Large Open Economy Monetary Policy Shocks and the Developing World

8/16/2019 Large Open Economy Monetary Policy Shocks and the Developing World

1/63

Large Open Economy Monetary Policy Shocks and The

Developing World: The Case of Sub-Saharan Africa

Jeremy Kronick

Department of Economics

International Business School

Brandeis University

415 South St.Waltham, MA, 02453

[email protected]

January 31, 2014

Abstract

This paper adds to the empirical side of the international transmission literature by com-

paring how large open economy contractionary monetary policy shocks originating in the Euro

Area (“EU”) and US aff ect a set of thinly investigated sub-Saharan African (“SSA”) countries,who diff er in many respects including exchange rate regime. I depart from standard structural

vector autoregression (“SVAR”) methodologies and use a two-stage model whereby EU and US

monetary policy shocks series are extracted from a factor model in the first stage and used as

regressors in a SSA country SVAR model in the second stage. Results diff er depending on which

of the EU or US is the shocking country, and the type of exchange rate regime the SSA country

possesses. Floating exchange rate countries are more negatively impacted by US shocks as the

bulk of their external debt is denominated in US dollars, while fixed exchange rate countries are

more sensitive to EU shocks due to large trading relationships with the EU and being pegged

to the Euro. Policy implications are discussed.

1


2/63

1 Introduction

Many papers have looked at international transmission of monetary policy shocks and evaluated

both the economic eff ects as well as the transmission mechanisms (see, e.g. Kim (2001); Canova

(2005); Mackowiak (2006, 2007); Kozluk and Mehrotra (2009); Jannsen and Klein (2011)). However,

the results have been mixed in two primary regards, and this paper sets out to shed some light on

these issues in the context of countries that have been thinly investigated: first, whether or not we

should expect a positive or negative correlation between the foreign monetary policy shock and the

domestic GDP response, and second whether we should look to trade balance or interest rates to

help explain the transmission. The literature has also been quiet on how exogenous monetary policy

shocks from two diff erent large open economies aff ect the same set of developing economies. For

least developed countries (“LDC”), such as the sub-Saharan African (“SSA”) countries in this paper,these questions have become increasingly relevant as the financial liberalization these economies

have undertaken in the past 20 years has caused their economies to be more sensitive to foreign

shocks.

I study 11 SSA countries including the floating exchange rate economies of South Africa, Ghana,

Kenya, Tanzania, Uganda, Mauritius, and the fixed exchange rate economies of Gabon, Cote

D’Ivoire, Botswana, Senegal, and Rwanda. The possible output eff ects and transmission mech-

anism will in part depend on the type of exchange rate regime in these countries. Each country

has a large trading relationship with the foreign countries in this study, namely the Euro Area

(“EU”) and US. If trade is a strong transmission mechanism and substitution eff ects are dominant,

a contractionary monetary policy shock will cause an appreciation of the Euro or US dollar relative

to the floating SSA currency thereby making SSA goods cheaper, causing SSA trade and GDP to

increase. This eff ect is muted for fixed exchange rate countries. However, if income eff ects are

dominant, the contractionary shock in the EU or US will cause a slowdown in that economy and

a decrease in purchases of SSA goods causing trade and GDP to fall regardless of exchange rate

regime.

These countries also have a significant portion of their external debt denominated in either the Euro

or US dollar. For floating exchange rate countries a contractionary monetary policy shock could lead

to an appreciation of the foreign currency with respect to the domestic currency, thereby making

debt that much more expensive. Should a floating exchange rate country respond by increasing

interest rates in order to off set the appreciation, they risk harming the rest of the economy. To

the extent that floating exchange rate economies respond in this fashion, interest rates act as the

transmission mechanism. For fixed exchange rate countries the concern is competitiveness of capital

flows. If a fixed exchange rate country does not increase interest rates following a foreign increase,

they risk losing some of that capital flow competitiveness, and will be forced to defend their currency

by buying it back with foreign reserves or by using some form of capital control. 1 To the extent

1This represents the impossible trinity for fixed exchange rate countries. Specifically the fact that having a fixedexchange rate, free capital movement, and independent monetary policy are not simultaneously possible.

2


3/63

that capital inflows matter, these countries will increase interest rates at the risk of damaging the

economy as a whole. Therefore, as the preceding paragraphs indicate, a priori , it is not clear whether

trade, interest rates, or both, play a role in the transmission of foreign monetary policy shocks.

To answer the uncertainties in the literature on the international transmission of monetary policy

shocks, I use a structural vector autoregression (“SVAR”). The typical SVAR methodologies in the

international transmission literature have struggled to account for the following issues simultane-

ously: a combination of appropriately identified foreign monetary policy shocks and interaction of

recipient country variables, while at the same time reducing overparameterization when using small

samples sizes. In order to deal with each of the issues, I propose a novel two-stage model. In the

first stage I estimate a factor model similar to Barakchian and Crowe (2013) who argue that many

of the SVAR models used to identify US monetary policy shocks do not account for the forward-

looking nature of current US monetary policy thus creating impulse response functions inconsistentwith economic theory. Using their factor model for both the EU and US, I extract the shock series.

The US series is very similar to the one in Barakchian and Crowe (2013) while the EU series is

new to the literature to the knowledge of this paper. I then follow Romer and Romer (2004) in

the second stage, and cumulate the shock series in order to produce an I(1) variable. I then insert

this I(1) shock series into a SVAR model for the SSA country of interest treating the shock series

as a regressor. By performing these two stages, I have simultaneously identified a truly exogenous

monetary policy shock series, and reduced the issue of overparameterization.

Using this two-stage model, results indicate that significant diff erences exist depending on whether

the EU or US is the shocking country, and whether or not the recipient SSA country is a floatingor fixed exchange rate economy, this latter result diff ering from some other papers in which results

for the two types of exchange regimes are not significantly diff erent.2 Following the contractionary

monetary policy shock in the EU, floating exchange rate countries have mixed results with some

experiencing falls in GDP while others do not. Despite large trading relationships with the EU,

interest rates are the dominant transmission mechanism for these countries, with some countries

using countercyclical interest rate policy to off set the contractionary shock. The countries using this

countercyclical policy seem to be the countries with lower levels of capital flows relative to GDP,

and are thus less concerned with capital flow competitiveness.

Fixed exchange rate regime countries, in general, seem likely to experience significant real economy

contractions following the EU shock with interest rates and trade acting as transmission mechanisms.

It appears that contractionary shocks in the EU lead to less purchases of SSA goods by the EU,

causing trade balance and real GDP to fall. In an eff ort to keep capital flows coming in, and in an

attempt to lower price levels to off set negative trade eff ects, these countries increase their interest

rates as well, causing a further fall in GDP.3

2Canova (2005) states that the diff erences between the two regimes are more due to magnitude than transmissionmechanism or output response.

3Without the ability to adjust exchange rates, countries attempt to lower domestic prices to off set these tradeeff ects. These countries experience mixed results in successfully lowering prices, and even those that are successful,are, for the most part, not able to fully off set the negative trade eff ects.

3


4/63

Following the US shock, a diff erent story arises. For floating exchange rate countries, contractions

are likely to occur as a dominant income eff ect in the US leads to a fall in trade, and an increase in

SSA interest rates in an attempt to keep the real debt value unchanged further harms the economy.

The sensitivity of these countries to trade with the US despite lower levels of trade is a by-product

of their concern over debt. By increasing interest rates to off set debt burden increases, they lose

the ability to try and depreciate the currency further in order to off set the dominance of the income

eff ect in the US.4

Fixed exchange rate countries seem to be able to avoid contractionary results following the US

shock. As neither trade or interest rates act as a transmission mechanism, some other factor not

investigated in this analysis is playing a significant role. One example might be aid. Despite the

US’ policy of attempting to help the developing world by moving away from aid and towards trade 5,

SSA countries remain the largest recipients of US aid.6

The implication of this analysis for policymakers in these SSA countries is that external monetary

policy shocks are going to matter, with the country originating the shock and the exchange rate

regime of the recipient country playing pivotal roles in determining the output response as well as

the importance of trade and interest rates as transmission mechanisms. Diversifying debt burdens

away from US dollar domination as well as considering a move away from fixed exchange rate regimes

as these countries continue to develop seem to be policies to consider as ways of hedging some of

the potential foreign contractionary monetary policy shock risk.

The rest of this paper is structured as follows: Section 2 develops and explains the two stage model,

including identification and empirical methodology decisions. Section 3 analyzes the results forthe main specification. Section 4 discusses robustness checks as well as an extension. Section 5

concludes.

2 Model

In this section I introduce the novel, two-stage method of evaluating international transmission of

monetary policy shocks. I then evaluate the EU and US monetary policy shocks extracted from the

first stage factor model in order to confirm that they are appropriate, before explaining in greater

detail the second stage SVAR model setup for the SSA countries.

The bulk of papers in the international transmission literature use SVAR methodologies that require

a potential trade-off between the combination of appropriately identified exogenous foreign mone-

tary policy shocks with interaction of recipient country variables, while at the same time reducing

overparameterization in small sample datasets. For example, Kim (2001) uses an extensive model for

4Domestic price levels do not respond quick enough to help off set the income eff ect with increasing interest rates.5In 2000, the US Congress approved the African Growth and Opportunity Act (“AGOA”), a piece of legislation

focused on trade.6Of the 40 countries still receiving over $100 MN in aid a year in 2011, 21 of them are in SSA. See US AID for

details.

4


5/63

the US in order to identify the monetary policy shocks by adopting the setup of CEE (1996, 1999),

where a recursive ordering for the US includes, in order, a US real economy variable, a US price

level indicator, a commodity price index, and the US federal funds rate (“FFR”). However, due to

overparameterization, Kim (2001) is forced to separately add the recipient country variables which

results in not being able to consider how the interaction between the recipient country variables

following the monetary policy shock aff ects the results.

Mackowiak (2007) on the other hand uses a block-block model whereby a fully identified US model

is used in the same SVAR as a recipient country model. This in theory should account for both

the lack of a truly identified exogenous monetary policy shock, as well as being able to investigate

the interaction of multiple recipient country variables following the shock. The issue with this type

of modeling is that it dramatically increases the number of parameters being investigated, and in

small datasets such as the one being used in this paper, this potentially becomes infeasible (see e.g.Jannsen and Klein (2011)).

2.1 Stage 1 Identification

As shown by shown by Barakchian and Crowe (2013), performing the CEE (1996) identification

scheme for the US produces a real economy puzzle if viewed in the period from 1988-2008. Specifi-

cally, a contractionary monetary policy shock in the US will increase real GDP contrary to economic

theory. These authors argue that identifying exogenous monetary policy shock series in a SVAR

setting requires one to be fairly confident in both the information set being used, as well as the

contemporaneous relationships being assumed. The CEE (1996) method fails in two regards: first

the assumption that the monetary policy decision can be determined from looking at only a real

economy variable, a price index, and a commodity price index is likely to be incorrect in a complex

US economy, while secondly not accounting for the forward looking nature of US monetary policy

likely leaves out important details. The authors argue that looking at financial market data on

futures rates, such as the eurodollar futures rates for the EU and the federal funds futures rates

for the US, at the time of their respective central banking meeting will produce better results for

identifying truly exogenous monetary policy shocks. Specifically, any change in futures rates from

right before the meetings with results directly after the meetings, will represent the true exogenous

monetary policy shocks, as at the time of the meeting the private sector should have access to allinformation going into the decisions being made at the European Central Bank (“ECB”) or Federal

Open Market Committee (“FOMC”).

More formally, Barakchian and Crowe (2013) assume we can identify the exogenous monetary policy

shocks as:

S t = f (Ωt) + st (1)

where S t is the monetary policy stance and f is a linear function that describes the relationship

5


6/63

between the monetary policy stance and the information set Ω.7

Let there be two diff

erent measures of the private sector’s beliefs or expectations for the monetarypolicy stance taken by the ECB or FOMC, namely one right before the meeting, t−1 Ŝ t, and one

right after any policy announcement, t Ŝ t. Each of these measures are noisy versions of the private

sector’s true expectations:

t−1 Ŝ t = E

pt−1[S t] + ξ t−1 = E

pt−1[f (Ωt)] + ξ t−1 (2)

t Ŝ t = E

pt [S t] + ξ t = S t + ξ t = f (Ωt) + ξ t (3)

We make the assumption that ξ t − ξ t−1 = 0 which implies that any noise is unchanged during theperiod representing the policy announcement. By further assuming that E pt−1[f (Ωt)] − f (Ωt) = 0,

we are saying that the private sector’s beliefs about the information set being used by the ECB or

Fed is right. Under these two assumptions if we subtract equation (2) from (3) we get:

t Ŝ t − t−1 Ŝ t = st (4)

which means that an appropriate proxy for the shock, st, can be represented by the change in the

private sector’s belief concerning the stance of the ECB or Fed around the time of the announcement.

The question then is what data to use to represent the private sector’s beliefs. In the spirit of Soderstrom (2001), Kuttner (2001), and Faust et al. (2004), Barakchian and Crowe (2013) use the

federal funds futures contracts to represent the private sector’s beliefs, and this paper does the same,

adding the eurodollars futures contracts for the EU. Therefore any change in a particular futures

rate will be given by equation (4).

As in Barakchian and Crowe (2013) I create this shock series variable for six diff erent futures

contracts with horizons from one to six months. Using a range of maturities as opposed to simply

a one month futures contract allows the use of information from diff erent series that will perhaps

improve on the noise from looking at only one given horizon. 8

I then insert this shock series into a simple factor model estimated using maximum likelihood as

follows. Let T and N be the time series and cross-section components respectively where T will

be the sample size t = 1, 2,...,T and i = 1, 2, ...N where N = 6 represents the six diff erent futures

rates being used. The factor structure therefore of observation si,t will be:

7The authors are essentially arguing that both f () and Ω are problematic for previous US results using the CEE(1996) setup.

8Note that equation (4) needs to be weighted for the 1-month futures rate to reflect how many days in the currentmonth will be aff ected by the given change (see Kuttner (2001). Formally, the equation will be st =

M

M −d(t Ŝ t−t−1 Ŝ t)

where M is days in a given month, while d represents the date of the announcement.

6


7/63

si,t = φ0

iπt + ei,t (5)

If we let s and e be T × N matrices, the factor model in matrix form can be written as:

s = πφ0

+ e (6)

where π = (π1, π2,...πT ) is the T × r matrix of factors, and φ = (φ1, φ2,...φN )0

is the N × r factor

loading matrix.

Running this factor model for both the EU and US suggest that two factors are sufficient, the first

of which represents the exogenous monetary policy shock originating in these countries. Figure 1

shows the EU structural monetary policy shocks extracted from the factor model for the periodunder analysis, 1999q1 - 2011q4, while Figure 2 shows the EU real economy response to these

contractionary monetary policy shocks. As we see, the real economy response follows conventional

economic wisdom.

Figure 1: EU Monetary Policy Shock Series

Figure 2: EU Monetary Policy Shock and Real Economy Reaction

Figure 3 shows the structural US monetary policy shocks extracted from the factor model for the

period under analysis 1992q1 - 2008q2.9 This shock series is highly correlated with the shock series

9Ideally this paper would have looked at 1992q1 - 2011q4 for both the EU and US shocks. I start in 1999q1 for

7


8/63

produced in Barakchian and Crowe (2013). Furthermore, Figure 4 indicates that these contrac-

tionary monetary policy shocks produce the expected results for real GDP in the US.10

Figure 3: US Monetary Policy Shock Series

Figure 4: US Monetary Policy Shock and Real Economy Reaction

2.2 Stage 2 Identification

The shock series’ that we have extracted come from diff erenced futures rates, and after testing are

confirmed stationary, I(0), variables. However, to use these shock series as regressor variables in

the second stage SVAR, we need them to be I(1) variables as the SSA variables comprising the

remainder of the SVAR are in levels. I follow Romer and Romer (2004) and cumulate the US

monetary policy shock series turning it into an I(1) variable.

Formally, I assume that the SSA economy can be described by an equation with the following

structural form:

G(L)Y t = C (L)X t + et (7)

the EU due to the introduction of the Euro. I stop in 2008q2 for the US because they hit the zero lower bound and Idid not want to introduce non-linearities into the model. Further discussion on time periods will occur in Section 3.

10Note: EU and US monetary policy shock series is monthly due to the futures series’ being monthly. As theyrepresent stock series I take the last month of each quarter to create the quarterly series’.

8


9/63

where Y t is a n×1 data vector of endogenous SSA variables, Y 0

t = [ShocktTradetRGDP tCP I tXRtIRt],

X t is a data vector of foreign exogenous variables, X 0

t = [Commt Oilt], G(L) is a n×n matrix poly-

nomial in the lag operator, C (L) is a n × k matrix polynomial in the lag operator, and lastly, et is

a n × 1 vector of structural disturbances, with var(et) = Λ, where Λ is a diagonal matrix in which

the diagonal elements represent variances of structural disturbances and are by definition mutually

uncorrelated.

The corresponding reduced-form VAR to (1) can be written as:

Y t = A(L)Y t−1 + B(L)X t + εt (8)

where the A(L) and B(L) are both matrix polynomials, εt are reduced form errors with var(εt) = Σ.

If we let H be the coefficient matrix of contemporaneous terms, and let J (L) be a matrix of non-

contemporaneous terms, we get:

G(L) = H + J (L) (9)

Therefore the relationship between the structural and reduced-form equations can be written as:

A(L) = −H −1J (L) (10)

B(L) = H −1C (L) (11)

The error term relationship can be written as εt = H −1et or et = H εt, implying the following:

Σ = H −1ΛH −10

(12)

Using the Choleski or recursive decomposition, H is a triangular matrix where the proper ordering

of the variables is required. In the non-recursive modeling, maximum likelihood estimates of H and

Λ are necessary and are obtained from a sample estimate of Σ. Note that in the right hand side of equation (12) there are n × (n + 1) free parameters that require estimation. Furthermore, Σ has

n × (n + 1)/2 parameters, so if we normalize n diagonal elements of H to be 1, we require at a

minimum n × (n − 1)/2 restrictions on H in order to have full identification.

As the focus of this study is not only on real economy movements but also whether the transmission

mechanism is trade or interest rates I use the following variables for Y t:

Y 0

t = [Shockt T radet RGDP t CP I t XRt IRt] (13)

9


10/63

where Shockt represents either the EU or US monetary policy shock cumulated to be an I(1) variable,

Tradet is real bilateral trade balance with either the EU or US, RGDP t is real GDP11, C P I t is the

consumer price index, XRt, is the nominal bilateral exchange rate with either the EU or US, and

IRt represents an appropriate nominal central bank interest rate used by the given SSA country.

For the exogenous variables, I follow Cheng (2006) and use both a commodity price index which

excludes oil, as well as an oil specific index:

X 0

t = [Commt Oilt] (14)

As Bernanke and Blinder (1992) among others argue, the important insight of the SVAR research

is that one need only appropriately identify the shock of interest in order to determine the impulse

responses of other macroeconomic variables due to the mutually uncorrelated nature of the error

terms. Therefore, by having identified the large open economy contractionary monetary policy

shock in stage 1, and inserting it in the first position of the stage 2 domestic model given that it

is unlikely to be aff ected by any SSA variables contemporaneously, the ordering of the other stage

2 domestic variables is irrelevant. I therefore include the foreign monetary policy shock and other

SSA variables of interest for the given country as follows:

1 0 0 0 0 0

a21 1 0 0 0 0

a31 a32 1 0 0 0

a41 a42 a43 1 0 0

a51 a52 a53 a54 1 0

a61 a62 a63 a64 a65 1

εShock

εtrade

εrgdp

εcpi

εxr

εir

=

b11 0 0 0 0 0

0 b22 0 0 0 0

0 0 b33 0 0 0

0 0 0 b44 0 0

0 0 0 0 b55 0

0 0 0 0 0 b66

eShock

etrade

ergdp

ecpi

exr

eir

(15)

3 Results

This section will analyze the extent to which a contractionary monetary policy shock in each of

the EU and US aff ects a subset of SSA countries including floating exchange rate economies South

Africa, Tanzania, Uganda, Mauritius, Ghana, Kenya, and fixed exchange rate economies Botswana,

Rwanda, Senegal, Cote D’Ivoire, and Gabon.12 The data period analyzed diff ers depending on

whether we are looking at the EU shock or the US shock. For the EU I use the period following

the introduction of the Euro in 1999q1, up until 2011q4. For the US I use the period beginning in

1992q1, around the time capital flows began to increase towards developing countries, and I stop in

2008q2 when the US hits the zero lower bound. I do not extend past the zero lower bound as my

11May also represent a proxy as quarterly GDP figures are not available for all countries. See Appendix C for adescription.

12Appendices E and F describe the exchange rate regimes and central bank mandates of these countries over theperiod under analysis.

10


11/63

overall time period would include non-linearities which opens up other potential analytical issues.

As the EU does not hit the zero lower bound I am able to use up until 2011q4. Given the availability

of data for each country in question, the period will change slightly for a given country.13

The first question to answer is what will happen to the overall real economies of these SSA countries,

i.e. real GDP or some equivalent measure. The literature is mixed on what to expect depending

on which of substitution versus income eff ects dominate.14 Should substitution eff ects dominate,

contractionary monetary policy shocks in the EU or US will cause appreciations in their currency

relative to SSA country currency, leading to increased imports including SSA goods, resulting in

increased trade balance and GDP for these SSA countries. Should income eff ects dominate, there

will be contractions in GDP but the story of why will depend on whether trade or interest rates are

the dominant transmission mechanism. If trade is dominant then contractionary monetary policy

shocks will lead to falls in real GDP in the EU or US causing them to import less, meaning tradeand GDP will fall in SSA countries. If interest rates are dominant then increases in EU or US

interest rates will cause increases in the world interest rate, increases in SSA interest rates, and

a fall in world aggregate demand, causing ambiguous trade results but falls in GDP from falls in

domestic demand.

As Table 1 shows these countries have significant trading ties with each of the EU and US, though,

relationships with the EU are much higher for all countries.15 This may lead the researcher to

expect stronger eff ects following a EU shock with trade playing the role of important transmission

mechanism. However, as Table 2 indicates, almost all countries have significant external debt as

a percentage of GNI and much of this debt is denominated in either Euros or US dollars with theUS dollar debt representing a larger share in most. As a contractionary monetary policy shock in

the EU or US will lead to an appreciation of its currency, the debt burden of floating exchange

rate countries will increase unless they increase their interest rates as well. So the decision for

floating exchange rate countries is do you allow the EU or US currency appreciation to occur in

the hopes that trade will increase or do you respond with increases in interest rates in order to

off set the appreciation thus keeping your debt burden stable but potentially harming other parts

of your economy. Fixed exchange rate countries do not have to concern themselves with exchange

rate issues, however if interest rates increase in the EU or US, to the extent that these countries

are concerned with capital inflows, and one can see the increased importance since the beginning

of the 1990s as indicated by Figure 616, they may consider increasing their interest rates as well

to remain competitive. Given the dominance of the US dollar as the debt denomination for both

13See Appendix D for more detail on time periods for SSA countries.14Kim (2001), Kozluk and Mehrotra (2009), and Jannsen and Klein (2011) all find contractions for their small open

economies, while Canova (2005) and Mackowiak (2006) both find expansionary responses.15Despite the much larger trading relationships with the EU, the US often ranks as one of the larger trading

partners for these countries and is thus still relevant.16This foreign direct investment data reflects net inflows of investment to acquire a lasting management interest in an

enterprise operating in an economy that does not belong to the investor. This data therefore reflects committed FDI.Data comes from www.indexmundi.com courtesy of the IMF, United Nations Conference on Trade and Development,and official national sources.

11


12/63

fixed and floating countries, as well as the eff ect on capital flows that the US’ importance as the

setter of world interest rates will have, if interest rates are the primary transmission mechanism,

this may lead one to assume that US shocks will cause stronger eff ects on these SSA countries than

EU shocks.

Therefore, a priori, it is unclear whether we should expect real economy contractions for these SSA

countries, and whether trade and/or interest rates will act as a transmission mechanism. It is also

not obvious which of the EU or US shock is likely to have the larger impact. To further complicate

expectations, as Mishra and Montiel (2012) point out, domestic monetary policy transmission has

been relatively ineff ective in SSA countries, including the countries in this study. Therefore the

question is, if domestically monetary policy does not work, how can we expect international trans-

mission to have any eff ect at all? Some authors (see, e.g. Canova (2005)), though, have found that

US shocks have larger eff ects on foreign countries than those countries’ domestic monetary policieshave had on their own economies. We therefore need to investigate the empirical evidence to reach

conclusions on these questions of interest.

Percent (%) of Total Trade Exports EU Imports EU Exports US Imports US

South Africa 21.38 27.73 9.34 9.39

Ghana 36.94 22.22 8.17 7.41

Kenya 16.63 16.01 4.31 5.93

Tanzania 23.94 13.20 2.35 3.53

Uganda 25.23 17.41 2.23 3.89

Mauritius 36.94 23.06 13.90 2.50

Gabon 18.62 59.21 42.13 7.66Cote D’Ivoire 44.74 37.74 7.82 3.83

Botswana 31.27 37.29 16.09 9.03

Senegal 29.89 44.81 0.49 4.18

Rwanda 26.25 19.96 4.39 6.41

Table 1: SSA Country Trade with EU and US

12


13/63

Percent (%) of Total External Debt Euro USD Debt as % of GNI*

South Africa 10.13 69.36 22.15

Ghana 11.92 62.63 72.99Kenya 16.75 44.41 52.50

Tanzania 4.60 46.52 79.79

Uganda 5.12 59.92 53.67

Mauritius 45.39 34.67 20.79

Gabon 41.74 27.53 69.69

Cote D’Ivoire 49.06 39.17 117.10

Botswana 5.59 16.55 9.98

Senegal 15.45 43.24 59.26

Rwanda 6.15 51.36 56.43*This percentage includes private debt as well

Table 2: SSA Country External Debt Denomination

Figure 5: SSA Country FDI Inflows as % of GDP (Floating Exchange Rate Regime)

13


14/63

Figure 6: SSA Country FDI Inflows as % of GDP (Fixed Exchange Rate Regime)

3.1 EU Shock Results

3.1.1 Main Results

I begin by looking at how the contractionary monetary policy shocks from the EU aff ect the real

GDP’s of the 11 SSA countries and determine which mechanism(s), trade and/or interest rates,

are at the heart of the results. The discussion will focus on a comparison of floating versus fixed

exchange rate economies.

Tables 3 and 4 show the real GDP responses to a contractionary EU monetary policy shock, how

much of the variance in GDP can be explained by the shock, as well as if trade and/or interest

rate patterns match those we see in the real economy thus acting as transmission mechanism. The

arrow represents a significant change in real GDP in a given direction. The confidence intervals

used to determine the significance are based on asymptotic standard errors with 68% bands. The

’Y’, ’N’, imply that trade and/or interest rates do or do not act as transmission mechanisms for the

EU shock.17

17Figures for each country can be found in Appendix A

14


15/63

Real GDP Change FEVD* Trade as Transmission Mechanism Interest Rate asTransmission

MechanismSouth Africa - 1.2% N Y

Ghana ↓ 23.1% N Y

Kenya ↑ 12.0% N Y

Tanzania ↑ 7.6% Y Y

Uganda ↓ 2.6% N Y

Mauritius ↓ 9.3% N Y* Average forecast error variance decomposition over the first 8 quarters after the shock.

Table 3: Floating Exchange Rate Country Responses Following EU Shock

Real GDP Change FEVD Trade as Transmission Mechanism Interest Rate asTransmission

Mechanism

Gabon ↑ 4.3% N N

Cote D’Ivoire ↓ 5.1% Y Y

Botswana ↓ 6.5% Y Y

Senegal ↓ 6.2% Y Y

Rwanda ↓ 7.4% N N

Table 4: Fixed Exchange Rate Country Responses Following EU Shock

Table 3 indicates that floating exchange rate countries have mixed results when it comes to experi-

encing real GDP contractions following a contractionary monetary policy shock in the EU, whereas

Tables 4 indicates that fixed exchange rate countries seem to be more likely to incur economic con-

tractions. Furthermore, floating exchange rate countries are much more likely to have interest rates

as the dominant transmission mechanism, as opposed to trade. Fixed exchange rate countries seem

likely to be sensitive to both.

The implication for floating exchange rate countries seems to be that those countries who are willing

to use interest rates as an off setting tool will be successful. Instinctively one might assume that

those countries with higher debt burdens both in absolute terms and in terms of being denominatedin Euros would be more willing to increase their rates as a way to off set the increased cost from

currency appreciation. However, there is no clear pattern in Table 2 indicating that more debt

denominated in Euros leads to a more likely interest rate increase. It does appear though that the

countries that are increasing their interest rates, thus causing an economic slowdown, are those that

are more concerned with capital inflow competitiveness. Figure 5 shows that Ghana, Uganda, and

Mauritius have been the most positively aff ected by financial liberalization since the early 1990s,

and have higher FDI/GDP percentages in general than the other floating countries.18 They may

18Tanzania also has high capital flows and, despite their initial increase in GDP, they experience a quick return to

15


16/63

therefore be more sensitive to capital flow reversals leading them to increase their interest rates in

an anticipatory response.

Fixed exchange rate countries, unlike their floating exchange rate counterparts, will not be able to

experience a depreciation in their currency relative to the Euro that would allow them to off set

some of the negative income eff ect on trade stemming from the EU. Increasing interest rates seems

to be an attempt to counter the fall in trade by causing prices to decrease making SSA goods

cheaper given the fixed exchange rates. However, as seen in Figure 24 in Appendix A, the success

of lowering CPI is mixed. These countries may also be increasing their interest rates in an attempt

to keep capital flow competitiveness strong. In any case, the increase in interest rates compounds

the contractionary GDP eff ect at home.

3.1.2 Floating and Fixed Exchange Rate Country Examples

To put the general results in context I show a floating and fixed exchange rate country example.

Specifically, I use Uganda for the floating exchange rate countries that experience GDP contractions,

and Senegal for the fixed exchange rate countries.

As can be seen in Figure 7, Uganda experiences an immediate contraction, falling by approximately

0.2%. The recovery is slow, not returning to the zero line for almost four years. Most papers looking

at domestic transmission in Uganda find insignificant eff ects on real GDP (see e.g. Cheng (2006)).

The implication is then that international transmission of monetary policy shocks seems to have

a larger eff ect than domestic shocks of the same nature, consistent with other papers includingCanova (2005).

Trade is clearly not a strong transmission mechanism as it moves in almost the complete opposite

direction as GDP. Interest rates on the other hand seem to be a strong transmission mechanism as

the approximately 50 bps initial increase causes GDP to fall while the subsequent lowering seems

to help GDP begin to recover. The increase in interest rates seems to be due to Uganda’s desire

to keep capital flows from leaving. However, upon seeing the economy shrink, the central bank

responds with a lowering of these rates.

Figure 7: Uganda Economic Response to EU Shock

the zero line due to an increase in interest rates.

16


17/63

For Senegal, a fixed exchange rate country, we again see a fall in GDP, this time of approximately

0.5%, a larger fall than for Uganda, in line with the idea that floating exchange rates should work as

a measure of protection (see, e.g. Canova (2005)). Trade seems to work as a transmission mechanism

with a bit of a delay. Trade increases in period 1 as does GDP. However the falls in trade in periods

2 and 3 do not cause declines in real GDP until periods 3 and 4. In the later periods trade tends

towards zero as does GDP. Interest rates also appear to work as a transmission mechanism. The

initial and period 1 decreases in interest rates totaling approximately 8 bps cause the quick rebound

in GDP. This quicker recovery is due to both an increase in trade in period 1 as well as the fact that

fixed exchange rate countries are able to lower their interest rates without creating more expensive

external debt levels. Uganda seems to be more aff ected by this more expensive debt when they

lower rates leading to a slower recovery. So while fixed exchange rate countries are more aff ected

by the initial fall, they are able to experience quicker recoveries.

Figure 8: Senegal Economic Response to EU Shock

3.2 US Shock Results

3.2.1 Main Results

Tables 5 and 6 below show a similar analysis to the one performed for the EU.


Mechanism

South Africa ↓ 1.7% N N

Ghana ↓ 28.4% N N

Kenya ↑ 6.4% Y Y

Tanzania ↓ 3.7% Y Y

Uganda ↓ 5.6% Y Y

Mauritius ↓ 17.9% Y Y

Table 5: Floating Exchange Rate Country Responses Following US Shock

17


18/63


MechanismGabon - 1.2% N N

Cote D’Ivoire ↑ 24.6% N N

Botswana ↑ 8.0% N N

Senegal ↑ 1.1% N N

Rwanda - 0.6% Y N

Table 6: Fixed Exchange Rate Country Responses Following US Shock

The results indicate that floating exchange rate countries can expect economic contractions following

a US contractionary monetary policy shock in contrast to the mixed results following the EUshock. For these countries, both trade and interest rate seem to play important roles in real GDP

movements. The story seems to indicate that the US contractionary monetary policy shock leads

to a fall in trade as the income eff ect dominates the US economy. In an attempt to keep the debt

burden from becoming more expensive due to an appreciation of the US currency, these countries

respond with an increase in their interest rate further damaging the economy. The implication for

these countries is that diversifying both trading relationships and, more importantly, debt burden,

could help these countries avoid a declining real economy response to a contractionary monetary

policy shock in the US.19

For fixed exchange rate countries, the story is vastly diff

erent from the EU shock. Whereas followingthe EU shock these countries could expect a contraction with interest rates and trade working

as transmission mechanisms, in the case of the US shock these countries are able to avoid the

contraction and it appears that neither trade or interest rate is a transmission mechanism. The

relatively smaller trade relationships with the US compared to the EU could help explain trade’s

insignificance, though the EU/US ratio isn’t significantly diff erent for these countries compared with

the floating exchange rate countries. It could be that some other non-investigated factor is playing

a role here. One example might be aid. The US, as of 2011, still spends approximately 30% of all its

aid relief on SSA countries. Therefore, despite interest rate and trade movements suggesting GDP

should fall and it does not, it appears aid or some factor like it is overwhelming any trade or interest

rate eff ects. The implication is that either the pivot the US made in the early 2000s from aid totrade (see, e.g. AGOA) has not been successful as of yet for fixed exchange rate countries, or aid

continues to flow in at too high levels to allow trade and interest rates to play a significant role. As

US aid to SSA countries has remained relatively stable even during US recessions, contractions in

the US may not lead to a lowering of aid, which might help explain the expansionary real economy

responses in some of these SSA countries.20

19I say ’more importantly’ because if these countries did not increase interest rates they may have been able toenjoy some of the off setting currency eff ects on trade similar to the case following the EU shock.

20Aid data comes from US AID.

18


19/63

Note that there is a reversal in negative response of real GDP to US shocks from the case of EU

shocks. Whereas fixed exchange rate countries were more negatively impacted by EU shocks com-

pared with floating exchange rate countries, the reverse is true following the US shocks. Following

the EU shock, this result is explained for fixed exchange rate countries by their higher reliance on

the EU as their currency is fixed to the Euro and they have large trading relationships with them.

Following the US shock, the result for floating exchange rate country sensitivity is explained by

the importance of the debt burden and the increase in cost of this debt should the US currency

appreciate relative to the SSA currency.

3.2.2 Floating and Fixed Exchange Rate Country Example

As before, I look at floating and fixed exchange rate country examples to put this discussion incontext. I use Tanzania for the floating exchange rate countries and Gabon for the fixed exchange

rate countries.

As Figure 9 indicates, Tanzania experiences an immediate contraction of around 0.2% matching

Uganda’s contraction following the EU shock. There is again a slow recovery for the economy. Other

papers (see e.g. Buigut (2009)) on Tanzania have found insignificant eff ects on the real economy

from a domestic monetary policy shock implying that international transmission of monetary policy

shocks is again a stronger eff ect.

Trade is a strong mechanism in this case falling early on before recovering. Interest rates also act

as a strong mechanism as the initial and period 1 increase totaling approximately 100 bps seemsto contribute to the fall in the real economy, while the subsequent lowering of interest rates at a

steady pace causes the slow recovery.

Figure 9: Tanzania Economic Response to US Shock

As Figure 10 indicates, Cote D’Ivoire does not experience a contraction, in fact they experience an

expansion following the shock before quickly falling back down to the original starting point. As can

be seen neither trade or interest rates explain the GDP pattern that we see. Trade experiences an

increase, though barely significant, after GDP is already back down to its initial level. The initial

interest rate fall might help explain some of the increase in GDP, however, at a decrease of only 1 or

2 bps it is hard to imagine that this is a major cause. Further, the subsequent increase in interest

19


20/63

rates occur after GDP has already fallen back to its initial levels. Here is where aid may play a role.

US aid to Cote D’Ivoire has been relatively fixed, even during the recession since 2008, so despite

the contraction in the US, aid to Cote D’Ivoire may have been sufficient to allow for the initial

increase in real GDP before slowing down and causing GDP to go back to initial levels. Despite

returning to initial levels, aid may have been enough to off set any negative trade or interest rate

eff ects.21

Figure 10: Cote D’Ivoire Economic Response to US Shock

4 Robustness Checks

My first robustness check looks at overall trade balances instead of the bilateral trade data that

has been used to this point. Bilateral trade was used in the primary specification due to the size

of the trading relationships with the EU and US. Overall trade balances will allow us to determine

whether additional trading partners play a strong enough role to eff ect the main conclusions.

My next robustness check then looks at whether the overparameterization issue of the block-block

method is significant enough to change any of the primary results using this methodology only as

it concerns the EU shock. In this model, a SVAR methodology is used for both blocks which will

create a shock series with a diff erent genesis than the one in the primary specification. The factor

model cannot be used in a block-block setting so we turn to the SVAR setup to extract the foreign

exogenous monetary policy shock series. The reason this analysis is only done for the EU shock is

due to the fact that, as Barakchian and Crowe (2013) show, generating the exogenous shock series

for the US using the CEE (1996) SVAR setup produces a real economy response that is inconsistent

with theory.22 Running this setup for the EU does not produce the same puzzle. Therefore, not

only can we analyze how the overparameterization issue plays out, we are also able to compare how

the shock series from the factor model and the shock series from the SVAR model aff ect the SSA

countries, given that they both act as proper monetary policy shocks judged by the results on the

EU itself.

Lastly, as an extension I create aggregate measures for the floating and fixed exchange rate countries

21Aid to Cote D’Ivoire from the US in 2011 was $133 million which represented 0.55% of GDP. Furthermore, totalaid was $1.4 billion which amounted to 7.92% of GDP.

22Barakchian and Crowe (2013) also show that other well-known setups also produce the same real economy puzzle.

20


21/63

using geometric means as in Kim (2001).23 This will allow me to see whether generalizing results

for a typical floating and fixed country matches the findings when we look at countries individually.

4.1 Overall Trade Balance - EU Results

Tables 7 and 8 show the results for both the floating and fixed exchange rate countries following

the EU shock.24


Mechanism


Ghana ↓ 15.0% Y NKenya ↑ 8.3% N Y

Uganda ↓ 4.1% Y Y




Mechanism

Gabon ↑ 14.3% N Y

Cote D’Ivoire ↓ 7.7% N NBotswana ↓ 6.9% N Y

Senegal - 1.6% Y Y

Rwanda ↓ 10.9% N N


Table 7 indicates that, for floating exchange rate countries, we can expect real economic contractions

following the EU contractionary monetary policy shock. Whereas there were mixed results in the

primary specification, contractions are more assured when including overall trade balances. Trade

is stronger as a transmission mechanism while interest rates remain important. The implication isthat the countries that used interest rate policy countercyclically in the primary specification are

not this time. By including trade relationships with all countries, these floating countries must now

concern themselves with the interest rate response of all countries not just the EU. If all countries

raise their interest rates following the EU shock it is not just the trade, debt, and capital flow

relationships with the EU that matter, it is with all countries. Perhaps anticipating that a large

open economy like the EU will have a strong eff ect on most countries’ interest rates, they respond

23For interest rates and trade balance, arithmetic mean is used.24Note that Tanzania is left out of the analysis. This is due to weak real eff ective exchange rate data.

21


22/63

as well with an increase in their rates in order to keep capital inflows competitive, to the detriment

of the real economy both through increased interest rates and off setting any exchange rate eff ects

that may slow down the income eff ect in the EU that drives down trade. In comparing Tables 3

and 7 one notices that not too much in-country changes, except for trade.

For fixed exchange rate countries, the story for real GDP remains the same in the sense that we

see contractions in general. However, including other countries’ trade relationships with these fixed

exchange rate countries, the negative trade eff ect seems to be mitigated. Interest rates are increased

in order to remain competitive in the capital flows market. Comparing Tables 4 and 8 shows very

little in-country changes to the results.

4.2 Overall Trade Balance - US Results

Tables 9 and 10 show the results for both the floating and fixed exchange rate countries following

the US shock.


Mechanism


Ghana ↓ 9.2% Y N

Kenya ↑ 3.8% N Y

Uganda ↓ 6.5% Y Y

Mauritius ↓ 32.2% N Y

Table 9: Floating Exchange Rate Country Responses Following US Shock


Mechanism

Gabon ↑ 4.8% Y Y

Cote D’Ivoire ↑ 35.3% N N

Botswana ↑ 7.0% Y N

Senegal ↑ 13.7% Y YRwanda ↓ 10.1% N N

Table 10: Fixed Exchange Rate Country Responses Following US Shock

Table 9 indicates that, for floating exchange rate countries, we can expect real economic contractions

following the US contractionary monetary policy shock. This result is consistent with the primary

specification. Furthermore, interest rates remain a strong transmission mechanism while trade

becomes a bit weaker. Trade becomes a bit weaker because by including other countries’ trade

relationships with these floating exchange rate countries, I am allowing for these new countries to

22


23/63

expand trade with SSA when it becomes more expensive for them to import from the US due to the

US dollar appreciation following the US shock. This increased trade helps off set the fall in trade

from the US. However, as GDP contractions are still prominent the interest rate is a dominant

mechanism. Interest rates are increased in order to keep debt levels from increasing and ensure

capital flow competitiveness. Comparing Tables 5 and 9 shows very little movement of in-country

results for these countries except for trade.

For fixed exchange rate countries, they continue to be able to avoid a real economic contraction

following the US shock but trade now becomes more helpful as a transmission mechanism as other

countries import more due to the increased cost of US goods. Following the US shock, fixed exchange

rate countries remain less negatively impacted by the foreign monetary policy shock compared with

the floating exchange rate countries as in the primary specification. Again, there are little in-country

changes to the results except for trade.

4.3 Block-Block Method - EU Results

The block-block methodology contains a large open economy block, which uses the CEE (1996,

1999) monetary policy shock identification method, as well as an SSA block containing the same

SSA variables as in the primary specification. As we know from Barakchian and Crowe (2013) that

the CEE (1996) setup produces a real economy puzzle for the US, we perform this analysis for the

EU shock only.

Formally, the bivariate block VAR model can be written as:

" yit

wt

# =

" Ai

11(L) Ai

12

0 A22(L)

# " yit−1

wt

#+

" Ai

13(L)

A23(L)

#xt +

" εit

et

# (16)

where the error terms have the following property: (εit, et)0 ∼ (0,Σi), where Σi = blockdiag(Σεi ,Σe).

We can rewrite the structural model underlying this reduced-form version as:

" Gi0 Ai00 H 0

# " yitwt

# =

" Bi11(L) Bi120 B22(L)

# " yit−1wt

#+" 0

B24(L)#

x1t+

" Bi13(L)0

#x2t++

" uitvt

#(17)

where the error terms now have a diff erent property, namely (uit, vt)0

∼ (0, I ). In this setup, wt

represents the EU block variables, while yit represents the given SSA country block. Exogenous

variables are represented by xit.25,26

25The zero in the bottom left-hand corner of the LHS matrix reflects the fact that SSA variables are unlikely tohave any eff ect on the EU variables.

26In order to reduce estimation time, I run this block-block model in two stages where the EU model is run first,the shock series is extracted and cumulated, and inserted into the SSA block similar to the primary specificationmethod.

23


24/63

The variables in the EU block are as follows:

w0

t = [y cpi index ir] (18)

where y represents real GDP, cpi is a consumer price index, index is a commodity price index

which includes all commodities including oil, and ir is the EU 3-month interbank rate, EURIBOR.

As can be seen in Figure 12, the results on the EU economy from a EU monetary policy shock

are as predicted by economic theory, specifically a real economy contraction from a contractionary

monetary policy shock.

Figure 11: EU Monetary Policy Shock Series

Figure 12: EU Monetary Policy Shock and Real Economy Reaction

The variables in the SSA block are the same as the ones in the primary specification where the

shock is generated from estimation of the EU SVAR block:

y0

it = [Shockt T radet RGDP t CP I t XRt I Rt] (19)

The exogenous variables aff ecting the EU, x1t, include the US federal funds rate. For the SSA

countries, the exogenous variables, x2t, are represented by two separate commodity price indices,

specifically one that includes all commodities other than oil and one that contains only oil.

Tables 11 and 12 show the results for floating and fixed exchange rate countries when the block-block

analysis is performed.

24


25/63


MechanismSouth Africa - 1.1% N N

Ghana ↓ 4.6% N Y

Kenya - 0.9% N Y

Tanzania - 1.6% N Y

Uganda ↓ 4.6% N Y




Mechanism

Gabon ↓ 8.6% N Y

Cote D’Ivoire ↓ 2.9% N Y

Botswana ↑ 14.4% Y Y

Senegal ↓ 8.4% Y Y

Rwanda ↓ 7.4% Y N


Similar to the primary specification, floating exchange rate countries have mixed results when it

comes to experiencing real GDP contractions whereas fixed exchange rate countries seem to be more

negatively impacted. This can be reflected not only in the direction of the GDP movement but also

in how much of the change in real GDP can be explained by the shock. For fixed exchange rate

countries the FEVD is in general higher for these countries compared with the floating countries.

Furthermore, consistent with the primary specification results, floating exchange rate countries are

more likely to have interest rates as the dominant transmission mechanism, instead of trade whereas

fixed exchange rate countries are sensitive to both. Given that the main results of the block-block

model are all consistent with the primary specification results the economic story remains the same.

The implication of this similarity, along with the fact that the EU economy responds as expected

to a contractionary monetary policy shock, is that it does not seem to matter if we generate the

EU shock series using the CEE (1996) setup or with the factor model Barakchian and Crowe (2013)

setup. The reason for this could be due to the fact that the ECB, relative to the Federal Reserve, is

in its early stages and may not be as forward-looking as the Federal Reserve has become. Therefore,

in the same way that the CEE (1996) setup worked for the US for many years it still works for the

EU. Going forward however, as the ECB develops, a shock series such as I have generated using

futures rates in a factor model may become more appropriate.

25


26/63

4.4 Typical Floating and Fixed Country - EU Results

Many papers that have investigated multiple countries’ reaction to a foreign shock have used atypical country approach whereby they create aggregate measures using the geometric mean of

their variables of interest.27 This allows them to investigate one set of results that represent a

typical country in their analysis as opposed to looking at all countries individually. I create both

a floating and fixed exchange rate group using the same division of countries for each group as in

the rest of the paper. Given my investigation of individual countries I can compare how looking at

these countries as one entity changes my conclusions.

Figures 13 and 14 detail the results for both groups following a contractionary monetary policy

shock in the EU.

Figure 13: Floating Exchange Rate Typical Country Economic Responses to EU Shock

Figure 14: Fixed Exchange Rate Typical Country Economic Responses to EU Shock

As Figure 13 shows for floating exchange rate countries, a GDP contraction does not occur. In the

main specification, we saw mixed results with some countries experiencing contractions while others

saw expansions. Clearly, if we had simply averaged the results we would have been left to believe

expansions would occur. Furthermore, the results for the transmission mechanisms are completely

reversed. Trade is a strong transmission while interest rates are not when we look at a typical

floating exchange rate country, which is in direct contrast to the primary specification.

27Kim (2001), Canova (2005), Mackowiak (2006, 2007) among others.

26


27/63

Figure 14 for fixed exchange rate countries indicates that while there is fall in the real economy

towards the end of the sample, the first two years economic expansions are occurring. This is

opposite to the contractionary conclusion reached from the individual country analysis. Both trade

and interest rates are transmission mechanisms which match the finding in the primary specification.

The implication of both the floating and fixed exchange typical country results following the EU

shock is that if we look at these countries only under the lens of aggregate measures we are likely to

conclude in ways that may not match generalizations formed by looking at each country individually.

Large diff erences between these countries, even within each group, causes aggregating to mask these

country idiosyncrasies.

4.5 Typical Floating and Fixed Country - US Results

Figures 15 and 16 detail the results for typical countries in both sets of exchange rate regimes

following a contractionary US monetary policy shock.

Figure 15: Floating Exchange Rate ’Typial’ Country Economic Responses to US Shock

Figure 16: Fixed Exchange Rate ’Typial’ Country Economic Responses to US Shock

As Figure 15 indicates for floating exchange rate countries, one can expect immediate and long-

lasting economic contractions following the US shock. The significant contractions is consistent with

the individual country results, with the length of the contraction being an additional trait, similar

to South African results. Trade patterns identically match GDP patterns indicating the importance

of trade as a transmission mechanism. Interest rates appear to have a strong initial eff ect, and

27


28/63

the increase in period 1 seems to limit the increase in GDP coming from strong trade increases.

The subsequent patterns line up as well so it appears that interest rates work as a transmission

mechanism in addition to trade. Both trade and interest rates as transmission mechanisms is

consistent with the primary specification.

Fixed exchange rate countries also experience an economic contraction though with a delay that

follows a small uptick. This contraction is not consistent with the primary specification. Trade is

not a mechanism for these countries consistent with the primary specification, while interest rates

are, which is a departure. Therefore, fixed exchange rate countries are most aff ected following US

monetary policy shocks by generalizing results into a typical country. Large diff erences across these

countries could again explain the problems arising from aggregation.

5 Conclusion

This paper looks at how two diff erent large open economy monetary policy shocks, specifically the

EU and US, aff ects a set of SSA economies including floating exchange rate countries South Africa,

Ghana, Kenya, Tanzania, Uganda, and Mauritius, and fixed exchange rate economies Gabon, Cote

D’Ivoire, Botswana, Senegal, and Rwanda. I use a two-stage model where the first stage involves

extracting the contractionary monetary policy shock series for the EU and US from a factor model

in the spirit of Barakchian and Crowe (2013), cumulating the series so it is I(1), and using it as

a regressor as in Romer and Romer (2004) in the second stage SSA economy SVAR. This paper

uses thinly investigated countries to shed light on the debate in the literature as to whether there

is a positive or negative correlation between the foreign monetary policy shock and domestic real

economy response, as well as whether trade and/or interest rates play a role in the international

transmission of the shock.

Results indicate that there are diff erences depending on who the originator of the shock is, i.e. the

EU or US, and whether the country has a floating or fixed exchange rate regime. Following the EU

shock, floating exchange rate countries experience mixed results with some incurring real economic

contractions while others do not. As interest rate is the only transmission mechanism, it appears

that countries who have more reliance on capital flows are the ones increasing their interest rates

in response to the foreign shock, thus enhancing their capital flow competitiveness, but sacrificingreal economic growth. Those that do not experience economic contractions use countercyclical

interest rate policy eff ectively to off set the shock. For fixed exchange rate countries, they are in

general sensitive to the foreign shock and experience real economic contractions. Both interest rates

and trade act as transmission mechanisms. Their inability to aff ect exchange rates causes them to

increase interest rates in an attempt to lower prices in order to off set some of the fall in trade from

income eff ect dominance in the EU. However, prices are lowered with mixed success leading to fall

in real GDP from both a continued decrease in trade and a lowering of domestic demand due to the

interest rate hike.

28


29/63

Following the US shock, the floating exchange rate countries are more sensitive to the foreign shock,

in general experiencing contractions with both trade and interest rates working as transmission

mechanisms. These countries are less willing to use countercyclical interest rate policy as the

increase in US interest rates will lead to an appreciation of the US currency making debt burdens

more expensive unless they increase domestic interest rates. As most of these countries have the

bulk of their debt denominated in US dollars they are more sensitive to this issue. Fixed exchange

rate economies are less eff ected by US monetary policy shocks, experiencing GDP expansions or

insignificant GDP changes. As neither interest rates or trade play a transmission role, some other

factor, such as aid, plays a dominant role in preventing economic contractions.

The policy implications are vast. For floating exchange rate countries, both EU and US shocks can

possibly cause economic contractions with these countries being more sensitive to US shocks due to

high levels of debt denominated in US dollars. Therefore, attempts to diversify debt may allow thesecountries to use countercyclical interest rate policies as a means of avoiding a real economy slowdown

from a contractionary monetary policy shock. abroad For fixed exchange rate countries, their

reliance on the EU both in terms of their currency peg, and large trading relationships, makes them

sensitive to contractions there, whereas the same is not true with the US. Therefore, consideration

should be paid to whether a fixed exchange rate regime is the best policy moving forward as it

leads to a lack of flexibility when dealing with other countries where significant relationships exist;

a situation that is more likely to exist with the US as these countries develop.

29


30/63

A Primary Specification Figures

A.1 Macroeconomic Responses to EU Monetary Policy Shock - Floating Fixed

Exchange Rate Countries.

Figure 17: SSA Country GDP Response to EU Monetary Policy Shock (Floating Exchange RateRegime)

Figure 18: SSA Country GDP Response to EU Monetary Policy Shock (Fixed Exchange RateRegime)

30


31/63

Figure 19: SSA Country Trade Balance Response to EU Monetary Policy Shock (Floating ExchangeRate Regime)

Figure 20: SSA Country Trade Balance Response to EU Monetary Policy Shock (Fixed ExchangeRate Regime)

31


32/63

Figure 21: SSA Country Interest Rate Response to EU Monetary Policy Shock (Floating ExchangeRate Regime)

Figure 22: SSA Country Interest Rate Response to EU Monetary Policy Shock (Fixed ExchangeRate Regime)

32


33/63

Figure 23: SSA Country CPI Response to EU Monetary Policy Shock (Floating Exchange RateRegime)

Figure 24: SSA Country CPI Response to EU Monetary Policy Shock (Fixed Exchange RateRegime)

33


34/63

Figure 25: SSA Country Exchange Rate Response to EU Monetary Policy Shock (Floating ExchangeRate Regime)

Figure 26: SSA Country Exchange Rate Response to EU Monetary Policy Shock (Fixed ExchangeRate Regime)

34


35/63

A.2 Macroeconomic Responses to US Monetary Policy Shock - Floating Fixed

Exchange Rate Countries.

Figure 27: SSA Country GDP Response to US Monetary Policy Shock (Floating Exchange RateRegime)

Figure 28: SSA Country GDP Response to US Monetary Policy Shock (Fixed Exchange RateRegime)

35


36/63

Figure 29: SSA Country Trade Balance Response to US Monetary Policy Shock (Floating ExchangeRate Regime)

Figure 30: SSA Country Trade Balance Response to US Monetary Policy Shock (Fixed ExchangeRate Regime)

36


37/63

Figure 31: SSA Country Interest Rate Response to US Monetary Policy Shock (Floating ExchangeRate Regime)

Figure 32: SSA Country Interest Rate Response to US Monetary Policy Shock (Fixed ExchangeRate Regime)

37


38/63

Figure 33: SSA Country CPI Response to US Monetary Policy Shock (Floating Exchange RateRegime)

Figure 34: SSA Country CPI Response to US Monetary Policy Shock (Fixed Exchange Rate Regime)

38


39/63

Figure 35: SSA Country Exchange Rate Response to US Monetary Policy Shock (Floating ExchangeRate Regime)

Figure 36: SSA Country Exchange Rate Response to US Monetary Policy Shock (Fixed ExchangeRate Regime)

39


40/63

B Robustness Check Figures

B.1 Overall Trade Balance Robustness Check Figures

B.1.1 Macroeconomic Responses to EU Monetary Policy Shocks - Floating and Fixed

Exchange Rate Countries

Figure 37: SSA Country GDP Response to EU Monetary Policy Shock (Floating Exchange Rate

Regime)


40


41/63



41


42/63



42


43/63



43


44/63



44


45/63

B.1.2 Macroeconomic Responses to US Monetary Policy Shocks - Floating and Fixed


Figure 47: SSA Country GDP Response to US Monetary Policy Shock (Floating Exchange RateRegime)

Figure 48: SSA Country GDP Response to US Monetary Policy Shock (Fixed Exchange RateRegime)

45


46/63

Figure 49: SSA Country Trade Balance Response to US Monetary Policy Shock (Floating ExchangeRate Regime)

Figure 50: SSA Country Trade Balance Response to US Monetary Policy Shock (Fixed ExchangeRate Regime)

46


47/63

Figure 51: SSA Country Interest Rate Response to US Monetary Policy Shock (Floating ExchangeRate Regime)

Figure 52: SSA Country Interest Rate Response to US Monetary Policy Shock (Fixed ExchangeRate Regime)

47


48/63

Figure 53: SSA Country CPI Response to US Monetary Policy Shock (Floating Exchange RateRegime)

Figure 54: SSA Country CPI Response to US Monetary Policy Shock (Fixed Exchange Rate Regime)

48


49/63

Figure 55: SSA Country Exchange Rate Response to US Monetary Policy Shock (Floating ExchangeRate Regime)

Figure 56: SSA Country Exchange Rate Response to US Monetary Policy Shock (Fixed ExchangeRate Regime)

49


50/63

B.2 Block-Block Robustness Check Figures

B.2.1 Macroeconomic Responses to EU Monetary Policy Shocks - Floating and Fixed


Figure 57: SSA Country GDP Response to EU Monetary Policy Shock (Floating Exchange RateRegime)


50


51/63



51


52/63



52


53/63



53


54/63



54


55/63

C Data

Most of the data, especially for the SSA countries comes from the IMF’s International Financial

Statistics (“IFS”) database.

Creating a measure of real aggregate output in SSA countries is a difficult task given the lack

of data availability and the reliability of the data that does exist. However, there are some SSA

countries in which quarterly data for real GDP is available, including Botswana, Mauritius, and

South Africa. Furthermore, annual GDP data exists for Gabon, Cote D’Ivoire, and Senegal. I then

use quarterly industrial production index (“IPI”) data for these countries in order to interpolate

the annual GDP data creating a proxy quarterly GDP dataset.28 In addition, a paper by Opoku-

Afari and Dixit (2012) creates quarterly GDP data for Kenya, Rwanda, Tanzania, and Uganda

using high-frequency data to construct a comprehensive index of economy activity. Lastly, Ghana’scentral bank has created a real composite index of economic activity that is used in this paper

without interpolation due to the lack of annual GDP data during the middle years of the sample.

My primary specification first looks at the strength of the trade balance as a transmission mechanism

for the foreign monetary policy shock. Bilateral trade balance data (exports and imports) was

obtained from the IFS. The data was taken from the perspective of the EU and US, and is all in

USD.29Given the influence of large open economies such as the EU or US on the trade decisions of

other countries, I also check to see whether world trade data changes any of the primary specification

conclusions. Global exports, imports, and trade balance data also comes from the IFS.

The domestic interest rate diff ers by country. All interest rates were gathered from IFS. They tendto be broken down in two categories: the central bank discount rate, and a money market rate. In

industrial countries the rate generally used in transmission of monetary policy shocks work is some

form of short-term rate (usually a money market rate) set directly or indirectly by the central bank,

used by banks in interbank trading (e.g. FFR). In some countries (e.g. Canada) a target band is

set with the interbank trading rate in the middle and the upper and lower parts equidistant from

the middle. The upper band is the amount the central bank charges directly to the banks (the

central bank discount rate). Depending on availability for the SSA countries I use either the money

market rate or the central bank rate.30 I do not use real interest rates in SSA countries because I

am evaluating how a nominal interest rate shock from a foreign country aff ects an interest rate in

the SSA countries and I want to compare similar items. However, controlling for price level with

28I use the Denton least squares method for interpolation. Using IPI to interpolate GDP data has been performedin other papers including Owyang et al. (2013). The close relationship between IPI and GDP, as noted by Nilsson(1987) and others, has indicated an appropriateness to interpolating in this fashion.

29Trade balance data for the EU is for all European Union countries. Ideally only Euro Area countries would havebeen used, as is the case for all other variables, but this data was not available in aggregate and, given the changingnature of the Euro Area over time, constructing the data could have potentially lead to measurement error. Further,as the European Union acts as one trading block with a common external tariff , perhaps trade data on all EuropeanUnion countries is more appropriate.

30Used money market rates where possible. If not used central bank discount rate. Exception: Kenya where alending rate is used as there is not enough data for central bank rates. During the time data is available for thecentral bank rate the correlation is high at approximately 0.65.

55


56/63

nominal interest rates allows us to look at only the exogenous non-price reaction of the interest

rate.31

Two diff erent exchange rates are used. In the primary specification where the bilateral trade balance

is analyzed, the nominal bilateral exchange rate (domestic currency per US dollar or Euro) is used.

In the extension using global trade data, the real eff ective exchange rate (“REER”) is used. Since

trade is in real terms, exchange rate should usually be as well. However, since it is the nominal

bilateral exchange rate that countries with pegs are looking at, and are adjusting to, when they

make nominal interest rate changes, this paper uses it when evaluating bilateral trade data. In the

case of global trade, an eff ective exchange rate is necessary, and here it makes sense to use the real

eff ective exchange rate as it works like a price on goods. Bilateral exchange rate comes from IFS.

The data comes in currency per USD so I also obtain the USD/Euro relationship from the IFS and

make the conversion for analyzing the EU shocks. The data for the REER comes from a workingpaper, Darvas (2012a,b). The REER is CPI-based and relies on a wide range of trading partners.32

To account for foreign inflationary shocks, I include the world commodity price index from IFS

which includes all primary commodities including fuel. In the SSA model I use a commodity price

index without oil and an oil-specific index to separate out the eff ects. Including a foreign inflationary

variable has been shown by Kim and Roubini (2000), among others, to prevent some puzzles from

appearing in the results.

GDP, trade balance, exports, and imports are all in real terms.33 All are deflated by the GDP

deflator, where possible. If GDP deflator is not available then variables are deflated by CPI. All

variables are in logarithms except for interest rates (in percent per annum terms) and trade balance.

For global trade balance data I put it in terms of percent of GDP averaged over the sample.34,35

The foreign variables (i.e. EU and US variables) include real GDP, CPI, and an appropriate interest

rate. For the US, the real GDP is calculated using nominal GDP and the GDP deflator, both

obtained from the St. Louis FED. US CPI is also obtained from the St. Louis FED. The interest

rate for the US is the federal funds rate obtained from IFS. For the EU, the real GDP is also

calculated using the nominal GDP and deflat

Large Open Economy Monetary Policy Shocks and the Developing World

Documents

Transcript of Large Open Economy Monetary Policy Shocks and the Developing World