate pass-through, the existing literature mainly
focuses on developed countries and the evidence for developing countries, par-
ticularly for those in SSA, is quite limited. This paper thus aims at contribut-
ing to the limited evidence regarding the impact of exchange rate shocks to
prices in sub Saharan Africa (SSA) by considering the case of Ethiopia. To this
end, we employ a Structural Vector Autoregressive (SVAR) model where identi-
fication of the structural shocks is achieved using a combination of short-run
and long-run restrictions. In the case of the latter, the restrictions are derived
from a small open economy macro model.
In the theory model, which we use as a guideline in specifying the long-run
restrictions needed for identification, an effort is made to reflect the existing
realities of SSA countries in general and Ethiopia in particular. Specifically,
we modified this standard open macroeconomic model to take into account the
fact that the Ethiopian economy has a closed capital account (has limited/no
private capital inflows) and is heavily dependent on external financial flows
(foreign aid and remittances) to finance its norm trade deficit.
The results from estimated cumulative structural impulse response func-
33

tions are starkly consistent with the theoretical predictions, giving us confi-
dence in the model and the identification strategy we have employed in this
paper. The impulse response of domestic prices to a one standard devia-
tion nominal exchange rate shock (depreciation/devaluation) is found to be
positive and statistically significant, which leaves domestic prices at a perma-
nently higher level in about two years. The corresponding rate of exchange
rate pass-through to aggregate CPI starts out low but quickly increases to
more than 80 percent in about four quarters and approaches 96 percent in
approximately a two-year period. While impulse responses of food prices fol-
low the same pattern as aggregate CPI, the exchange rate pass-through to
non-food prices is found to be relatively lower, approaching 47 percent. The
higher exchange rate pass-through to the domestic price that we found for
Ethiopia is consistent with predictions of producer currency pricing models.
Given the potentially low importance of the Ethiopian economy to foreign ex-
porters, this is not a very difficult hypothesis to accept. Moreover, the fact that
the Ethiopian economy is characterized by a persistent exchange rate, a high
import dependence and an inflationary environment can also explain the large
exchange rate pass-through estimate we find in this paper.
Despite the large exchange rate pass-through estimate, shocks to the nom-
inal exchange rate are found to account for only about 7 to 8 percent of the
variation in inflation in three to four years. In longer horizons, a larger pro-
portion of the variation in domestic inflation is instead found to be explained
by shocks to BOP, aggregate supply and aggregate demand, in their respec-
tive order. On the other hand, shocks to money supply are found to play a
very limited role in explaining the variation in the general inflation, which is
contrary to our expectation. While this remains true in the case of variance
decompositions of food inflation, the variance decomposition of non-food infla-
tion reveals a totally different picture. In the case of non-food prices, shocks to
money supply explain a larger percentage of the variation in non-food inflation
in Ethiopia.
Overall, the results have their own implication for the effectiveness of mon-
etary policy. In particular, the high rate of ERPT found in Ethiopia shows that
impacts of changes in the nominal exchange rate (devaluation measures) on
real variables will be lower than anticipated. In view of this, the devaluation
measures aimed at improving (maintaining) the country’s export competitive-
ness may not be able to achieve the desired targets, since such measures will
largely be matched by a rise in the domestic price level. This is also reflected
in the actual over time trends of CPI, the nominal and real exchange rate
depicted in Figure C17.
Therefore, even if devaluation measures are sometimes unavoidable, par-
ticularly in order to maintain competitiveness in the face of inflationary pres-
sures or excess demand in foreign currency, taking such measures per se
cannot be successful in achieving the desired target unless accompanied by
other macroeconomic measures that can put inflation in check.
34

Appendix A Components of Aggregate Demand
Below, we define components of the aggregate demand equation in order to
show how the two shocks (shocks associated with trade balance and external
financial flows) are introduced in the standard IS equation:
Ct=C0,t+σYt+FC
t(A1)
It=I0,t−ρRt+FI
t(A2)
Gt=G0,t+FG
t(A3)
EXt=EX0,t+ηxQt(A4)
IMt=IM0,t−ηmQt+θYt(A5)
Defining autonomous net exports as X0,t=EX0,t−IM0,tand ηx+ηm=η, we
can define the current account as:
CAt=ηQt−θYd
t+X0,t
| {z }
Net exports=NX
+Ft= ∆Rest(A6)
where: Ft=FC
t+FI
t+FG
tand the respective components FC
t,FI
t,FG
tshow
the amount (in Ethiopian Birr) of external financial flows used to finance
consumption, investment and government expenditure. Moreover, σand ρin
Equations A1 and A2, respectively, refer to the propensity to consume out
of disposable income and the response of investment to changes in the real
interest rate R. On the other hand, ηand θin Equation Equation A6 capture
the response of the trade balance to movements in the real exchange rate (Qt)
and the propensity to import from disposable income, respectively.22 Lastly,
X0,tand Ftin Equation A6, respectively, refer to the autonomous level of net
exports (NX) and the exogenous financial flows needed to finance the trade
deficit. Since Ethiopia has a closed capital account, the trade deficit is by
and large financed through foreign currency earnings from foreign aid and
remittances. In the short to medium run, this will be equal to a change in
the country’s foreign reserves ∆Rest. That is, when foreign financial flows Ft
22The real exchange rate is given by Q=StP∗
t
Ptwhere Stis the nominal exchange rate here
defined as the price of the foreign currency (USD) in terms of the local currency (Birr) and
a rise in Stimplies a depreciation/devaluation of the Ethiopian Birr; Ptis the domestic price
level and P∗
tis the foreign price level and it is normalized to one.
35

exceed the country’s trade deficit, there will be a positive change in foreign
reserves and vice versa.
Given its components listed above, the aggregate demand equation, assum-
ing a good’s market equilibrium (Yd
t=Ys
t=Yt), can be given as:
Yd
t=C0,t+σYd
t+I0,t−ρRt+G0,t+Ft+ηQt−θYd
t+X0,t
| {z }
NX = ∆Rest−Ft
(A7)
Yd
t=1
1−σ+θ[C0,t+I0,t+G0,t
| {z }
Dt
+X0,t+Ft−ρRt+ηQ](A8)
Appendix B Long-run equilibrium
Long-run Price Level
pt=λ
ρ+ρλ +ηλ "dt+xt+ft−λ+γρ
γλ yt+ρ+ηλ
λmt+ηet−ρE[pt+1]#(B1)
pt=λ(dt+xt+ft)
ρ+ρλ +ηλ −(λ+γρ)yt
ργ +ρλγ +ηλγ +(ρ+ηλ)mt
ρ+ρλ +ηλ +λη(et)
ρ+ρλ +ηλ −λρE[pt+1]
ρ+ρλ +ηλ (B2)
Forwarding the first-order difference equation in price given in Equation B2
one period, we get:
pt+1=λ(dt+1+x0,t+1+ft+1)
ρ+ρλ +ηλ −(λ+γρ)yt+1
ργ +ρλγ +ηλγ +(ρ+ηλ)mt+1
ρ+ρλ +ηλ
+λη(et+1)
ρ+ρλ +ηλ −λρE[pt+2]
ρ+ρλ +ηλ (B3)
Substituting pt+1back into pt, we get:
36

pt=λdt+(1 −η)(xt+ft)+η∆rest
ρ(1 +λ)+mt
1+λ−yt λ+γρ −λγηθ
γρ(1 +λ!
+E"λ2dt+1+(1 −η)(x0,t+1+ft+1)+η∆rest+1
ρ(1 +λ)2#+λEmt+1
(1 +λ)2
−λ λ+γρ −λγηθ
γρ(1 +λ)2!Eyt+1+λ
1+λ2
E[pt+2](B4)
pt=λ(dt+xt+ft)
ρ+ρλ +ηλ −(λ+γρ)yt
ργ +ρλγ +ηλγ +(ρ+ηλ)mt
ρ+ρλ +ηλ +λη(et)
ρ+ρλ +ηλ
−λρ
ρ+ρλ +ηλ "λ(dt+1+x0,t+1+ft+1)
ρ+ρλ +ηλ −(λ+γρ)yt+1
ργ +ρλγ +ηλγ +(ρ+ηλ)mt+1
ρ+ρλ +ηλ
+λη(et+1)
ρ+ρλ +ηλ −λρE[pt+2]
ρ+ρλ +ηλ#(B5)
pt=λ
ρ+ρλ +ηλ "dt+xt+ft+λρ
ρ+ρλ +ηλ(dt+1+x0,t+1+ft+1)#(B6)
−λ+γρ
ργ +ρλγ +ηλγ "yt+λρ
ρ+ρλ +ηλ yt+1#+ρ+ηλ
ρ+ρλ +ηλ "mt+λρ
ρ+ρλ +ηλmt+1#
+λη
ρ+ρλ +ηλ "et+λρ
ρ+ρλ +ηλet+1#+ λρ
ρ+ρλ +ηλ!2
E[pt+2](B7)
A repeated iteration of this can be summarized as:
pt=λ
ρ+ρλ +ηλ lim
T→∞
T−1
X
k=t




 λρ
ρ+ρλ +ηλ!k−t
E[dk+x0,k+fk]





−λ+γρ
ργ +ρλγ +ηλγ lim
T→∞
T−1
X
k=t




 λρ
ρ+ρλ +ηλ!k−t
Eyk





+ρ+ηλ
ρ+ρλ +ηλ lim
T→∞
T−1
X
k=t




 λρ
ρ+ρλ +ηλ!k−t
Emk





+λη
ρ+ρλ +ηλ
lim
T→∞
T−1
X
k=t




 λρ
ρ+ρλ +ηλ!k−t
Eek





+lim
T→∞
T
X
k=t+1




 λρ
ρ+ρλ +ηλ!k−t
Epk




(B8)
37

A bounded solution requires that:
lim
T→∞
T
X
k=t+1




 λρ
ρ+ρλ +ηλ!k−t
Epk





=0and (B9)

λ
ρ+ρλ +ηλ lim
T→∞
T−1
X
k=t λρ
ρ+ρλ +ηλ!k−t
E[dk+x0,k+fk]
−λ+γρ
ργ +ρλγ +ηλγ lim
T→∞
T−1
X
k=t λρ
ρ+ρλ +ηλ!k−t
Eyk+ρ+ηλ
ρ+ρλ +ηλ
lim
T→∞
T−1
X
k=t λρ
ρ+ρλ +ηλ!k−t
Emk+λη
ρ+ρλ +ηλ
lim
T→∞
T−1
X
k=t λρ
ρ+ρλ +ηλ!k−t
Eek
<∞(B10)
Both of the above conditions will be satisfied given that λρ
ρ+ρλ+ηλ <1. Or,
more specifically, unless the absolute value of the log of price level grows at
an exponential rate of ρ+ρλ+ηλ
λρ or above. And the convergence of the second
term requires that the sequences in {dt,xt,ft,mt,yt,et}do not grow at an
exponential rate of ρ+ρλ+ηλ
λρ or above.
Letting vt={dt,xt,ft,mt,yt,et}
vt=vt−1+µv
t
vt+1=vt+µv
t+1
vt+2=vt+1+µv
t+2
vt+2=vt+µv
t+1+µv
t+2
etc... and
E(µv
t+i)=0for i≥1
lim
T→∞
T−1
X
k=t λρ
ρ+ρλ +ηλ!k−t
=ρ+ρλ +ηλ
ρ+ηλ
The long-run equilibrium price level can therefore be given by:
pt=λ
ρ+ηλ[dt+xt+ft]−λ+γρ
ρ+ηλ yt+mt+λη
ρ+ηλet(B11)
38

Appendix C Variables in Log Levels and Differ-
ences
Figure C10: Trends in Log GDP: In Log Levels and Differences
LGDP
1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013
21.2
21.4
21.6
21.8
22.0
22.2
22.4
22.6
22.8 Levels
1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013
-0.04
-0.02
0.00
0.02
0.04
0.06 Differences
39

Figure C11: Trends in Log Real Exchange Rate: In Log Levels and Differences
LRER
1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013
1.6
1.7
1.8
1.9
2.0
2.1
2.2
2.3
2.4
2.5 Levels
1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013
-0.20
-0.15
-0.10
-0.05
-0.00
0.05
0.10
0.15 Differences
Figure C12: Consumer Price: In Log Levels and Differences
LCPI
1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013
4.00
4.25
4.50
4.75
5.00
5.25
5.50
5.75
6.00
6.25 Levels
1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013
-0.10
-0.05
0.00
0.05
0.10
0.15
0.20 Differences
40

Figure C13: Trends in Log Nominal Exchange Rate: In Log Levels and Differences
LNER
1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013
1.6
1.8
2.0
2.2
2.4
2.6
2.8
3.0 Levels
1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013
-0.025
0.000
0.025
0.050
0.075
0.100
0.125 Differences
Figure C14: Trends in Log Real Money Demand: In Log Levels and Differences
LREAL_MONEYBAL
1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013
5.75
6.00
6.25
6.50
6.75
7.00
7.25
7.50 Levels
1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013
-0.15
-0.10
-0.05
0.00
0.05
0.10
0.15 Differences
41

Table C4: Summary of Unit Root Tests: ADF, Philip Perron and KPSS
ADF Philip Perron KPSS
Variables Obs Lags Z(t) p val Obs Lags Z(t) p val Obs Lags Test Stat
Log of Consumer Prices 82 5 1.355 0.997 87 3 1.619 0.998 88 6 1.198
Log Nominal Exchange Rate 83 4 -0.388 0.912 87 3 0.456 0.983 88 6 1.218
Log Output 84 3 -1.392 0.863 87 3 -0.827 0.963 88 6 0.443
Log Real Exchange Rate 82 5 -1.242 0.655 87 3 -1.165 0.689 88 6 0.326
Log Real Money Demand 81 5 -0.832 0.810 86 3 -0.827 0.811 88 6 0.399
Log Coffee Price 82 5 -1.312 0.624 87 3 -1.922 0.322 87 6 1.245
Log Oil Price 83 4 -1.205 0.671 87 3 -1.098 0.716 88 6 1.271
KPSS Crit. values for H0: X=level stationary: 10%: 0.347,5%: 0.463, 1%: 0.739
KPSS Crit. values for H0: X=trend stationary: 10%: 0.119,5%: 0.146, 1%: 0.216
42

Table C5: Summary of Unit Root Tests in Differences: ADF, Philip Perron and KPSS
ADF Philip Perron KPSS
Variables Obs Lags Z(t) p val Obs Lags Z(t) p val Obs Lags Test Stat
Diff. Log Consumer Prices 82 4 -3.561 0.007 86 3 -5.591 0.000 87 5 0.520
Diff. Log Nominal Exchange Rate 83 3 -3.150 0.023 86 3 -6.662 0.000 87 6 0.174
Diff. Output 81 5 -2.887 0.047 86 3 -4.159 0.001 87 6 0.066
Diff. Real Exchange Rate 82 4 -3.575 0.006 86 3 -5.798 0.000 87 5 0.364
Diff. Real Money Demand 81 4 -3.972 0.002 85 3 -8.031 0.000 86 11 0.143
Diff. Coffee Price 82 4 -5.120 0.000 86 3 -7.088 0.000 87 3 0.073
Diff. Oil Price 83 3 -5.341 0.000 86 3 -6.597 0.000 87 9 0.116
KPSS Crit. values for H0: X=level stationary: 10%: 0.347,5%: 0.463, 1%: 0.739
KPSS Crit. values for H0: X=trend stationary: 10%: 0.119,5%: 0.146, 1%: 0.216
43

Table C6: Clemente-Monta ˜
n´
es-Reyes unit-root test with double mean shifts
CLEMAO2 CLEMIO2
Variables Obs Lags t-stat (Br1) t-stat (Br2) t-stat(ρ−1) Lags t-stat (Br1) t-stat (Br2) t-stat(ρ−1)
Log of Consumer Prices 80 2 15.345a3.776a-2.833 3 3.512a2.330b-2.941
Log of Nominal Exchange Rate 80 4 12.025a20.271a-3.550 12 1.631 5.751a-4.660
Log of Output 80 0 15.258a9.988a-2.822 12 3.666a1.044 -0.762
Log of Real Exchange Rate 80 12 6.855a-8.848a-2.384 1 1.922c-3.655a-3.698
Log of Real Money Demand 79 0 16.762a7.662a-3.170 0 2.168b3.569a-3.298
Log of Coffee Price 80 1 -6.153a9.554a-4.493 9 -3.033a3.761a-2.931
Log of Oil Price 80 1 7.942a16.139a-5.270 1 3.333a3.800a-4.825
Note: a, b and c indicate statistical significance at the 1%, 5%, and 10% level, respectively.
t-stat (Br1) and t-stat (Br2) are t-statistics values for the two break points.
t-stat (ρ−1) is the test statistics for the unit root test (ρ−1), the 5% critical value= -5.490
44

Table C7: Johansen test for co-integration
Rank Trace Stat. 5% Crit. 1% Crit.
0 124.89 87.31 96.58
1 54.26 62.99 70.05
2 18.31 42.44 48.45
3 10.42 25.32 30.45
4 3.60 12.25 16.26
Deterministic terms: Restricted Trend.
Table C8: Test for Autocorrelation (Lagrangian Multiplier Test)
lag chi2 df Prob >chi2
1 33.1260 25 0.12795
2 32.7648 25 0.13705
3 26.3637 25 0.38840
4 29.5620 25 0.24111
5 35.5554 25 0.07862
6 29.6241 25 0.23867
H0: no autocorrelation at lag orders.
Figure C15: VAR Stability Test
−1 −.5 0 .5 1
Imaginary
−1 −.5 0 .5 1
Real
Roots of the companion matrix
45

Table C9: Tests for Normality
Jarque-Bera test Skewness test Kurtosis test
Equation chi2 df Prob >chi2Skewness chi2 df Prob >chi2Kurtosis chi2 df Prob >chi2
Diff LnGDP 34.769 2 0.000 -.41633 2.427 1 0.11929 6.0398 32.342 1 0.000
Diff LnRER 2.807 2 0.24572 .18167 0.462 1 0.49666 2.1815 2.345 1 0.12568
Diff LnReal Money Bal. 4.834 2 0.08920 -.29024 1.179 1 0.27749 1.9782 3.654 1 0.05592
Diff LnNER 20.769 2 0.00003 .91339 11.680 1 0.00063 4.6114 9.089 1 0.00257
Diff LnCPI 0.429 2 0.80701 -.15932 0.355 1 0.55108 2.8551 0.073 1 0.78637
All 63.607 10 0.000 16.103 5 0.00655 47.504 5 0.000
46

Figure C16: Trends in Aggregate CPI, Food and Non-Food Price Indices
1998 2000 2002 2004 2006 2008 2010 2012 2014
3.5
4.0
4.5
5.0
5.5
6.0
6.5
LnFood PI
LnNon-Food PI
LnCPI
1998 2000 2002 2004 2006 2008 2010 2012 2014
-0.4
-0.3
-0.2
-0.1
-0.0
0.1
0.2
0.3
Diff LnFood PI
Diff LnNon-Food PI
Diff LnCPI
47

Table C10: Major Groups in the 2006 and 2011 based CPI and their Weights at the
Country Level
Weights
Items December 2006=100 December 2011=100
1. Food and Non-Alcoholic Beverage 0.57 0.53
2. Non-Food 0.43 0.47
Alcoholic Beverages and Tobacco, Nar -
cotics
2.5 4.85
Clothing and Footwear 8.32 6.62
Housing, Water, Electricity, Gas and
Other Fuels
20.56 16.34
Furnishing, Housing Equipment and
maintenance of the house
3.75 5.41
Health 1.11 1.08
Transport 2.5 2.8
Communication - 1.08
Recreation and Culture 1.1 0.6
Education - 0.45
Restaurant and Hotels - 5.46
Miscellaneous Goods and Services 3.2 2.56
Total 100 100
48

Figure C17: Trends in CPI as well as Nominal and Real Exchange Rates in Ethiopia:
1993Q1-2014Q4
5 10 15 20
NER and RER (Birr/USD)
100 200 300 400 500
CPI, 2005=100
1995q1 2000q1 2005q1 2010q1 2015q1
date
CPI NER RER
Figure C18: Trends in Inflation in Ethiopia (percent change, corresponding period
previous year): 1993Q1-2014Q4
−20.00 0.00 20.00 40.00 60.00
Inflation (y/y, %)
1995q1 2000q1 2005q1 2010q1 2015q1
date
49

Appendix D Additional Results
Figure D19: Cumulative Impulse Response of the CPI to Nominal Exchange Rate
and Money Supply Shocks: using Choleski Decomposition
Nom. Exchange Rate Shock
Money Supply Shock
Consumer Price Index
5 10 15 20
0.010
0.012
0.014
0.016
0.018
0.020
0.022
5 10 15 20
0.002
0.003
0.004
0.005
0.006
0.007
0.008
50

Table D11: Decomposition of Variance for Aggregate Output Supply
Supply BOP Demand Money SS Nom. ExRate
Horizon Shock Shock Shock Shock Shock
1 79.0 3.2 5.0 2.9 2.1
[62.4,88.9] [ 0.3,13.7] [ 0.5,16.8] [ 0.3, 8.7] [ 0.2, 9.0]
2 83.9 2.4 4.9 1.5 1.7
[69.1,92.3] [ 0.4,10.7] [ 0.8,14.9] [ 0.5, 4.3] [ 0.4, 6.7]
3 87.5 1.9 3.9 1.5 1.3
[74.8,94.3] [ 0.3, 8.6] [ 0.7,11.7] [ 0.7, 3.1] [ 0.3, 5.0]
4 90.2 1.6 3.0 1.2 1.1
[79.6,95.7] [ 0.3, 6.9] [ 0.6, 9.1] [ 0.5, 2.4] [ 0.3, 4.0]
8 95.6 0.8 1.4 0.5 0.6
[90.4,98.1] [ 0.2, 3.3] [ 0.3, 4.0] [ 0.2, 1.1] [ 0.2, 1.9]
12 97.3 0.5 0.8 0.3 0.4
[94.2,98.8] [ 0.1, 2.0] [ 0.2, 2.4] [ 0.1, 0.7] [ 0.1, 1.2]
16 98.1 0.3 0.6 0.2 0.3
[96.0,99.1] [ 0.1, 1.4] [ 0.1, 1.7] [ 0.1, 0.5] [ 0.1, 0.9]
Table D12: Decomposition of Variance for Real Exchange Rate
Supply BOP Demand Money SS Nom. ExRate
Horizon Shock Shock Shock Shock Shock
1 22.8 46.6 2.2 9.8 11.3
[ 9.4,38.8] [26.4,65.3] [ 0.2, 9.4] [ 4.0,18.3] [ 1.5,27.4]
2 22.5 57.1 2.3 5.2 7.0
[ 8.1,40.7] [36.1,74.7] [ 0.6, 8.5] [ 1.9,11.0] [ 1.2,17.2]
3 22.8 64.1 1.7 2.9 4.3
[ 7.4,42.4] [42.5,80.6] [ 0.5, 6.2] [ 1.2, 5.9] [ 0.9,11.9]
4 22.6 67.8 1.3 2.0 3.0
[ 7.0,43.2] [45.7,84.0] [ 0.4, 4.8] [ 0.9, 4.0] [ 0.7, 8.7]
8 25.0 70.9 0.6 0.8 1.3
[ 7.4,47.9] [47.5,88.4] [ 0.2, 2.2] [ 0.3, 1.5] [ 0.3, 3.6]
12 26.2 71.4 0.4 0.4 0.7
[ 7.6,50.4] [47.1,89.8] [ 0.1, 1.3] [ 0.2, 0.9] [ 0.2, 2.1]
16 26.8 71.5 0.3 0.3 0.5
[ 7.8,51.8] [46.6,90.4] [ 0.1, 0.9] [ 0.1, 0.7] [ 0.1, 1.5]
51

Table D13: Decomposition of Variance for Demand for Real Money Balances
Supply BOP Demand Money SS Nom. ExRate
Horizon Shock Shock Shock Shock Shock
1 6.6 41.4 4.3 3.3 36.9
[ 0.7,23.1] [21.9,61.5] [ 0.4,17.6] [ 0.5, 9.0] [12.9,50.8]
2 12.8 42.2 4.0 2.0 31.5
[ 2.5,32.3] [22.8,61.9] [ 0.6,15.8] [ 0.5, 5.5] [10.8,45.6]
3 14.0 43.3 3.7 1.2 30.0
[ 2.6,35.1] [24.0,62.5] [ 0.6,15.2] [ 0.4, 3.0] [11.3,43.7]
4 16.3 43.0 3.6 0.8 28.2
[ 3.2,38.7] [23.9,62.5] [ 0.6,14.4] [ 0.3, 2.0] [10.7,42.5]
8 20.8 41.5 3.3 0.3 25.0
[ 4.6,45.4] [22.5,62.2] [ 0.5,13.4] [ 0.1, 0.8] [10.3,40.4]
12 22.6 40.6 3.1 0.2 23.9
[ 5.1,48.2] [21.7,61.7] [ 0.5,13.2] [ 0.1, 0.5] [ 9.8,39.9]
16 23.4 40.2 3.0 0.1 23.3
[ 5.4,49.4] [21.2,61.6] [ 0.5,13.3] [ 0.0, 0.3] [ 9.5,39.7]
Table D14: Decomposition of Variance for Nominal Exchange Rate
Supply BOP Demand Money SS Nom. ExRate
Horizon Shock Shock Shock Shock Shock
1 11.5 2.4 50.3 0.0 25.1
[ 1.8,29.4] [ 0.2,10.2] [18.3,75.5] [ 0.0, 0.0] [ 3.2,63.5]
2 9.9 4.8 43.6 2.2 26.9
[ 1.8,27.0] [ 1.2,15.3] [14.6,68.4] [ 0.7, 4.7] [ 4.0,65.2]
3 8.9 5.1 43.3 3.7 24.9
[ 1.5,26.0] [ 1.0,16.9] [15.0,67.8] [ 1.4, 7.1] [ 3.7,63.8]
4 8.5 5.7 41.0 4.6 25.4
[ 1.4,25.7] [ 1.1,18.7] [13.6,65.7] [ 1.9, 8.7] [ 4.0,63.5]
8 8.4 7.2 37.5 5.1 25.1
[ 1.3,27.0] [ 1.2,21.9] [11.7,63.3] [ 2.1, 9.7] [ 4.2,62.4]
12 8.7 7.8 36.0 5.2 24.8
[ 1.3,28.3] [ 1.2,23.6] [11.1,62.3] [ 2.1,10.0] [ 4.2,61.5]
16 8.8 8.2 35.3 5.3 24.3
[ 1.2,29.0] [ 1.2,24.7] [10.6,62.0] [ 2.1,10.0] [ 4.1,60.9]
52

References
Alexius, Annika and Erik Post, “Exchange rates and asymmetric shocks in
small open economies,” Empirical Economics, 2008, 35 (3), 527–541.
Aron, Janine, Ronald Macdonald, and John Muellbauer, “Exchange Rate
Pass-Through in Developing and Emerging Markets: A Survey of Concep-
tual, Methodological and Policy Issues, and Selected Empirical Findings,”
The Journal of Development Studies, 2014, 50 (1), 101–143.
Bacchetta, Philippe and Eric van Wincoop, “A theory of the currency denom-
ination of international trade,” Journal of International Economics, December
2005, 67 (2), 295–319.
Barhoumi, Karim, “How Structural Macroeconomic Shocks Can Explain Ex-
change Rate Pass-Through in Developing Countries? A Common Trend Ap-
proach,” Global Economy Journal, 2009, 9(2).
Blanchard, Olivier Jean and Danny Quah, “The Dynamic Effects of Aggregate
Demand and Supply Disturbances,” The American Economic Review, 1989,
79 (4), pp. 655–673.
Choudhri, Ehsan U. and Dalia S. Hakura, “Exchange rate pass-through to
domestic prices: Does the inflationary environment matter?,” Journal of In-
ternational Money and Finance, 2006, 25 (4), 614 – 639.
Clarida, Richard and Jordi Gali, “Sources of real exchange-rate fluctuations:
How important are nominal shocks?,” Carnegie-Rochester Conference Series
on Public Policy, December 1994, 41 (1), 1–56.
Clemente, Jess, Antonio Montas, and Marcelo Reyes, “Testing for a unit
root in variables with a double change in the mean,” Economics Letters,
1998, 59 (2), 175 – 182.
Dornbusch, Rdiger, “Exchange Rate Economics: Where Do We Stand?,”
Brookings Papers on Economic Activity, 1980, 11 (1, Tenth Anniversary Is-
sue), 143–206.
Estima,RATS Version 8: User’s Guide, Estima, 2010.
Frimpong, Siaw and Anokye Mohammed Adam, “Exchange Rate Pass-
Through in Ghana,” International Business Research, April 2010, 3(2).
Gali, Jordi, “How Well Does the IS-LM Model Fit Postwar U.S. Data?,” The
Quarterly Journal of Economics, 1992, 107 (2), pp. 709–738.
Geiger, Michael and Chorching Goh, “Ethiopia Economic Update Overcom-
ing Inflation,Raising Competitiveness,” Technical Report 20026, The World
Bank 2012.
53

Hossain, A. Akhand, “Granger-Causality Between Inflation, Money Growth,
Currency Devaluation and Economic Growth in Indonesia, 1951-2002,” In-
ternational Journal of Applied Econometrics and Quantitative Studies, 2005,
2(3), 45–68.
IMF, “The Federal Democratic Republic of Ethiopia: Staff Report for the 2014
Article IV Consultation,” IMF Staff Country Reports 14/303, International
Monetary Fund October 2014.
Ito, Takatoshi and Kiyotaka Sato, “Exchange Rate Pass-Through and Do-
mestic Inflation: A Comparison between East Asia and Latin American
Countries,” Discussion papers 07040, Research Institute of Economy, Trade
and Industry (RIETI) June 2007.
and , “Exchange Rate Changes and Inflation in Post-Crisis Asian
Economies: Vector Autoregression Analysis of the Exchange Rate Pass-
Through,” Journal of Money, Credit and Banking, 2008, 40 (7), 1407–1438.
Juselius, K.,The Cointegrated VAR Model: Methodology and Applications Ad-
vanced Texts in Econometrics, OUP Oxford, 2006.
Loening, Josef L., Dick Durevall, and Yohannes A. Birru, “Inflation dynam-
ics and food prices in an agricultural economy: the case of Ethiopia,” Policy
Research Working Paper Series 4969, The World Bank 2009.
McCarthy, Jonathan, “Pass-through of exchange rates and import prices
to domestic inflation in some industrialized economies,” Technical Report
2000.
, “Pass-Through of Exchange Rates and Import Prices to Domestic Inflation
in Some Industrialized Economies,” Eastern Economic Journal, Fall 2007, 33
(4), 511–537.
Melesse, Wondemhunegn Ezezew, “Exchange Rate Transmission into Sec-
toral Consumer Price Inflation in Ethiopia-SVAR Approach,” Journal of Eco-
nomic and Financial Modelling, 2014, 2(1), 37–54.
Mwase, Nkunde, “An Empirical Investigation of the Exchange Rate Pass-
Through to Inflation in Tanzania,” IMF Working Papers 06/150, Interna-
tional Monetary Fund June 2006.
Obstfeld, Maurice, “Floating Exchange Rates: Experience and Prospects,”
Brookings Papers on Economic Activity, 1985, 16 (2), 369–464.
Razafimahefa, Ivohasina Fizara, “Exchange Rate Pass-Through in Sub-
Saharan African Economies and its Determinants,” IMF Working Papers
12/141, International Monetary Fund June 2012.
Shambaugh, Jay, “A new look at pass-through,” Journal of International
Money and Finance, June 2008, 27 (4), 560–591.
54

Sims, Christopher A., “Are forecasting models usable for policy analysis?,”
Federal Reserve Bank of Minneapolis Quarterly Review, 1986, (Win), 2–16.
and Tao Zha, “Error Bands for Impulse Responses,” Econometrica, Septem-
ber 1999, 67 (5), 1113–1156.
Sissoko, Yaya and Sel Dibooglu, “The exchange rate system and macroeco-
nomic fluctuations in Sub-Saharan Africa,” Economic Systems, 2006, 30 (2),
141 – 156.
Taylor, John B., “Low inflation, pass-through, and the pricing power of firms,”
European Economic Review, June 2000, 44 (7), 1389–1408.
Wehinger, GertD., “Causes of Inflation in Europe, the United States and
Japan: Some Lessons for Maintaining Price Stability in the EMU from a
Structural VAR Approach,” Empirica, 2000, 27 (1), 83–107.
World Bank, “Third Ethiopia Economic Update : Strengthening Export Perfor-
mance through Improved Competitiveness,” World Bank Other Operational
Studies 20026, The World Bank June 2014.
Younger, Stephen D., “Testing the Link Between Devaluation and Inflation:
Time Series Evidence from Ghana,” Journal of African Economies, 1992, 1
(3), 369–394.
55