Accepted Manuscript

Asymmetric real exchange rates and poverty: The role of

remittances

Nicholas Apergis, Arusha Cooray

PII: S1566-0141(17)30356-4
DOI: doi:10.1016/j.ememar.2018.02.001
Reference: EMEMAR 544
To appear in:
Received date: 14 September 2017
Revised date: 18 January 2018
Accepted date: 13 February 2018

Please cite this article as: Nicholas Apergis, Arusha Cooray , Asymmetric real exchange
rates and poverty: The role of remittances. The address for the corresponding author
was captured as affiliation for all authors. Please check if appropriate. Ememar(2017),
doi:10.1016/j.ememar.2018.02.001

This is a PDF file of an unedited manuscript that has been accepted for publication. As
a service to our customers we are providing this early version of the manuscript. The
manuscript will undergo copyediting, typesetting, and review of the resulting proof before
it is published in its final form. Please note that during the production process errors may
be discovered which could affect the content, and all legal disclaimers that apply to the
journal pertain.
 ACCEPTED MANUSCRIPT

Asymmetric real exchange rates and poverty: The role of remittances

Nicholas Apergis

Department of Banking & Financial Management, University of Piraeus, Greece

[email protected]

PT
Arusha Cooray

RI
Business School, University of New South Wales, Australia

SC
[email protected]
NU
MA
E D
PT
CE
AC

Acknowledgment: The authors need to express their gratitude to the Editor of the journal for

his valuable comments and suggestions that enhanced the merit of this work. Needless to say,

the usual disclaimer applies.

1
 ACCEPTED MANUSCRIPT

Asymmetric real exchange rates and poverty: The role of remittances

ABSTRACT

This paper explores the asymmetric effect of real exchange rate changes on poverty through the

remittance channel for a panel of 99 countries, spanning the period 1980-2015. Considering a

PT
threshold partial adjustment modelling approach, the results document that real exchange rate

depreciations exert a stronger positive effect on poverty through remittances. The results are expected

RI
to be of substantial importance in the case of emerging and developing countries in designing

SC
exchange rate and inflation policies that affect the poverty levels of their population through the

mechanism of remittances.
NU
Keywords: real exchange rates, remittances, poverty, asymmetric effects
MA

JEL Classification: F24, F31, O11

E D
PT
CE
AC

2
 ACCEPTED MANUSCRIPT

1. Introduction

Remittance inflows into the developing economies have increased over tenfold from US$

31,058 million to US$ 581,640 over the 1990 to 2015 period, overtaking overseas

development aid (ODA) flows, and accounting for the second largest foreign exchange

inflow next to foreign direct investments (World Bank, 2015). Remittances account for a very

PT
large share of GDP in certain smaller economies, i.e., 36% of GDP in Tajikistan, 29% of

GDP in Lesotho, 25% in Samoa, 23% in Moldova, 21% in Kyrgyzstan, and 20% in Tonga,

RI
Nepal and Lebanon. Remittances account for 10% and above of GDP in 26 low and middle

SC
income economies contributing to the promotion of development in many of these countries

(World Bank, 2016). Accordingly, remittances have been termed the ‘new development
NU
mantra’ (Kapur, 2005), while unlike foreign direct investments and foreign aid flows,

remittances are received directly by households, and therefore, are a ‘bottom up’ source of
MA

development finance (De Haas, 2005). The magnitude and significance of these flows has led

to a heightened interest in the role played by them in the development process, as well as in
E D

poverty alleviation in the developing economies.

PT

Thus, the dependence on remittances by many households in the low and middle
CE

income economies gives rise to the following question: how do changes in remittances affect

the livelihoods of those in the low and middle income economies? Remittance receipts
AC

however, are affected by changes in the exchange rate. Lin (2010) in the case of Tonga,

highlights that a percentage point of real appreciation in the Tongan currency against the

remitting country currency is related with a two percentage point fall in remittance receipts.

Similarly, Nekoei (2013) finds that migrants earnings fall by 0.92 percent in response to a

10% real appreciation in the US dollar. In a study on the response of Filipino migrants to

exchange rate shocks, Yang (2008) documents that an appreciation of a migrant’s currency

3
 ACCEPTED MANUSCRIPT

against the Philippine peso increases migrant remittances. Olubiyi and Kehinde (2015) find

that an expected depreciation of the real exchange rate lead to a fall in remittance inflows.

Accordingly, exchange rate variations can change remittances receipts, thus, affecting

poverty levels in an economy. The overvaluation of a currency can adversely affect the

macro-economy by discouraging export competitiveness and increasing current account

PT
deficits1. Given the emphasis on the alleviation of poverty by the United Nations and the

Millennium Development Goal (MDG) of eradicating extreme poverty and hunger, the goal

RI
of the present study is to empirically investigate the effect of real exchange rate movements

SC
on poverty, through the specific channel of remittances. Paya et al. (2003) argue that

exchange rate movements may not be symmetric, implying that changes in exchange rates
NU
need not have a symmetric effect on poverty through remittances. Asymmetric exchange rate
MA

adjustments have significant welfare and policy implications (Meyer and Cramon-Taubadel,

2004). If upward movements in the exchange rate lead to declines in remittances and

increases in poverty, then in that sense, asymmetric exchange rate movements are expected to
E D

lead to a different distribution of welfare as well. Similarly, if an exchange rate depreciation

PT

has a greater impact on remittances, leading to a fall in poverty, then this would help to

alleviate poverty. Given the high reliance of many emerging economies on remittances, the
CE

present study additionally seeks to extend upon the literature by investigating whether
AC

changes in real exchange rates affect asymmetrically poverty through remittances.

Overall, the contribution of the present study is threefold: (1) it investigates, to the

best of the authors’ knowledge, for the first time, whether changes in real exchange rates

asymmetrically affect poverty through remittances; (2) it examines the asymmetric effect of

1
Studies also show that large inflows of remittances are associated with real exchange rate appreciation and loss
of export competitiveness (Lopez et al. 2007), and the Dutch disease effect (Acosta et al. 2007).

4
 ACCEPTED MANUSCRIPT

the real exchange on poverty through remittances for the panel of countries under

investigation by disaggregating them by income group – low income and low middle income

and high middle income, as the effects could differ across different income groups; (3) it

makes use of the Nonlinear Auto-Regressive Distributed Lag (NARDL) model proposed by

Shin et al. (2014), which has important advantages over others methodologies that model

jointly cointegration and asymmetries. In particular, this modelling approach provides greater

PT
flexibility in allowing for combinations of I(1) and I(0) variables by making use of a bounds

RI
testing procedure for the presence of the equilibrium vector, while it is not constrained by the

SC
requirement of cointegrating models that all variables are I(1).

Evidence has shown that remittances have enabled low income households to
NU
reduce poverty and income inequality (Bertoli and Marchetta 2014, Naatus 2014, Hatemi-J

and Uddin 2014, Gaaliche and Zayati 2014, Adams and Page, 2003; Taylor et al., 2005;
MA

Acosta et al., 2008), smoothen consumption by reducing vulnerability to adverse shocks

(Yang and Choi, 2007), increase their propensity to save (Adams, 2002), provide capital for
D

micro enterprises (Woodruff and Zenteno, 2007), and invest in farm technology (Stark and
E

Lucas, 1988). Ratha (2007) finds that remittances, by directly supplementing the income of
PT

low and middle income households, influence poverty both directly and indirectly through
CE

multiplier effects. A 10% increase in remittances is found to lead to a 3.5% fall in the share of

poor people at the macroeconomic level, and a fall in the poverty head-count ratio at the
AC

microeconomic level in a number of countries; for example by 11 percentage points in

Uganda, by 6 percentage points in Bangladesh and by 5 percentage points in Ghana (Ratha,

2007).

Notwithstanding the rapidly growing literature on remittances, the emphasis of

the literature hereto (discussed in the next section), has been on the effect of remittance

inflows on poverty (Bertoli and Marchetta 2014, Naatus 2014, Hatemi-J and Uddin 2014,

5
 ACCEPTED MANUSCRIPT

Gaaliche and Zayati 2014, Adams, 2001; Adams and Page, 2003; Taylor et al., 2005; Acosta

et al., 2008; Brown and Jimnez, 2008; Coombes and Ebeke, 2011), as well as on the effects

of remittance receipts on the exchange rate (Caceres and Saca, 2006; Lopez et al., 2007;

Acosta et al., 2007; Barjas et al., 2010). A few studies have examined the impact of exchange

rate changes on remittance receipts (Olubiyi and Kehinde 2015, Yang, 2008; Lin, 2010;

Nekoei, 2013), while another strand of the literature investigate the link between remittances

PT
and the exchange rate, i.e. the Dutch disease effect (Rahman et al. 2013, Acosta et al., 2007).

RI
Surprisingly, there are no empirical studies which investigate the impact of real exchange rate

SC
changes on poverty asymmetrically through the specific channel of remittances.

The rest of this paper is structured as follows. Section 2 discusses the literature.
NU
Section 3 presents the data and methodology. Section 4 discusses the empirical results and
MA

section 5 concludes.
D

2. Background
E
PT

Remittances have, in general, been found to lead to the enhancement of welfare for

remittance receiving households, thus, mitigating both poverty and inequality. Adams (2001)
CE

finds that in rural Egypt, remittance receiving households spend 42.5% of such revenues on
AC

building a new home. In a study of households receiving remittances in Turkey, Koc and

Onan (2004) conclude that of all households receiving remittances, about 80% spend them on

improving their standard of living. Acosta et al. (2008) show that for every percentage point

increase in the remittances to GDP ratio, the poverty headcount is reduced by 0.4% for a

group of ten Latin American and Caribbean countries. Similarly, Taylor et al. (2005) find

evidence in favour of a significant effect of remittances on equalizing income inequality in

Mexico. Their results suggest that in West-Central Mexico which has the largest number of

6
 ACCEPTED MANUSCRIPT

international migrants, a 10% increase in international remittances reduces the total income

Gini index by 0.3%. Evidence also supports the hypothesis that remittances act as an

insurance cover for those experiencing rainfall shocks (Yang and Choi, 2007).

Brown and Jimnez (2008) explore the impact of migration and remittances on poverty

and income distribution in two Pacific Islands, namely, Fiji and Tonga, and find that the

findings depend on the choice of the economic methodology used, as well as on the history of

PT
migration in the countries in question. The impact of remittances on poverty alleviations is

RI
found to be more significant when more rigorous income estimates are used. Esquivel and

SC
Huerta-Pineda (2006) use three different measures of poverty for Mexico, i.e., food-based,

capabilities-based and assets-based2, and note that remittances receiving households are able
NU
to reduce the probability of being in food-based and capabilities-based poverty by 8 and 6

percentage points, respectively; however, they are unable to significantly reduce asset-based
MA

poverty levels.

At the macroeconomic level, Adams and Page (2003) observe that in the case of the
E D

Middle East and North Africa (MENA) countries, both remittances and government
PT

employment have contributed to low poverty rates and reduced income inequality, with a 10

percentage point increase in the share of remittances to GDP reducing the poverty headcount
CE

ratio by 5.7%. In a similar study covering a sample of 71 countries, Adams and Page (2005)
AC

suggest that remittances contribute to reducing the depth, the level and the intensity of

poverty, with a 10% increase in migration leading to a 2.1% fall in the share of people living

below the poverty line of $1.00 a day. Bertoli and Marchetta (2014) examine the effect of

migration on the incidence of poverty in Ecuador. They find that migration leads to a fall in
2
A household was considered as poor if the household’s income was less than 672.25 Pesos in 2002 which is
the lowest income required to buy a basket of food. A household was considered to be in capabilities-based
poverty if that household could not cover his basic expenses on food, health, and education, according to an
officially defined basket; finally, a household was considered to be in assets-based poverty if he could not cover
his expenses on food, health, education, dressing, home and public transportation (Esquivel and Huerta-Pineda,
2006).

7
 ACCEPTED MANUSCRIPT

the incidence of poverty among migrant households. Also in a study between migration

remittances received by 262 municipalities and the level of extreme poverty in El Salvador,

Naatus (2014) similarly finds that remittances lead to a fall in the level of extreme poverty in

municipalities. Hatemi-J and Uddin (2014) in a study of the relationship between remittances

and poverty in Bangladesh over the 1976-2010 period, find a bi-directional relationship

between the two variables. They find that the effect of poverty reduction on remittances is

PT
stronger than the reverse impact. Similar findings are documented by Gaaliche and Zayati

RI
(2014). In a study of the bi-directional causality between remittances and poverty reduction

SC
for 14 emerging and developing countries over 1980-2012, they observe a bi-directional

relationship between poverty and remittances. They also find that the causal effect of poverty
NU
reduction on remittance is stronger than the reverse effect. Employing a panel dataset

covering 1970 - 2000, Portes (2009) observe that an increase in remittances is associated
MA

with a fall in income inequality. While a 1 percent increase in remittances is associated with

an increase in the first decile’s income by approximately 0.43 percent, a 1 percent increase in
D

remittances is associated with a 0.10 percent decrease in the income of the top 10 percent of
E

the population in low income countries.

PT

Coombes and Ebeke (2011) illustrate that remittances reduce consumption instability
CE

and act as an insurance policy in the event of negative shocks, such as natural disasters,
AC

exchange rate and financial instability, in a panel of developing economies. They also argue

that remittances play a stronger role in acting as a stabilizer in the event of negative shocks in

countries with weaker financial systems.

Evidence also shows that the decision to remit is influenced by changes in the

exchange rate. Employing a panel of nine Western Hemisphere countries, Higgins et al.

(2004) test the hypothesis that remittances of immigrants respond to risk variables, in

particular, to exchange rate uncertainty. Their results support the idea that migratory flows

8
 ACCEPTED MANUSCRIPT

are driven by investment motives. The studies by Lianos (1997) and El-Sakka and McNabb

(1999) corroborate the findings of Higgins et al. (2004) in that exchange rates affect a

migrant’s decision to remit. Lianos (1997) examines the flow of migrant remittances from

Germany, Belgium, and Sweden to Greece. He attributes the ongoing devaluation of the

Greek drachma as a factor which causes migrants to postpone sending remittances back

home. Similarly, in a study of Egypt, El-Sakka and McNabb (1999) highlight that both

PT
exchange rates and interest rates differentials play an important role in remittances flowing in

RI
through certain formal channels. Consequently, if the decision of a migrant to remit is based

SC
on changes in the exchange rate, then migrant remittances act as a channel through which the

exchange rate affects poverty.

NU
Another strand in the literature documents that changes in the exchange rate affect
MA

remittance receipts. Lin (2010) finds that a percentage point real appreciation in the Tongan

currency against the remitting country currency is related with a 2 percentage point fall in
D

remittance receipts into Tonga. Nekoei (2013) in a study of Mexican workers in the US,
E

suggests that a 10% appreciation of the US dollar leads to a fall in migrants annual earnings
PT

by 0.92 percent, while Yang (2008) in a study of the response of Filipino migrants to
CE

exchange rate shocks, notes that stronger positive shocks, i.e. an appreciation of a migrant’s

currency against the Philippine peso, leads to larger increases in households’ remittances
AC

receipts. A 10% increase in Philippine Pesos per unit of foreign currency was found to

increase Peso remittances by 6%. Rahman et al. (2013) examine the long-run and short-run

effects between emigrants’ nominal remittances in U.S. dollars into Mexico and the Peso-

Dollar nominal exchange rates using high frequency data. They find that changes in the

exchange rate lead to larger effects on changes in remittances in the long run as compared to

the opposite. Olubiyi and Kehinde (2015) using a choice-theoretical model find that the real

9
 ACCEPTED MANUSCRIPT

exchange rate negatively effects remittances. Hence, an expected depreciation of the real

exchange rate lead to a fall in remittance inflows.

In spite of the fact that a number of studies investigate the impact of remittances on

the exchange rate, i.e. real exchange rate appreciation and loss of export competitiveness

(Lopez et al., 2007), and the Dutch disease effect (Caceres and Saca, 2006; Acosta et al.,

PT
2007; Barajas et al., 2010), much less attention has been paid to the impact of the exchange

rate on remittances. Additionally, although the literature investigates separately the influence

RI
of remittances on poverty, and the influence of exchange rates on remittances, there are no

SC
studies which investigate the effect of exchange rate changes on poverty through the

remittance channel. Given the high reliance of many economies on remittances, the
NU
livelihood of many is severely affected by exchange rate changes. Therefore, the present
MA

study extends upon the literature by investigating the following hypothesis: Changes in

exchange rates affect poverty non-linearly through the remittance channel.

E D
PT

3. Methodology and data

3.1. The methodology of threshold partial adjustment models

CE

This part of the methodology describes a dynamic panel threshold model to allow for

asymmetries in the process of pass-through from changes in the real exchange rate to poverty
AC

through the mechanism of remittances. This particular novel methodology has been

recommended by Dang et al. (2012). This methodological approach is able to consistently

estimate heterogeneous speeds of adjustment in different regimes, as well as to properly test

for the threshold effect. It derives the GMM estimators and describes how the threshold

parameter is estimated. More specifically, it combines time series methodological approaches

in relevance to threshold modelling (Chan, 1993; Hansen, 2000; Caner and Hansen, 2004)

10
 ACCEPTED MANUSCRIPT

with the existing GMM literature (Alvarez and Arellano, 2003). Their methodological

approach has a number of advantages: i) it allows the threshold parameter to be estimated

endogenously within the model, ii) it allows countries to switch regime over-time conditional

on the splitting of real exchange rates into their positive and negative part, iii) it does not

suffer from the generated regressors problem and the resulting estimation and inference

complexities, especially in dynamic panels, and iv) it allows for complex adjustment

PT
mechanisms, whereby countries may not only adjust at heterogeneous rates (short-run

RI
asymmetry), but also adjust toward heterogeneous poverty targets (long-run asymmetries).

SC
The estimation is carried out in two steps. In particular, in the first step of the

empirical analysis, the methodology decomposes the real exchange rate into its positive and
NU
negative partial sums of increases and decreases:

t t
MA

RER+t = ΣΔRER = Σ max(ΔRER , 0)

+
j j
j=1 j=1
D

and
E
PT

t t
Σ
RER-t = ΔRER-j = Σ
min(ΔRERj, 0)
j=1 j=1
CE

Accordingly, the threshold partial adjustment specification can be modelled as follows:

AC

p1 p2
ΔREMit = [ai + Σbj+ΔRER+i(t-j) + Σ dkΔREMi(t-k)] I(qit≤c) +
j=0 k=1

p1 p2
[ai + Σbk-ΔRER-i(t-k) + Σ dkΔREMi(t-k)] I(qit>c)+ vit (1)
j=0 k=1

11
 ACCEPTED MANUSCRIPT

where I{⋅} is an indicator function taking the value of 1 if the event is true (i.e., positive real

exchange rates), and 0 otherwise (i.e., negative real exchange rates), αi captures country-

specific fixed effects characteristics, while vit is the well-behaved error term with a zero mean

and constant variance. Qit is the (regime-switching) transition variable, such as GDP growth.

When guided by altruistic motives, remittances aim to support recipients in their daily

expenditure and/or compensate them for catastrophic events. By contrast, when guided by

PT
investment motives, remittances aim to take advantage of any opportunities for profits in the

RI
home country. In both cases, however, GDP growth is the driver that determines the future

SC
course of remittances (Chami et al., 2005; Browne and Mineshima, 2007). Therefore, in this

respect, the first regime is defined when the growth of GDP is below the threshold, while the
NU
second regime is defined when the growth of GDP is above the threshold. C represents the

threshold parameter, while for simplicity, the transition variable, qit, is assumed to be
MA

stationary and exogenous. REM denotes the remittances/GDP ratio and RER is the real

exchange rate, with v being the error term. The superscripts (+) and (–) stand for the positive
D

and negative partial sums decomposition as defined above, while the symbols p1 and p2
E

denote the respective lag orders for the dependent variable and the remaining variables in the
PT

distributed lag part, respectively. The symmetry is tested by using a Wald test of the null
CE

hypothesis: Σb+ = Σb-.

AC

In the second step, the empirical analysis decomposes the estimated remittances

from Equation (1) into their positive and negative partial sums of increases and decreases:

t t
Σ
RÊM+t = ΔRÊM+j = Σ
max(ΔRÊMj, 0)
j=1 j=1

and

12
 ACCEPTED MANUSCRIPT

t t
Σ
RÊM-t = ΔRÊM-j = Σ
min(ΔRÊMj, 0)
j=1 j=1

and the new threshold panel dynamic specification can be modelled as follows:

q1 q2
ΔPOVit = [βi + Σcj+ΔREM+i(t-j) + ΣfkΔPOVi(t-k)] I(qit≤c)

PT
j=0 k=1

RI
q1 q2
Σ Σ
[βi + ck-ΔREM-i(t-k) + fkΔPOVi(t-k)] I(qit>c)+ wit (2)

SC
j=0 k=1
NU
where w is the new error term and βi captures country fixed effects. In this new case, the

transitional variable q is in relevance to GDP growth again, as this variable has been detected
MA

by the relevant literature to primarily drive the course of poverty levels (Kraay 2006;

Ravallion 2012; among others). Once again, the superscripts (+) and (–) stand for the positive
D

and negative partial sums decomposition of remittances as defined above, while the symbols
E

q1 and q2 q3 denote the respective lag orders for the dependent variable and the remaining
PT

variables in the distributed lag part, respectively. The symmetry is tested again by using a
CE

Wald test of the null hypothesis: Σc+ = Σc-.

AC

3.2. Data

Annual data on the RER are the effective exchange rate; data on poverty

(POV) are the percentage of population below $2.00 a day. Here, a real exchange rate

depreciations will lead to higher remittances reducing poverty, while lower remittances

caused by real exchange rate appreciation will tend to increase poverty. Remittances (REM)

are defined as the addition of personal transfers and compensation of employees (World Bank

13
 ACCEPTED MANUSCRIPT

2014). These include current transfers by migrant workers and wages and salaries earned by

non-resident workers. To the ends of the empirical analysis we use the remittances to GDP

ratio. The study spans the period 1980-2015 period for 99 emerging countries (see the

Appendix for the countries under investigation). Their selection was based on data

availability for the entire time span. The data cover a panel of low and middle income

economies from Sub-Saharan Africa, East Asia and the Pacific, Eastern Europe, Central Asia,

PT
Latin America and the Caribbean, the Middle East and North Africa, and South Asia. The

RI
data for the RER come from Datastream, while those for POV and REM from World Bank’s

SC
World Development Indicators.
NU
4. Empirical analysis
MA

The first step of the empirical analysis examines the presence of cross-section

dependence. In the cases where cross-section dependence is not sufficiently high, a loss of

power might result if second-generation panel unit root tests that allow for cross-section
E D

dependence are employed. Therefore, before selecting the appropriate panel unit root test, it
PT

is crucial to provide some evidence on the degree of residual cross-section dependence. The

cross-sectional dependence (CD) statistic by Pesaran (2004) is based on a simple average of

CE

all pair-wise correlation coefficients of the OLS residuals obtained from standard augmented
AC

Dickey-Fuller regressions for each variable in the panel. Under the null hypothesis of cross-

sectional independence, the CD test statistic follows asymptotically a two-tailed standard

normal distribution. The results reported in Table 1 uniformly reject the null hypothesis of

cross-section independence, providing evidence of cross-sectional dependence in the data

given the statistical significance of the CD statistics regardless of the number of lags (from 1

to 4) included in the ADF regressions.

[Insert Table 1 about here]

14
 ACCEPTED MANUSCRIPT

Next, two second-generation panel unit root tests are employed to determine the degree of

integration in the respective variables. The Pesaran (2007) panel unit root test does not

require the estimation of factor loading to eliminate cross-sectional dependence. Specifically,

the usual ADF regression is augmented to include the lagged cross-sectional mean and its

first difference to capture the cross-sectional dependence that arises through a single-factor

model. The null hypothesis is a unit root for the Pesaran (2007) test. The bootstrap panel unit

PT
root tests by Smith et al. (2004) utilize a sieve sampling scheme to account for both the time

RI
series and cross-sectional dependence in the data through bootstrap blocks. All four tests by

SC
Smith et al. (2004) are constructed with a unit root under the null hypothesis and

heterogeneous autoregressive roots under the alternative hypothesis. The results of these
NU
panel unit root tests are reported in Table 2 and support of the presence of a unit root in both

variables under consideration.

MA

[Insert Table 2 about here]

In the next step, the empirical analysis involves the GMM estimation of the asymmetric pass-
D

through of the RER to REM perform the Wald test for detecting both the short- and long-run
E

symmetry, while the optimal number of lags is selected on the basis of the SIC information
PT

criterion. Table 3 reports the results. Countries are classified into the first (real appreciations)
CE

regime, when the value of the transition variable is less than or equal to the estimated

threshold value, and into the second regime (real depreciations), when the value of the
AC

transition variable is greater than the estimated threshold value. Panel A reports the short-run

dynamics, as well as the Wald test statistic for the null hypothesis of short-run symmetry. It

also reports the implied speed of adjustment and the threshold value. Furthermore, the results

illustrate the conventional AR(2) and Sargan test statistics, which allow us to check out for

the validity of the instruments used in the GMM regressions. It is worth noting that both tests

are not rejected at the 1% significance level, suggesting that all GMM regressions use valid

15
 ACCEPTED MANUSCRIPT

instruments. Panel B reports the long-run coefficients, along with the Wald test statistics

under the null of long-run symmetry. Note that these tests enable us to shed light on the

interesting question of whether countries in different (growth) regimes adjust toward

heterogeneous target remittances. Finally, the findings also display the GDP growth

characteristics of countries classified into the first and second regime. In terms of symmetry

tests, the results clearly illustrate that the null hypothesis of both long- and short-run

PT
symmetry is clearly rejected at the 1% level.

RI
[Insert Table 3 about here]

SC
The next step in the empirical analysis estimates the symmetric and asymmetric pass-through

of the estimated remittances (i.e., REM) from Equation (1) to poverty (i.e., POV) through the
NU
employment of the model Equation (2); Table 4 reports the new GMM findings. Once again

the optimal number of lags is selected on the basis of the SIC information criterion. Countries
MA

are classified into the first (remittances coming from real appreciations) regime and into the

second regime (remittances coming from real depreciations). Panel A reports the short-run
D

dynamics, as well as the Wald test statistic for the null hypothesis of short-run symmetry.
E

Both AR(2) and Sargan tests indicate that the null hypothsis is not rejected at the 1%
PT

significance level, suggesting that all GMM regressions use valid instruments. Panel B
CE

reports the long-run coefficients, along with the Wald test statistics under the null of long-run

symmetry. Note that these tests enable us to shed light on the interesting question of whether
AC

countries in different (growth) regimes adjust toward heterogeneous target poverty levels. In

terms of symmetry tests, the results clearly illustrate that the null hypothesis of both long- and

short-run symmetry is clearly rejected at the 1% level. Furthermore, the estimates indicate

that both in the short- and in the long-run the impact of remittances on poverty is stronger in

the case where the estimated remittances come from a real depreciation regime.

[Insert Table 4 about here]

16
 ACCEPTED MANUSCRIPT

This final part of the empirical analysis explores the validity of the above findings in three

separate country panels based on the World Bank income classification: low income, lower

middle income, and upper middle income countries (given that the majority of remittances

are in relevance to these country groups) to infer the (asymmetric) impact of remittances on

poverty through the real exchange rate factor. First, Table 5 reports the estimation of the

asymmetric pass-through of the RER to REM. The income classification findings clearly

PT
provide evidence in favour that the null hypothesis of both short- and long-run symmetry is

RI
clearly rejected at the 1% level across all three income country panels. The strongest results

SC
are associated with the low income countries panel, denoting that real exchange rate changes

have a strongest impact on remittances heading to countries within this sample, which could
NU
be potentially attributed to the fact that low income levels in countries within this sample are

vital for the sustainability of life in countries where remittances are used to enhance the
MA

people’s welfare significantly. According to Coombes and Ebeke (2011), remittances reduce

consumption instability and act as an insurance policy in the event of negative shocks, such
D

as natural disasters, exchange rate and financial instability, in a panel of developing

E

economies. In low income countries, evidence points to remittance flows being associated
PT

more with the altruistic motive, I an investment motive which potentially could explain this.
CE

[Insert Table 5 about here]

Finally, Table 6 illustrates the asymmetric effect of estimated remittances on poverty across
AC

the three income groups. Once again, the results provide supportive evidence to those reached

earlier. More specifically, the presence of asymmetry is confirmed, with the results remaining

robust across the three income groups, while the impact of remittances on poverty is stronger

in the case of low income countries, followed by the lower middle and upper middle

countries, when these estimated remittances are in relevance to countries coming from real

depreciated currencies. Additionally, a policy predicament faced by countries is dealing with

17
 ACCEPTED MANUSCRIPT

a large influx of remittance inflows to curb inflationary pressure. A number of studies show

that certain countries devalue their currency in response to these inflows. For example, the

nominal devaluations in Costa Rica, Guatemala, Honduras and Nicaragua were 11.9%, 5.3%,

10.1% and 11.1%, respectively, over the period 1995-2000 (Inter-American Development

Bank, 2007). Evidence also documents that in some countries, Central Bank’s intervention

policies in the exchange rate market, have been asymmetrical, limiting the nominal

PT
appreciation; for example, Tajikistan (IMF, 2006), the world’s largest remittance receiving

RI
country in terms of the remittances to GDP ratio. This potentially explains why exchange rate

SC
depreciations tend to induce a larger effect on poverty through remittances compared to

exchange rate appreciations.

NU
[Insert Table 6 about here]
MA

5. Conclusion

This study assessed the asymmetric effects of real exchange rate changes on poverty for a
D

sample of 99 countries through the channel of remittances. The findings highlighted that real
E

depreciations exerted a stronger (negative) effect on remittances than real appreciations. The
PT

magnitude and significance of these flows has led to heightened interest in the role played by
CE

remittances in development and poverty alleviation the in developing economies.

Remittances to the developing countries have helped to increase foreign exchange earnings,
AC

foreign reserves, and service debt. They have also contributed to reduce current account

deficits in many countries – see for example, Kireyev (2006) in the context of Tajikistan. To

the extent that changes in the exchange rate is asymmetrical, with depreciations fostering a

stronger impact on remittances compared to appreciations, such exchange rate depreciations

would help to increase foreign exchange reserves and improve a country’s current account

balance. Similarly, exchange rate depreciations would promote higher levels of investments

and economic growth. In particular, if depreciations have a stronger impact on remittances, as

18
 ACCEPTED MANUSCRIPT

many migrants invest their remittance income in small scale businesses, real estate and other

assets, exchange rate depreciations are expected to reduce credit constraints and alleviate

poverty. Thus, policymakers in emerging economies should think carefully of the

consequences of exchange rate changes as they can affect remittances receipts and the

standard of living for households.

PT
References

RI
SC
Acosta, P., Calderon, C., Fajnzylber, P., Lopez, H., 2008. What is the impact of international

remittances on poverty and inequality in Latin America? World Development 36, 89-
NU
114.
MA

Acosta, P., Lartey, E., Mandelman, F., 2007. Remittances and the Dutch disease. Federal

Reserve Bank of Atlanta Working Paper, No 2007-8a.

D

Adams, R., 2001. The economic impact and uses of international remittances in rural Egypt.
E

Economic Development and Cultural Change 39, 698-722.

PT

Adams, R., 2002. Precautionary saving from different sources of income: evidence from rural
CE

Pakistan. World Bank Policy Research Working Paper, No. 2761, Washington, DC.

Adams, R., Page, J., 2005. Do international migration and remittances reduce poverty in
AC

developing countries? World Development 33, 1645-1669.

Adams, R., Page, J., 2003. Poverty, inequality and growth in selected Middle East and North

Africa countries, 1980-2000. World Development 31, 2027-2048.

Alvarez, J., Arellano, M., 2003. The time series and cross-section asymptotics of dynamic

panel data estimators. Econometrica 71, 1121-1159.

19
 ACCEPTED MANUSCRIPT

Barajas, A., Chami, R., Fullenkamp, C., Garg, A., 2010. The global financial crisis and

workers’ remittances to Africa: what’s the damage? IMF Working Paper, No. 10/24,

Middle East and Central Asia Department.

Bertoli S and Marchetta F (2014) Migration, Remittances and Poverty in Ecuador, Journal of

Development Studies, 50, 1067-1089.

PT
Brown, R., Jimenez, E., 2008. Estimating the net effects of migration and remittances on

poverty and inequality: comparison of Fiji and Tonga. Journal of International

RI
Development 20, 547-571.

SC
Browne, C., Mineshima, A., 2007. Remittances in the Pacific region. IMF Working Paper,
NU
WP/07/35, Washington: International Monetary Fund.
MA

Caceres, L., Saca, N., 2006. What do remittances do? Analysing the private remittance

transmission mechanism in El Salvador. IMF Working Paper.

D

Caner, M., Hansen, B.E., 2004. Instrumental variable estimation of a threshold model.
E

Econometric Theory 20, 813-843.

PT

Chami, R., Barajas, A., Cosimano, T., Fullenkamp, C., Gapen, M., Montiel, P., 2008.
CE

Macroeconomic consequences of remittances. IMF Occasional Paper, No. 259,

Washington: International Monetary Fund.

AC

Chan, K.S., 1993. Consistency and limiting distribution of the least squares estimator of a

threshold autoregressive model. Journal of Annual Statistics 21, 520-533.

Coombes, J., Ebeke, C., 2011. Remittances and household consumption instability in

developing countries. World Development 39, 1076-1089.

20
 ACCEPTED MANUSCRIPT

Dang, V.A., Kim, M., Shin, Y., 2012. Asymmetric capital structure adjustments: new

evidence from dynamic panel threshold models. Journal of Empirical Finance 19, 465-

482.

De Haas, H., 2005. International migration, remittances and development: myths and facts.

Third World Quarterly 26, 1269-1284.

PT
El-Sakka, M., McNabb, R., 1999. The macroeconomic determinants of emigrant remittances.

World Development 27, 1493-1502.

RI
SC
Gaaliche M and Zayati M. 2014. The causal relationship between remittances and poverty

reduction in developing country: using a nonstationary dynamic panel data, Atlantic

NU
Review of Economics, 1, 1-12.
MA

Hansen, B.E., 2000. Sample splitting and threshold estimation. Econometrica 68, 575-603.

Hatemi-J A and Uddin G S (2013) On the causal nexus of remittances and poverty reduction,
D

Applied Economics, 46, 374-382.

E
PT

IMF, 2006. Republic of Tajikistan: sixth review under the poverty reduction and growth

facility. IMF Staff Country Report, No. 06/62.

CE

Inter-American Development Bank, 2007. Central American report: number 3: 2004-2006.

Subregional Integration Report Series, Argentina, Buenos Aires.

AC

Kapur, D., 2005. Remittances: the new development mantra? In Maimbo, S., Ratha, D.,

(Eds.) Remittance Development Impact and Future Prospects, World Bank,

Washington.

Kireyev, A., 2006. The acroeconomics of remittances: the case of Tajikistan. IMF Working

Paper, No. 06/2.

21
 ACCEPTED MANUSCRIPT

Koc, I., Onan, I., 2004. International migrants’ remittances and welfare status of the left

behind families in Turkey. International Migration Review 38, 78-112.

Kraay, A., 2006. When is growth pro-poor? Evidence from a panel of countries. Journal of

Development Economics 80, 198-227.

Lianos, T., 1997. Factors determining migrant remittances: the case of Greece. International

PT
Migration Review 31, 72-87.

RI
Lin, H., 2010. Determinants of remittances: evidence from Tonga. IMF Working Paper, No.1,

SC
Asia and Pacific Department.

Lopez, H., Molina, L., Bussolo, M., 2007. Remittances and the real exchange rate. World
NU
Bank Policy Research Working Paper, No. 4213, Washington.
MA

Meyer, J., Von Cramon-Taubadel, S., 2004. Asymmetric price transmission: a survey.

Journal of Agricultural Economics 55, 581-611.

D

Naatus M K., 2014. Remittances and poverty: A complex relationship, evidence from El-
E

Salvador, Advances in Management and Applied Economics, 4, 1-8.

PT

Nekoei, A., 2013. Immigrant’s labour supply and exchange rate volatility. American
CE

Economic Journal: Applied Economics 5, 144-164.

Olubiyi E and Kehinde K (2015) Does exchange rate affect remittances in Nigeria?, Review
AC

of Finance and Banking, 7, 31-45.

Paya, I., Venetis, I., Peel, D., 2003. Further evidence on PPP adjustment speeds: the case of

effective real exchange rates and the EMS. Oxford Bulletin of Economics and Statistics

65, 421-437.

Pesaran, M.H., 2007. A simple panel unit root test in the presence of cross-section

dependence. Journal of Applied Econometrics 22, 265-312.

22
 ACCEPTED MANUSCRIPT

Pesaran, M.H., 2004. General diagnostic tests for cross section dependence in panels.

Cambridge Working Papers in Economics, No. 435 and CESifo Working Paper, No.

1229.

Portes L S V (2009) Remittances, poverty and inequality, Journal of Economic Development,

34, 127-140.

Rahman M, Foshee A and Mustafa M (2013) Remittances exchange rate nexus: The US-

PT
Mexico case, Journal of Developing Areas, 47, 63-74.

RI
Ratha, D., 2007. Leveraging remittances for development. Policy Brief, World Bank,

Washinghton, DC.

SC
Ravallion, M., 2012. Why don’t we see poverty convergence? American Economic Review
NU
102, 504-523.

Smith. V., Leybourne, S., Kim, T.H., 2004. More powerful panel unit root tests with an
MA

application to the mean reversion in real exchange rates. Journal of Applied

Econometrics 19, 147-170.

D

Stark, O., Lucas, R., 1988. Migration, remittances and the family. Economic Development
E
PT

and Cultural Change 36, 465-481.

CE

Swamy, G., 1981. International migrant workers’ remittances: issues and prospects. World

Bank Working Paper, No. 481, Washington.

AC

Taylor, J.E., Mora, J., Adams, R., Lopez-Feldman, A., 2005. Remittances, inequality and

poverty: evidence from rural Mexico. Working Paper, No. 05-003, Department of

Agricultural and Resource Economics, University of California, Davis.

Woodruff, C., Zenteno, R., 2007. Migration networks and microenterprises in Mexico.

Journal of Development Economics 82, 509-528.

23
 ACCEPTED MANUSCRIPT

World Bank, 2016. Migration and remittance database.

http://go.worldbank.org/092X1CHHD0.

World Bank, 2014. World development indicators. World Bank, Washington.

Yang, D., 2008. International migration, remittances and household investment: evidence

from Philippine migrants and exchange rate shocks. Economic Journal 118, 591-630.

PT
Yang, D., Choi, H., 2007. Are remittances insurance? Evidence from rainfall shocks in the

RI
Philippines. World Bank Economic Review 21, 219-248.

SC
NU
MA
E D
PT
CE
AC

24
 ACCEPTED MANUSCRIPT

Table 1
Cross-section dependence (CD) tests.
Lags

Variables 1 2 3 4

LREM [0.00]a [0.00]a [0.00]a [0.00]a

RER [0.00]a [0.00]a [0.01]a [0.00]a

POV [0.00]a [0.00]a [0.00]a [0.01]a

PT
Notes: Under the null hypothesis of cross-sectional independence the CD statistic is distributed as a two-tailed
standard normal. Results are based on the test of Pesaran (2004). Figures in parentheses denote p-values.

RI
Significance levels: a(1%).

SC
NU
MA
E D
PT
CE
AC

25
 ACCEPTED MANUSCRIPT

Table 2

Panel unit root tests.

Pesaran Pesaran Smith et al. Smith et al. Smith et al. Smith et al.
Variable
CIPS CIPS* t-test LM-test max-test min-test

REM -1.17 -1.26 -1.24 3.03 -1.21 1.33

ΔREM -5.83a -6.02a -5.61a 22.08a -6.99a 7.30a

PT
RER -1.25 -1.25 -1.34 3.11 -1.32 1.32

ΔRER -5.89a -6.04a -7.03a 21.82a -7.93a 8.10a

RI
POV -1.24 -1.30 -1.26 2.83 -1.24 1.25

ΔPOV -5.53a -5.90a -6.26a 21.19a -6.80a 6.97a

SC
Notes: Δ denotes first differences. A constant is included in the Pesaran (2007) tests. Rejection of the null
NU
hypothesis indicates stationarity in at least one country. CIPS* = truncated CIPS test. Critical values for the
Pesaran (2007) test are -2.40 at 1%, -2.22 at 5%, and -2.14 at 10%, respectively. “a” denotes rejection of the
null hypothesis. Both a constant and a time trend are included in the Smith et al. (2004) tests. Rejection of the
MA

null hypothesis indicates stationarity in at least one country. For both tests the results are reported at lag = 4.
The null hypothesis is that of a unit root.
E D
PT
CE
AC

26
 ACCEPTED MANUSCRIPT

Table 3

Short- and long-run GMM estimates and symmetry tests (The effect of the real exchange rate on

remittances).

__________________________________________________________________________

Variables Coefficient(1st regime) p-value Coefficient(2nd regime) p-value

___________________________________________________________________________

PT
Panel A. Short-run

RI
REM(-1) 0.697 [0.00] 0.728 [0.00]

RER+(-1) 0.144 [0.00]

SC
RER-(-1) NU -0.164 [0.00]

Speed of adjustment 0.451 0.542

Threshold 0.075
MA

(Wald) symmetry test 43.61 [0.00]

Sargan 30.44 [0.65]

D

AR(2) -1.295 [0.33]

E

Panel B. Long-run
PT

RER+ 0.493 [0.00]

CE

RER- -0.766 [0.00]

(Wald) symmetry test 50.32 [0.00]

AC

Transition variable

GDP growth -0.04 0.05

__________________________________________________________________________

Notes: AR(2) test is a test for the second-order serial correlation, and is asymptotically distributed as N(0,1)
under the null of no serial correlation. Sargan test is a test for the validity of instruments and is asymptotically
distributed as χ2 under the null of valid instruments. As instruments lagged values up to 4 lags of the
independent variables have been used. The symmetry tests the null hypothesis that Σb+ = Σb- (either in the
short- or in the long-run).

27
 ACCEPTED MANUSCRIPT

Table 4

Short- and long-run symmetry tests (the effect of remittances on poverty).

__________________________________________________________________________

Variables Coefficient(1st regime) p-value Coefficient(2nd regime) p-value

___________________________________________________________________________

Panel A. Short-run

PT
POV(-1) 0.410 [0.00] 0.509 [0.00]

RI
RÊM+(-1) 0.296 [0.00]

RÊM+(-2) 0.112 [0.00]

SC
RÊM-(-1) -0.389 [0.00]
NU
RÊM-(-2) -0.192 [0.00]

Speed of adjustment 0.378 0.524

MA

Threshold 0.070

(Wald) symmetry test 40.85 [0.00]

D

Sargan 26.72 [0.74]

E

AR(2) -1.075 [0.41]

PT

Panel B. Long-run
CE

RÊM+ 0.428 [0.00]

RÊM- -0.796 [0.00]

AC

(Wald) symmetry test 64.18 [0.00]

Transition variable

GDP growth -0.033 0.042

__________________________________________________________________________

Notes: As instruments, lagged values up to 3 lags of the independent variables have been used. The symmetry
tests the null hypothesis that Σc+ = Σc- (either in the short- or in the long-run). The remaining are similar to
those in Table 3.

28
 ACCEPTED MANUSCRIPT

Table 5

Short- and long-run symmetry tests (The effect of the real exchange rate on remittances)-Income country panels.

______________________________________________________________________________________________________________________________________

Low income countries Lower middle income countries

P T Upper middle income countries

Variables Coefficient(1st regime) Coefficient(2nd regime) Coefficient(1st regime) Coefficient(2nd regime)

R I Coefficient(1st regime) Coefficient(2nd regime)

C
______________________________________________________________________________________________________________________________________

Panel A. Short-run

REM(-1) 0.718[0.00] 0.775[0.00] 0.630[0.00]

U S 0.675[0.00] 0.498[0.00] 0.531[0.00]

RER+(-1) 0.282[0.00]
A
0.237[0.00] N 0.172[0.00]

RER-(-1) -0.3295[0.00]
M -0.285[0.00] -0.204[0.00]

Speed of adjustment 0.580 0.644

E D 0.469 0.497 0.401 0.445

Threshold

(Wald) symmetry test

0.089

75.52[0.00]
P T 0.064

59.68[0.00]
0.051

46.37[0.00]

Sargan 32.25[0.59]

C E 28.75[0.70] 30.81[0.63]

AR(2)

Panel B. Long-run
-1.214[0.42]

A C -1.163[0.50] -1.151[0.52]

RER+ 0.575[0.00] 0.499[0.00] 0.432[0.00]

RER- -0.791[0.00] -0.663[0.00] -0.579[0.00]

(Wald) symmetry test 66.19[0.00] 56.25[0.00] 37.81[0.00]

29
 ACCEPTED MANUSCRIPT

Transition variable

GDP growth -0.05 0.04 -0.03 0.026 -0.01 0.022

____________________________________________________________________________________________________________________________________

T
Notes: Similar to those in Table 3.

I P
C R
U S
A N
M
E D
P T
CE
AC

30
 ACCEPTED MANUSCRIPT

Table 6

Short- and long-run symmetry tests (The effect of estimated remittances on poverty)-Income country panels.

_____________________________________________________________________________________________________________________________ _________

Low income countries Lower middle income countries

P T Upper middle income countries

Variables Coefficient(1st regime) Coefficient(2nd regime) Coefficient(1st regime) Coefficient(2nd regime)

R I Coefficient(1st regime) Coefficient(2nd regime)

C
______________________________________________________________________________________________________________________________________

Panel A. Short-run

POV(-1) 0.633[0.00] 0.679[0.00] 0.580[0.00]

U S
0.597[0.00] 0.486[0.00] 0.498[0.00]

RÊM+(-1) 0.360[0.00] 0.278[0.00]

A N 0.196[0.00]

RÊM-(-1) -0.395[0.00]
M -0.304[0.00] -0.211[0.00]

Speed of adjustment 0.520 0.574

D
0.450

E
0.503 0.368 0.391

Threshold

(Wald) symmetry test

0.070

69.42[0.00]
P T 0.054

53.30[0.00]
0.044

40.21[0.00]

Sargan 27.59[0.67]

C E 25.02[0.76] 32.38[0.54]

AR(2)

Panel B. Long-run
-1.105[0.35]

A C -1.140[0.52] -1.149[0.51]

RÊM+ 0.532[0.00] 0.458[0.00] 0.421[0.00]

RÊM- -0.644[0.00] -0.496[0.00] -0.457[0.00]

(Wald) symmetry test 60.25[0.00] 49.42[0.00] 35.64[0.00]

31
 ACCEPTED MANUSCRIPT

Transition variable

GDP growth -0.049 0.035 -0.03 0.029 -0.01 0.020

_____________________________________________________________________________________________________________________________ _______

T
Notes: Similar to those in Table 5.

I P
C R
U S
A N
M
E D
P T
C E
AC

32
 ACCEPTED MANUSCRIPT

Appendix

Country samples

__________________________________________________________________________

Full sample: 99 countries

Low Income Countries: 24 countries

PT
Afghanistan, Burkina Faso, Burundi, Cambodia, Central African Republic, Chad, Ethiopia,

RI
Gambia, Guinea, Haiti, Liberia, Madagascar, Malawi, Mali, Mozambique, Nepal, Niger,

SC
Rwanda, Sierra Leone, Somalia, Tanzania, Togo, Uganda, Zimbabwe.

Lower Middle Income Countries: 31 countries

NU
Armenia, Bangladesh, Bolivia, Cameroon, Cote d’Ivoire, Djibouti, Egypt, El Salvador,

Georgia, Ghana, Guatemala, Honduras, India, Indonesia, Kenya, Laos, Mauritania, Moldova,
MA

Morocco, Nicaragua, Nigeria, Pakistan, Philippines, Senegal, Sri Lanka, Sudan, Tajikistan,

Ukraine, Uzbekistan, Vietnam, Zambia.

D

Upper Middle Income Countries: 44 countries

E

Albania, Algeria, Angola, Azerbaijan, Belarus, Bosnia-Herzegovina, Botswana, Brazil,

PT

Bulgaria, China, Colombia, Costa Rica, Cuba, Dominican Republic, Ecuador, Fiji, Gabon,
CE

Grenada, Iran, Jamaica, Jordan, Kazakhstan, Lebanon, Libya, Malaysia, Maldives, Marshall

Islands, Mauritius, Mexico, Mongolia, Namibia, Panama, Paraguay, Peru, Romania, South
AC

Africa, St. Lucia, St. Vincent & Grenadines, Suriname, Thailand, Tonga, Tunisia, Turkey,

Turkmenistan.

33
 ACCEPTED MANUSCRIPT

Highlights

Asymmetric real exchange rates and poverty: The role of remittances

● The paper explores the asymmetric effect of real exchange rates on poverty

● The effect occurs through the remittance channel

● It employs a panel of 99 countries, spanning the period 1980-2014

PT
●The methodology considers a threshold partial adjustment model

RI
● Real exchange rate depreciations exert a stronger effect on poverty than appreciations

SC
NU
MA
E D
PT
CE
AC

34