Impact of Development Aid On Remittances Sent From Donor-Countries

October 2018
Volume: 3, No: 2, pp. 109 – 134

ISSN: 2059-6588
e-ISSN: 2059-6596
www.tplondon.com/rem
Article history: Received 8 June 2018; accepted 12 October 2018
Impact of development aid

on remittances sent from Sèna Kimm
donor-countries Gnangnon 
Abstract
This article examines the impact of the aggregate development aid supplied
by donors at the aggregate level of remittances sent from donor-countries.
The empirical analysis focuses on 22 donor-countries over the period 2000-
2015. It uses instrumental variables approaches and provide evidence that
donors' overall aid supply could be substitute or complementary with
remittances paid flows, depending on the level of remittances sent from
donor-countries, as well as on the share of migrants' stock on the total
population of the host-country.
Keywords: Remittances paid; Development aid; Donor-countries.
JEL Classification: F24, F35
Introduction
In light of the importance of migrants' remittances for beneficiaries in
home countries of migrants, many studies have been conducted on
migrants' remittances, in terms of both their determinants and their
impact on the receiving economies. Specifically, the studies on the
determinants of remittances have put emphasis on the
macroeconomic and microeconomics factors that influence
migrants' remittances received. These studies have either used
aggregate data to analyse the determinants of remittances by
relying on a framework of country/year or they have used a bilateral
setting, i.e., country-pair1/year to conduct the analysis. Among the
identified macroeconomic factors that explain the amount of
migrants' remittances received are the number of workers, the wage
rates, the economic situation in the host country of the migrants, the
economic situation in the migrants' country of origin, the exchange
rate in the home country of migrants, the relative interest rate
between the sending and the recipient countries, the institutional

Sèna Kimm Gnangnon, Economic Affairs Officer, World Trade Organization, Rue de
Lausanne 154, CH-1211 Geneva 21, Switzerland. E-mail: kgnangnon@yahoo.fr.
1 Country-pair refers to country of origin and country of destination of the migrants.
Copyright @ 2018 REMITTANCES REVIEW © Transnational Press London

110 Impact of development aid on remittances sent from donor-countries
quality as well as some demographic and financial/political

characteristics in the recipient-country (e.g., El-Sakka & McNabb,
1999; Freund and Spatafora, 2005; Gupta, 2005; Schiopu and Siegried,
2006; Chami et al, 2008; Adams, 2009). However, the bulk of these
studies have focused on the amount of remittances received in the
migrants' countries of origin, and insufficient attention has been paid
to the determinants (at the aggregate level) of the amount of
remittances sent from the countries of destination of migrants, in
particular developed countries.
At the same time, developed countries are known as the major
providers of development aid flows (also known as official
development aid, i.e., ODA) to developing countries, including
poorest ones, with a view to promoting development in these
countries. In recent years, very few studies, notably Bertoli et al. (2008)
and Fuchs et al. (2014) have, in their investigation of the determinants
of donors' ODA supply, explored whether the aggregate amount of
remittances paid (i.e., the total amount of remittances sent from these
donor-countries, i.e., developed countries) substitute or complement
ODA. Both have obtained that remittances do not exert a significant
impact on donors' overall aid supply: Bertoli et al. (2008) have proxied
remittances by the share of immigrant population on the country's
total population, and Fuchs et al. (2014) have used workers'
remittances and compensation of non-resident employees paid (i.e.,
sent from donor-countries) – henceforth referred to as remittances
paid - in their respective analysis. Nonetheless, these authors have
recognized the possibility of reverse causality from ODA to
remittances, i.e., ODA could in turn influence remittances paid.
The current article aims to complement the above-mentioned studies
that have focused on remittances received by looking now at the
side of countries from which remittances are sent (development aid
donor-countries), and examine whether such a reverse causality from
ODA to remittances exists. In other words, the article investigates, at
the aggregate level, the impact of donors' aid on remittances paid,
i.e., whether donors' development aid complements or substitutes to
remittances sent from donor-countries. To the best of our knowledge,
this is the first study on this matter, as other studies have looked at the
subject from the perspective of aid and remittances received by
recipient-countries. The article argues that the impact of
development aid on remittances paid would take place through its
impact on migration. The analysis is conducted on a sample of 21
donor-countries over the period 2000-2015, based on data
availability. Figures 1 and 2 provide a first glimpse on the relationship
between development aid flows i.e., the Net Aid Transfers, in
Gnangnon 111
percentage of donor-countries' Gross Domestic Product (GDP)

(denoted "NAT") and remittances paid, in percentage of donor-
countries' GDP ("REMIT"). Data used represent the average on the 21
donor-countries over the period 2000-2015. Specifically, Figure 1
presents the evolution over 2000-2015 of both types of capital flows,
while Figure 2 displays the cross-plot between these two types of
capital flows. Figure 1 shows that both Net Aid Transfers and
remittances paid have evolved more or less in the same direction over
the period. Net Aid transfers moved from 0.33% in 2000 to 0.46% in
2015, whereas remittances paid (% GDP) evolved from 0.46% in 2000
to 0.69% in 2015. Hence, this Figure clearly shows that the amount of
remittances sent from donor-countries has always exceeded the
amount of aid supplied by donors. Turning to Figure 2, the left-hand
side graph shows the correlation pattern between NAT and REMIT,
where the natural logarithm has not been applied to these variables.
The right-hand side graph presents the correlation pattern between
NAT and REMIT, where the natural logarithm has been applied to the
two variables. The application of the natural Logarithm to these two
variables in the right-hand side graph aims to show that it helps
address the outliers observed in the left-hand side graph. Both graphs
show a positive correlation pattern between NAT and REMIT.
However, the right-hand side graph seems to indicate that they may
be some outliers, which include countries where remittances paid (%
GDP) are higher than 1% [(= exponential (0)], and where remittances
paid (% GDP) are lower than 0.135% [(= exponential (-2)]. The
empirical exercise will duly take into account the possible effects of
these outliers.
The empirical analysis uses the instrumental variables approach and
shows evidence that the effect of net aid transfers on remittances
paid depends on the level of remittances-to-GDP ratio sent from
donor-countries to migrants' countries of origin, as well as on the share
of migrants' stock in the host-country's total population.
The rest of the paper is structured as follows. Section 2 discusses from
a theoretical perspective how donors' aid could influence
remittances paid. Section 3 presents the model that would help
examine empirically the impact of ODA on remittances paid, and
discusses the empirical strategy. Section 4 interprets the empirical
results. Section 5 undertakes a robustness check analysis. Section 6
concludes.
Discussion on the theoretical impact of donors' aid on
remittances paid
Recent years have witnessed a great attention to the relationship
between migration and aid. For example, further to the recent surge
of migrants flows from developing countries towards donor-countries,

the political statements 2 made by development aid donors,
including the European Union or European individual countries, tend
to the reflect the fact that development aid would substitute to
migration. As we will see below, the literature on the relationship
between development aid and migration is still inconclusive as to the
sign of this impact. In the current article, we argue that the
development aid provided by donors would influence remittances
sent from donor-countries (remittances paid) through its impact on
migration flows. In particular, we expect that development aid could
induce higher migration flows, and hence higher remittances paid, all
other things being equal. Similarly, development aid could induce
lower migration flows and consequently lower remittances paid, all
other things being equal.
Concerning the related-literature, there is from a theoretical
perspective no direct link between development aid and migration,
as development aid is expected to influence migration through its
indirect effects on the determinants of migration. Additionally, the
influence of development aid flows on migration flows is subject to
opposite forces so that its net impact is a priori unknown (Parsons and
Winters, 2014; Berthélemy et al., 2009). The hump-shaped pattern
hypothesis between receiving countries' development level (proxied
by their real per capita income) and migration flows has been put
forth in the literature to explain how development aid could affect
migration flows (see for example, Schiff, 1994; Hatton and Williamson,
2002; Adam and Page, 2003; Faini and Venturini, 1993; Berthélemy et
al., 2009; and Clemens, 2014 for this hump-shaped hypothesis). This
hypothesis combines two major, but opposite channels through which
development aid could influence migration, including the income
channel and the budgetary constraint channel. According to hump-
shaped hypothesis, for relatively low levels of Gross Domestic Product
(GDP) per capita, migration is positively associated with GDP per
capita. For relatively higher levels of GDP per capita, migration is
negatively correlated with GDP per capita. In other words, when initial
income is low, a rise in this income induces higher migration flows,
while the opposite is true for higher levels of income. This income
channel suggests that by raising households' disposable income (but
not necessarily the economy-wide growth rates), development aid
2 See for example the declaration of the UK Secretary of Defense in The Guardian, 21st
June 2015. See also the recent strengthening of the cooperation between the
European Union (EU) and the International Organization for Migration (IOM) on global
migration (https://ec.europa.eu/home-affairs/news/eu-and-iom-strengthen-ties-
global-migration_en)
Gnangnon 113
would discourage people from emigrating and hence, reduce

emigration. On the other hand, the budgetary constraint channel
posits that development aid could influence the impact of additional
wealth on financing migration costs for a larger share of the
population in the countries of origin. In other words, loosening budget
constraints makes migration more feasible. The combination of these
two channels could lead to the situation where at very low levels of
income per capita, the budgetary constraint channel dominates the
income channel, as development aid could generate higher
migration flows by allowing poor migrants to afford the migration
costs. However, as the income per capita rises, the income channel
progressively dominates the budgetary constraint channel.
More recently, Berthélemy et al. (2009) have demonstrated that
bilateral aid could influence migration flows through the 'attraction
effect'. The authors postulate that higher bilateral aid to a recipient-
country intensifies the donor-country's attractiveness for citizens of this
recipient-country. This is because aid policy implementation
facilitates bilateral contacts and increases the information
concerning the donor-country (including on labor market conditions)
available to potential migrants of the recipient-country. In turn, this
additional information would reduce the transaction costs attached
to the considered migration flows, and hence induce higher
migration flows.
On the empirical front, the results have shown mixed evidence. Faini
and Venturini (1993) have tested empirically the hypothesis that
income growth may fail to reduce emigration because it relaxes
credit-constraints, which tend to be especially binding in poorer
contexts. Hence, in their hypothesis, the budgetary constraint channel
dominates the income channel. By using migration data for Greece,
Portugal, Spain and Turkey, the authors have obtained empirical
support for this hypothesis. Berthélemy et al. (2009) have reported
strong empirical evidence that both bilateral aid and recipients' total
aid exert a positive and significant impact on migrant stocks. In a
more recent study, Lanati and Thiele (2017) have obtained a negative
and significant impact of bilateral aid on emigration rates, including
for the poorer half of recipient countries. The authors therefore
conclude that the budgetary constraints channel does not play a
significant role in shaping migration decisions. Furthermore, their
empirical analysis has revealed that the negative impact of
development aid on the emigration rate is higher when recipient-
countries are in conflict situations. Thus, it is not clear whether
development aid influences positively or negatively migration flows.
However, the current study is looking into the impact of aggregate

development aid supplied by donors on the remittances sent from

donor-countries, and not on migration flows to donor-countries. Thus,
while migration flows could be an important factor explaining the
impact of development aid supplied by donors on remittances sent
from donor-countries, they might not be necessarily the only factor
that matter. While the value of remittances sent from donor-countries
would reflect the volume of migrants in these donor-countries, this
value is surely not equivalent to the number of migrants in donor-
countries. This is because the value of remittances depends not only
on the number of migrants in the donor-countries, but also on other
factors such as the wages of remitters, and the real effective
exchange rate at which the money (remittances) is sent. For example,
all things being equal, compared to donor-countries where migrants'
earnings are relatively low, donor-countries with a low number of
migrants and with higher earnings could send an important amount
of remittances to their home-countries. Another channel through
which it development aid could influence remittances paid is through
distant feeling of the existing migrants to the friends and families left in
their home countries, which can be described as lost channel of
communication. This would likely lead to lower remittances paid.
Overall, the key issue is how development aid supplied by donors
responds to these remittances paid. From a theoretical perspective,
and in light of the discussion laid out above, it would be difficult to
anticipate whether donors' aggregate aid supply would influence
positively or negatively remittances paid.
Model specification and econometric strategy
Model specification
To investigate empirically how donors' aid supply influences
remittances paid, we postulate the following model:
𝐿𝑜𝑔(𝑅𝐸𝑀𝐼𝑇)𝑖𝑡 = 𝛼0 + 𝛼1 𝐿𝑜𝑔(𝑁𝐴𝑇)𝑖𝑡 + 𝛼2 𝐿𝑜𝑔(𝐺𝐷𝑃𝐶)𝑖𝑡 +
𝛼3 𝐿𝑜𝑔(𝐹𝐼𝑁𝐷𝐸𝑉)𝑖𝑡 + 𝛼4 𝐿𝑜𝑔(𝑅𝐸𝐸𝑅)𝑖𝑡 + 𝛼5 𝐿𝑜𝑔(𝑈𝑅)𝑖𝑡 +
𝛼6 𝐿𝑜𝑔(𝑆𝑇𝑂𝐶𝐾𝑀𝐺𝑅)𝑖𝑡 + 𝜇𝑖 + 𝜔𝑖𝑡 (1)
where i is the subscript associated with a country; t denotes the time-

period. The panel dataset used in the analysis contains 21 countries 3
over the annual period 2000-2015. The panel dataset has been
constructed on the basis of data availability. 𝛼0 to 𝛼6 are
3 The list of these countries includes: Australia, Austria, Belgium, Canada, Denmark,
Finland, France, Germany, Greece, Iceland, Ireland, Italy, Netherlands, New Zealand,
Norway, Portugal, Spain, Sweden, Switzerland, United Kingdom and United States.
Gnangnon 115
parameters to be estimated. 𝜇𝑖 are countries' fixed effects; 𝜔𝑖𝑡 is a

well-behaving error term.
The dependent variable "REMIT" stands for the remittances paid, in
percentage of the donor-country's GDP.
The development aid variable is measured by the Net Aid Transfers
(denoted "NAT"), which according to Dang et al., (2013)
approximates more closely the current budgetary outlays associated
with development aid. The "NAT" is calculated by subtracting from
gross official development assistance the repayments of principal,
interest payments and the cancellation of non-ODA loans (i.e., debt
relief). The exclusion of debt relief from the definition of aid is justified
by the fact that debt cancellation does not give rise to an actual
disbursement of funds and may even imply a double counting of aid
if the debt that is cancelled was granted on a concessional basis.
Control variables include the real per capita income ("GDPC") of a
donor-country; the depth of financial development ("FINDEV") in a
donor-country; the real effective exchange rate ("REER") prevailing in
a donor-country; the unemployment rate ("UR") in a donor-country;
and the share of the stock of migrants in the total population
("STOCKMGR") in a donor-country. The description and source of each
of these variables are provided in Appendix 1. The standard
descriptive statistics on these variables are reported in Appendix 2.
Following the literature on the macroeconomic determinants of
remittances (e.g., El-Sakka & McNabb, 1999; Freund and Spatafora,
2005; Gupta, 2005; Schiopu and Siegried, 2006; Chami et al, 2008;
Adams, 2009), we expect real per capita income - which reflects the
economic wealth of the host country of migrants - to be associated
with higher remittances paid. Similarly a higher depth of financial
development in the host-country of migrants would lead to higher
flows of remittances sent from migrants' host countries to their home-
countries. Concerning the real effective exchange rate, the expect
impact on remittances paid could be ambiguous. On the one hand,
an appreciation of the real effective exchange rate, for an
unchanged value of nominal remittances, would likely increase the
real value of remittances sent from donor-countries. Remittances
senders could also increase the nominal amount of remittances sent
in a context of a depreciation of the real effective exchange rate so
that the real (net) value of remittances sent would in fine increase.
Overall, it is difficult to anticipate the impact of real effective
exchange rate on remittances paid. Incidentally, higher
unemployment rates could reduce the chances for migrants to find a
job in their host country, and hence the amount of remittances that

they can send back to home. At the same time, even in the situation
of high unemployment, migrants might be willing to accept jobs that
would not be accepted by the native people of the host-countries.
Concerning the share of the stock of migrants of the total population,
we expect a rise in this share to be positively associated with higher
remittances paid. The inclusion of this variable in the model would not
only help control for its effect on remittances paid, but as we will see
later, it would also help examine the extent to which the effect of net
aid transfers on remittances paid depends on the share of the stock
of migrants in the total population.
Econometric strategy
To estimate model (1), we start with the Pooled Ordinary Least Squares
estimator where standard errors are corrected by means of the
Driscoll and Kraay (1998) technique to account for heteroscedasticity,
serial correlation and contemporaneous cross-sectional dependence
in the dataset. We henceforth refer to this estimator as "POLS-DK". The
results of this estimation are reported in column [1] of Table 1. We then
move onto estimate the same model by means of the within fixed
effects estimator where standard errors are also corrected by means
of the Driscoll and Kraay (1998) technique. This estimator is henceforth
denoted "FE-DK", and the outcome of the estimation is provided in
column [2] of Table 1.
However, the main problem with each of these two estimators is that
they do not address the endogeneity problem associated with the
reverse causality from remittances paid to donors' ODA flows. Indeed,
as mentioned above, remittances paid flows can influence donors'
total aid flows (see Bertoli et al. 2008; and Fuchs et al. 2014), situation
which would potentially generate an endogeneity bias in the current
empirical analysis. To address this endogeneity problem, we draw
from the literature on the determinants of donors' aid effort (see for
example, the interesting survey provided by Fuchs et al. 2014 on the
matter), and use in a fixed effects regression, a number of
macroeconomic determinants of donors' development aid supply as
instruments of the "NAT" variable. These instrumental variables include
the government expenditure (% GDP) in the donor-country, denoted
"EXPEND"; the total government revenue (% GDP) in a donor-country,
denoted "REV"; the government's ideology, i.e., its political orientation,
denoted "IDEOLOGY"; the existence of a system of checks and
balances, denoted "CHECKS", and a system of polarization, denoted
"POLARIZ". The description and source of each of these instrumental
variables are provided in Appendix 1, while descriptive statistics on
these variables are reported in Appendix 2. Note that the variable
"EXPEND" has been calculated as the difference between overall
Gnangnon 117
government expenditure and the net aid transfers, in % GDP. The

latter has been subtracted from the donor's overall government
expenditure because "NAT" is part of overall government expenditure.
The empirical literature on the determinants of ODA has shown that
government expenditure and revenue, including the budget deficit
are key determinants of donors' aid effort (ODA, in % GDP) (e.g., Faini,
2006; Bertoli, 2008; Gnangnon, 2013). The degree of government
polarization, and checks and balances reflect the divergence of
interests in domestic politics and are expected to influence donors'
aid budget (e.g., Fuchs et al., 2014). Indeed, Round and Odedokun
(2004) have argued that the aid budget would increase when
incongruent ideologies are represented within a government (e.g.,
coalition of parties in power), the opposition is strong, and the system
of checks and balances involves considerable veto power. In these
conditions, higher aid could be the result of various diverging interests
that need to be satisfied in the budget. Likewise, the political
(ideological) orientation of governments (e.g., social-democrat
versus libertarian-conservative) could influence donors' aid
expenditure. For example, conservative governments may allocate
more aid to promote national commercial interests, while progressive
governments may provide a similar amount of aid for altruistic reasons
(e.g., Round and Odedokun, 2004; Faini, 2006; Bertoli, 2008; Fuchs et
al. 2014).
Using these five instruments, we estimate model (1) by means of the
fixed effects instrumental variable, namely the two-stage least
squares (2SLS) within fixed effects approach (henceforth referred to
as "FE2SLS"). It is worth underlining that standard errors are also
corrected here using the Driscoll and Kraay (1998) technique. The
conclusions of the empirical analysis would primarily be derived from
the estimations that use this estimator. The empirical analysis therefore
proceeds as follows:
First, we estimate model (1) (using the "FE2SLS" approach) over the full
sample, without taking into account the outliers identified in Figure 2
and highlighted in section 1. The results of this estimation are reported
in column [1] of Table 2. Note that to save space, we have reported
the results of the first stage regression in Appendix 3. Additionally, to
save space, we have not reported the outcomes of the first stage
regressions for the other subsequent estimations (described below),
but present in the relevant tables the tests that help check the
strength and validity of the instruments used in the regressions. These
first stage regressions could be obtained upon request.
Second, we estimate another specification of model (1), which takes
into account the effect of the outliers on the dependent variable. To
do so, we construct a dummy variable (denoted "DUMOUTL"), which

takes the value 1 for the identified outliers, and "0", otherwise. We then
introduce this dummy variable, along with its interaction with the
variable "Log(NAT)" in model (1). The resulting model specification is
estimated by means of the 'FE2SLS' approach, and the outcome of
this estimation is reported in column [2] of Table 2. As the outliers
concern situations where remittances paid (% GDP) are higher than
1% (i.e., the natural logarithm of remittances (% GDP) is higher than 0),
and where remittances paid (% GDP) are lower than 0.135% (i.e., the
natural logarithm of remittances (% GDP) is lower than -2), we deepen
the analysis by examining how development aid affects remittances
paid for cases where remittances paid (% GDP) are higher than 1%,
and for cases where remittances paid (% GDP) are lower than 0.135%.
This would help understand whether the effect of development aid
on remittances paid depends on the category (among the two major
groups of outliers) to which the countries belong. To do perform this
analysis, we create two outlier-dummies (which are indeed the
components of the dummy "DUMOUTL". The first dummy, denoted
"DUMOUTLNEG" takes the value 1 if remittances paid (% GDP) are
lower than 0.135%, i.e., if the natural logarithm of remittances (% GDP)
is lower than -2. Concerned countries include Finland, Italy, Portugal
and Spain. It is worth noting that it is not all the years of the considered
period that these countries experience remittances paid (% GDP)
lower than 0.135%. The second dummy, denoted "DUMOUTLPOS"
takes the value 1 if remittances paid (% GDP) are higher than 1%, i.e.,
if the natural logarithm of remittances (% GDP) is higher than 0 (as per
the right-hand side of Figure 2). Concerned countries include Austria,
Belgium, Denmark, Ireland, Netherlands, Norway, Switzerland. It is also
important to note here that it is not all the years of the considered
period that these countries experience remittances paid (% GDP)
higher than 1%. Each of the two dummies is interacted with the
variable "Log(NAT)" in the model (1), and the two dummies along with
their interaction variables are all introduced in model (1). The
outcome of the estimation of this model specification by means of the
'FE-2SLS' estimator is presented in column [3] of Table 2.
Third, as noted in Section 2, we examine whether and if so, to what
extent the effect of development aid on remittances paid depends
on the share of the stock of migrants in the total population of the host
countries. To conduct this analysis, we estimate another variant of
model (1) in which we introduce an interaction variable between the
development aid and the share of migrants' stocks variables. The
results of this estimation are provided in Table 3.
Gnangnon 119
Interpretation of empirical results

Across the three columns of the Table 1, we observe that
development aid does not influence significantly remittances paid at
the 10% level. At the same time, estimates provided in column [3] of
Table 1 show that for outliers, net aid transfers influences positively
remittances paid, but the effect is significant only at the 10% level. The
positive and statistically significant coefficient of the "DUMOUTL"
variable suggests that net aid transfers exert a higher positive effect
in countries where remittances paid (% GDP) are lower than 0.135%
and higher than 1%, than in other countries. Focusing on results
concerning control variables in column [3] (as these estimates stem
from the estimation of model (1) taking into account fixed effects and
the effects of outliers), we obtain that only real per capita income
and financial development exert a significant effect (at the 10% level)
on remittances paid. Specifically, these two variables influence
positively remittances paid. However, as noted above these results
especially the ones reported in columns [2] and [3] could be biased
because of the potential endogeneity of the "NAT" variable.
Let us now turn to results presented in Table 2, where estimates are
unbiased as they stem from the estimation of model (1) that addresses
the endogeneity of the "NAT" variable through the use of the 'FE2SLS'
approach. We assess the strength of the set of instruments used in the
FE2SLS regression by reporting at the bottom of all three columns of
Table 2 the results of a number of diagnostic tests, including the
Kleibergen and Paap (2006) (KP) Wald F-statistic, along with the
Hansen J statistic. The Kleibergen and Paap (2006) Wald F-statistic (KP
Wald test) is evaluated against the Stock and Yogo (2005) critical
value that corresponds to the notion that 15% is the maximal rejection
rate that the researcher could tolerate if the true rejection rate is 5%.
Results across the three columns show that the KP Wald F-statistic far
exceeds the Stock and Yogo (2005) critical value at 5% for instruments
strength, thereby indicating that the instrumental variables are strong
instruments for the "NAT" variable. Moreover, the Sanderson-
Windmeijer (SW) first-stage Chi-Squared (and the associated p-value)
are results of test of underidentification of the endogenous regressor,
under the null hypothesis that the endogenous regressor is
unidentified. Across the three columns of this Table, the statistics (and
related p-values) reject the null hypothesis, thereby indicating that
the endogenous regressor is over-identified, based on the instruments
used in the regressions. In addition, the p-values associated with the
Hansen-J statistic amounts are far higher than the 10% level across the
three columns of Table 2. These therefore suggest that the
instrumental variables used in the analysis are valid, i.e., they are
uncorrelated with the error term, and are correctly excluded from the
estimated model. Overall, these diagnostic tests that help check the
validity and strength of the instrumental variables are fully satisfactory.
Turning to the estimates, we find from column [1] that donors' net aid
transfers influence negatively and significantly (at the 1% level)
remittances paid flows. This outcome is confirmed in column [2] of
Table 2 when we take into account outliers, although the coefficient
of the "Log(NAT)" variable in column [2] is slightly higher than that of
column [1] of the same Table. At the same time, for outliers, the effect
of development aid on remittances paid is positive and statistically
significant at the 1% level (see the coefficient of the interaction
variable "DUMOUTL*Log(NAT)". It is worth recalling here that
according to the descriptive statistics provided in Appendix 2,
remittances paid (% GDP) range between 0.024% and 3.807%. Based
on results in column [2], we conclude that for outliers, a 1 percentage
increase in the ratio of net aid transfers to GDP leads to a rise in
remittances paid to GDP ratio by 0.75 percentage. However, for non-
outliers (countries where remittances to GDP ratio is comprised
between 0.135% and 1%), a 1 percentage increase in the ratio of net
aid transfers to GDP leads to a 1.3% percentage decline in
remittances paid-to-GDP ratio. Put differently, for non-outliers, there is
a substitution effect between donors' aid and remittances flows sent
from donor-countries, as when donors' aid supply declines,
remittances increase, and vice-versa. In contrast, for outliers,
development aid is complementary with remittances paid. Results
concerning control variables (see column [2]) indicate that positive
drivers of remittances paid include a rise in the real per capita
income, a higher depth of financial development, a depreciation of
real effective exchange rate, and a higher unemployment rate. The
surprisingly positive effect of unemployment rate on remittances paid
may yet reflect differentiated effect across countries in the sample,
but it could also indicate that the effect of development aid on
remittances paid may be dependent on the prevailing level of
unemployment rate in the host countries of migrants (here, donor-
countries). We do not go deeper into this analysis, as it is not the main
purpose of the current study. The share of migrants' stock in the total
population does not exert any significant effect on the remittances
paid by these migrants. This result probably indicates that the effect
of development aid on remittances paid translates through its effect
on migration flows (and hence the prevailing stock of migrants in a
given year in the host country). Let us now turn to estimates provided
in column [3] of Table 2. We obtain that for non-outliers, development
aid influences negatively and significantly (at the 1% level)
remittances paid (see the coefficient of "Log(NAT)" variable). At the
same time, among outliers, in countries where remittances paid (%
Gnangnon 121
GDP) are lower than 0.135%, net aid transfers exert a positive and
significant effect (at the 1% level) (see the coefficient of the
interaction variable "DUMOUTLNEG*Log(NAT)"). Specifically, in these
countries, a 1 percentage increase in the ratio of net aid transfers to
GDP is associated with an increase in remittances paid to GDP ratio
by 1.12 percentage. However, for outlier-countries where remittances
(% GDP) are higher than 1%, there is on average, no significant effect
of development aid on remittances paid. Overall, these results tend
to suggest that while for non-outliers (countries where remittances
paid to GDP ratio is comprised between 0.135% and 1%), there is a
substitution between development aid and remittances paid, for
outliers, there is a complementarity between these two types of
capital flows, although this complementarity seems to be driven by
cases where remittances-to-GDP ratio is lower than 0.135% (for
countries where remittances is higher than 1%, there is no significant
effect of development aid on remittances paid).
We now consider the results reported in Table 3. As noted above, they
help examine the extent to which the effect of net aid transfers on
remittances paid depends on the share of the stock of migrants in the
total population in the host-countries. The results concerning the
diagnostics tests that help check the validity and strength of
instruments used in the regression are all fully satisfactory. In this Table,
we are particularly interested in the interaction terms associated with
the interaction variables, as well as with the coefficients of the
variable "Log(NAT)". We note that the coefficient of the "Log(NAT)" is
positive and statistically significant at the 1% level, while concurrently,
the interaction term associated with the interaction variable
"[Log(STOCKMGR)]*[Log(NAT)]" is negative and statistically significant
at the 1% level. Taking together, these two results suggest that for non-
outliers, the total effect of net aid transfers on remittances paid
changes sign above a certain threshold of the share of stock of
migrants in the host-country's total population. This threshold is given
by 15.43% [= exponential (3.453/1.262)]. To recall, the share of
migrants' stock in the host-country's total population ranges between
2.5% and 28.4%. This outcome therefore signifies that on average, for
non-outliers, the effect of net aid transfers on remittances paid is
positive for shares of stocks of migrants in the host-countries' total
population lower than 15.43%, and negative for shares of stocks of
migrants in the host-countries' total population higher than 15.43%.
However, this outcome may not provide the full extent of the effect
of development aid on remittances paid, as it represents an 'average'
outcome over all non-Outlier-countries of the full sample. In fact, the
effect of net aid transfers on remittances paid could exhibit various
effects (including in terms of magnitude, statistical significance and
sign of the estimated coefficient) across countries (non-outliers) in the

full sample, for different shares of the stock of migrants in the host-
country's total population. To get a better picture on this impact, we
provide in Figure 3, at the 95 per cent confidence intervals, the
evolution of the marginal impact of "NAT" on "REMIT" for different
shares of the stock of migrants in the host-country's total population
(for non-outliers). It is important to underline that the marginal impacts
that are statistically significant at the 95 per cent confidence intervals
are those encompassing only the upper and lower bounds of the
confidence interval that are either above or below the zero line. This
Figure indicates that the marginal impact of net aid transfers on
remittances paid could be positive or negative, and decreases as the
share of the stock of migrants in the non-outlier host-country's total
population countries increases. Specifically, the effect of
development aid on remittances paid is positive and significant in
those countries (among non-outliers) where the share of the stock of
migrants in the total population is lower than 13.76% [= exponential
(2.622093)]. For this set of countries, the magnitude of the effect of net
aid transfers on remittances paid is yet positive, but it declines as the
share of the stock of migrants in the total population increases. Put
differently, for those countries, net aid transfers exert a higher positive
effect on remittances paid in countries with lower share of stock of
migrants in the total population than in countries with higher share of
stock of migrants in the total population. For countries (among non-
outliers) that experience a share of the stock of migrants in the total
population higher than 16.7% [= exponential (2.815537)], the effect of
development aid on remittances paid is negative and significant, and
the higher the value of this share, the higher is the magnitude of the
negative effect of development aid on remittances paid. Thus, in non-
outlier countries where the share of the stock of migrants in the total
population is higher than 16.7%, there is a substitutability effect
between net aid transfers and remittances paid, and the degree of
this substitutability increases as the value of the share of the stock of
migrants in the total population rises. For countries (among non-
outliers) that experience a share of stock of migrants in the total
population between 13.76% and 16.7%, there is no significant effect
of development aid on remittances paid.
Turning to results concerning outliers in Table 3, we obtain that the
coefficient associated with the variable "[DUMOUTL*Log(NAT)]" is not
statistically significant at the 10% level, whereas the interaction term
related to the variable "[Log(STOCKMGR)]*DUMOUTL*[Log(NAT)]" is
positive and statistically significant at the 5% level. The combination of
these two results leads to the conclusion that, on average, for outlier-
countries, net aid transfers influence positively and significantly
Gnangnon 123
remittances paid, whatever the share of migrants' stock in the

population of the host-country. In addition, the higher this share, the
higher is the positive effect of net aid transfers on remittances paid.
As noted above, this 'average' outcome over the outliers may not fully
reflect how net aid transfers affect remittances paid for various shares
of migrants' stock in the total population of the host-country. To get a
better picture on this effect, we display in Figure 4, at the 95 per cent
confidence intervals, the evolution of the marginal impact of "NAT" on
"REMIT" for different shares of the stock of migrants in the host-country's
total population for outliers in the sample. The graph shows that this
marginal effect increases as the share of the stock of migrants in the
host-country's total population increases. However, it is yet always
positive, but not always statistically significant. In particular, this
marginal effect is statistically significant only for shares of the stock of
migrants in the host-country's total population higher than 11.9% [=
exponential (2.47701)]. Hence, outlier-countries that experience a
share of migrants' stock in the host-country's total population higher
than 11.9% experience a positive and significant effect of net aid
transfers on remittances paid, and the higher the share, the greater is
the positive effect of net aid transfers on remittances paid. At the
same time, outlier-countries with a share of migrants' stock in the host-
country's total population lower than 11.9% enjoy no significant effect
of net aid transfers on remittances paid.
Robustness check analysis
To check the robustness of results concerning the effect of net aid
transfers on remittances paid found in Table 2, especially in columns
[1] and [2] of this Table, we use another econometric estimator to
estimate the related specifications of model (1) (i.e., model (1) that
does not take into account outliers, and model (1) that takes into
account outliers). In particular, we use the Error Component two-
stage least squares, referred to as "EC2SLS", developed by Baltagi
(1981). It is a random effects estimator with a matrix weighted
average of the between 2SLS and fixed effects 2SLS approaches. It is
particularly more efficient in small samples and outperforms the
Generalized two-stage least square estimator (Baltagi and Liu, 2009).
The same instruments used for the estimations with the FE2SLS
approach are also employed here. The results of these estimations are
presented in Table 4. The p-values associated with the Sargan test of
overidentification of all instruments are higher than 10% across the two
columns of the Table, thereby indicating the validity of the instruments
used in the regressions. We note from column [1] of this Table that net
aid transfers affect negatively and significantly remittances paid.
However, the coefficient of the "Log(NAT)" variable in column [2]
(where outliers' effects are taking into account) of this Table more
than doubles that of column [1]. Specifically, for non-outliers, a 1

percentage increase in the net aid transfers-to GDP-ratio leads to a
1.35 percentage fall in the ratio of remittances paid to GDP. This result
is quite similar to the one obtained in column [2] of Table 2. In contrast,
for outliers, a 1 percentage increase in the net aid transfers-to GDP-
ratio induces a 1.75 percentage increase in the ratio of remittances
paid to GDP. Even though the magnitude of this effect is higher than
the one obtained in column [2] of Table 2 for outlier-countries, the two
effects go in the same direction, i.e., they are both positive and
statistically significant at least at the 5% level. Overall, these findings
confirm those of column [2] of Table 2, although with different
magnitudes. Results related to control variables in column [2] of Table
4 suggest that higher real per capita income and a depreciation of
the real effective exchange rate exert a positive and significant effect
on remittances paid. Incidentally, remittances paid to GDP ratio
exhibits a rising trend, as exemplified by the positive and statistically
significant coefficient of the "Trend" variable. The other control
variables do not appear to influence significantly (at least at the 10%
level) remittances paid. In particular, the lack of statistical significance
of the effect of the share of migrants' stock (in the host-country's total
population) on remittances paid probably reflects the fact that the
effect of net aid transfers on remittances paid translates through the
share of migrants' stock (in the host-country's total population).
Conclusion
In contrast with previous studies that have looked at the impact of
development aid on remittances received by recipient (developing
countries), this article examines the impact of the aggregate
development aid supplied by donors on the remittances flows sent
from donor-countries (remittances paid). The empirical analysis is
carried out focuses on 21 donor-countries (host countries of migrants)
over the period 2000-2015. It has shown evidence that donors' overall
aid supply could exert a negative or a positive and significant impact
on remittances paid, depending on the level of remittances-to-GDP
ratio sent from host-countries (i.e., the donor-countries), as well as on
the share of migrants' stock in the host-country's total population.
Specifically, for countries where remittances paid to GDP ratio is
comprised between 0.135% and 1%, development aid is a substitute
to remittances paid, especially when the share of the stock of
migrants in the total population is higher than 16.7%; the degree of this
substitutability increases as the value of the share of the stock of
migrants in the total population rises. This substitutability between net
aid transfers and remittances paid could be interpreted by the fact
that development aid reduces migration flows (hence the
importance of the income channel relative to the budgetary
Gnangnon 125
constraint channel in the relationship between aid and migration),

and consequently lead to lower remittances paid. However, when
the share of the stock of migrants in the total population is lower than
13.76%, net aid transfers influence positively remittances paid, i.e.,
there is a complementarity between these two types of capital flows.
For countries (among those with a ratio of remittances paid to GDP
between 0.135% and 1%) that experience a share of stock of migrants
in the total population between 13.76% and 16.7%, there is no
significant effect of development aid on remittances paid.
Considering now countries where the ratio of remittances paid to GDP
is lower than 0.135% or higher than 1%, we find that net aid transfers
influence significantly and positively remittances paid only for those
countries where remittances paid (% GDP) are lower than 0.135%, as
for the other countries, the effect is not statistically significant.
Furthermore, for those countries with a ratio of remittances paid to
GDP lower than 0.135% or higher than 1%, the effect of net aid
transfers on remittances paid is statistically significant only when the
share of migrants' stock in the host-country's total population is higher
than 11.9%. This statistically significant effect is in effect positive, and
increases as the share of migrants' stock in the host-country's total
population rises.
References
Adams JR, R.H. (2009). The Determinants of International Remittances in
Developing Countries. World Development, 37(1), 193–103.
Adams R., and Page J., Jr. (2003). International migration, remittances and poverty
in developing countries. World Bank Policy Research Working Paper 3179,
Washington, DC.
Baltagi, B. H. (1981). Simultaneous equations with error components. Journal of
Econometrics, 17 (2), 189-200.
Baltagi, B. H., and. Liu, L (2009). A Note on the Application of EC2SLS and EC3SLS
Estimators, In Panel Data Models. Center for Policy Research Working Papers
116, Center for Policy Research, Maxwell School, Syracuse University, United
States.
Beck T. C., Groff G., Keefer A. P., and Walsh P. 2001. New tools in comparative
political economy: The database of political institutions. World Bank Economic
Review, 15(1): 165–176.
Berthélemy J.-C., Beuran M. and Maurel, M. 2009. Aid and Migration: Substitutes or
Complements? World Development 37(10): 1589-1599.
Chami R., Barajas, A., Cosimano, T., Fullenkamp, C., Gapen, M., and Montiel, P.
2008. Macroeconomic Consequences of Remittances, IMF Occasional Paper
No. 259 (Washington: International Monetary Fund).
Clemens M. 2014. 'Does Development Reduce Migration?', Chapter 6 in Robert
Lucas (ed), International Handbook on Migration and Economic
Development, Edward Elgar.

Dang H-A., Knack S. and Rogers H. 2010. International aid and financial crises in
donor countries. European Journal of Political Economy, 32, pp 232–250. DOI:
https://doi.org/10.1016/j.ejpoleco.2013.08.003
Driscoll J. C., and Kraay A.C. 1998. Consistent Covariance Matrix Estimation with
Spatially Dependent Panel Data. Review of Economics and Statistics, 80(4):
549–560.
El-Sakka, M. I. T., and McNabb, R. 1999. The Macroeconomic Determinants of
Emigrant Remittances. World Development, 27(8): 1493-1502.
Faini R. 2006. Foreign aid and Fiscal Policy, CEPR Discussion Paper No. 5721.
Faini R., and Venturini A. 1993. Trade, aid and migrations: Some basic policy issues.
European Economic Review, 37(2-3): 435–442.
Freund C., and Spatafora, N. 2005. Remittances, Transaction Costs, Determinants,
and Informal Flows. World Bank Policy Research Working Paper, 3704.
Fuchs A., Dreher A., and Nunnenkamp P. 2014. Determinants of Donor Generosity:
A Survey of the Aid Budget Literature. World Development, 56, 172-199.
Gnangnon S.K. 2013. The consequences of fiscal episodes in OECD DAC countries
for aid supply. The Quarterly Review of Economics and Finance, 53(3): 302-313.
Gupta, P. (2005). Macroeconomic Determinants of Remittances: Evidence from
India. IMF Working Paper No. 05/224, International Monetary Fund,
Washington, D.C.
Hatton T., and Williamson J. G. 2002. What fundamentals drive world migration?.
In G. Borjas, & J. Crips (Eds.), Poverty, international migration and asylum (pp.
15–38). Hampshire, UK: Palgrave-Macmillan for WIDER.
Kleibergen F., and Paap R. 2006. Generalized Reduced Rank Tests Using the
Singular Value Decomposition. Journal of Econometrics 133, 97-126.
Lanati M., and Thiele R. (2017). The impact of foreign aid on migration revised.
Working Paper, MPW 2017/05, European University Institute, Italy.
Parsons C.R. and Winters L.A. 2014. 'International Migration, Trade and Aid: A
Survey'. In: R.E.B. Lucas (Ed.), International Handbook on Migration and
Economic Development. Edward Elgar Publishing.
Round J. I. and Odedokun M. 2004. Aid effort and its determinants. International
Review of Economics and Finance, 13(3): 293-309.
Schiff, M. 1994. How Trade, Aid, and Remittances Affect International Migration.
Policy Research Working Paper, no. 1376, the World Bank.
Schiopu I., and Siegfried N. 2006. Determinants of Workers’ Remittances: Evidence
for the European Neighbouring Region. European Central Bank Working Paper
Series 688.
Stock, J., Yogo, M. 2005. Testing for Weak Instruments in Linear IV Regression. In
Andrews D.W.K. Identification and Inference for Econometric Models. New
York: Cambridge University Press.
Tables and Appendices

Figure 1: Evolution of Net Aid Transfers (%) and Remittances Paid (%
GDP)
Gnangnon 127
0.8 0.5
0.7 0.45
0.4
0.6
0.35
0.5
REMIT
0.3
NAT
0.4 0.25
0.3 0.2
REMIT
0.15
0.2 NAT 0.1
0.1 0.05
0 0
2005
2010
2000
2001
2002
2003
2004
2006
2007
2008
2009
2011
2012
2013
2014
2015
Year
Source: Author
Figure 2: Correlation pattern between Net Aid Transfers (%) and

Remittances Paid (% GDP)
4
2
3
0
Lo gREMI T
REMIT
-2
1
0
-4
0 .5 1 1.5 -3 -2 -1 0 1
NAT LogN AT
REMIT Fitted values LogR EM IT Fitted values
Source: Author
Figure 3: Marginal Impact of "NAT" on "REMIT", for varying levels of
stock of migrants_Over non-Outliers in the full sample

Predictive Margins with 95% CIs

3
2
1
0
-1
1 1.5 2 2.5 3 3.5

Log(STOCKMGR)
Source: Author
Figure 4: Marginal Impact of "NAT" on "REMIT", for varying levels of
stock of migrants_Over Outliers in the full sample
Predictive Margins with 95% CIs
1
.5
0
-.5
1 1.5 2 2.5 3 3.5

Log(STOCKMGR)
Source: Author
Table 1: Impact of Net Aid Transfers on Remittances Paid
Estimator: Within Fixed Effects with Driscoll-Kraay (1998) standard errors
Gnangnon 129
POLS-DK FE-DK FE-DK

VARIABLES Log(REMIT) Log(REMIT) Log(REMIT)
(1) (2) (3)
Log(NAT) 0.00479 0.0443 -0.0313
(0.0814) (0.138) (0.139)
DUMOUTL*Log(NAT) 0.284*
(0.160)
DUMOUTL 0.368***
(0.114)
Log(GDPC) 1.132*** 2.804*** 2.992***
(0.151) (0.618) (0.625)
Log(FINDEV) -0.367*** 0.436*** 0.386***
(0.0775) (0.118) (0.108)
Log(REER) 0.660*** -0.483 -0.507
(0.204) (0.281) (0.302)
Log(UR) -0.317** 0.0344 0.0790
(0.114) (0.0644) (0.0638)
Log(STOCKMGR) 0.447** 0.0433 -0.0202
(0.164) (0.256) (0.259)
Constant -14.77*** -30.90*** -32.60***
(0.975) (6.060) (5.897)
Observations - 240 - 21 240 - 21 240 - 21

Countries
R-squared/ Within R-
0.351 0.3704 0.4216
squared
Note: *p-value < 0:1; **p-value < 0:05; ***p-value < 0:01. Robust Standard errors are in
parenthesis.


Estimator: Instrumental variables Fixed Effects with Driscoll-Kraay
(1998) standard errors
VARIABLES Log(REMIT) Log(REMIT) Log(REMIT)
(1) (2) (3)
Log(NAT) -0.858*** -1.298*** -1.021***
(0.133) (0.309) (0.215)
DUMOUTL*Log(NAT) 0.753***
(0.165)
DUMOUTLNEG*Log(NAT) 1.124***
(0.375)
DUMOUTLPOS*Log(NAT) -0.0372
(0.0485)
DUMOUTL 0.609***
(0.118)
DUMOUTLNEG 0.323
(0.303)
DUMOUTLPOS 0.434***
(0.0659)
Log(GDPC) 4.590*** 5.798*** 5.075***
(0.501) (1.179) (0.666)
Log(FINDEV) 0.579*** 0.481*** 0.253***
(0.0353) (0.0603) (0.0897)
Log(REER) -1.006*** -1.266*** -0.678***
(0.208) (0.386) (0.236)
Log(UR) -0.0103 0.112*** 0.206***
(0.0246) (0.0389) (0.0507)
Log(STOCKMGR) 0.336*** 0.159 0.0747
(0.0710) (0.181) (0.140)
Observations - Countries 240 - 21 240 - 21 240 - 21
R-squared 0.159 0.050 0.353
Joint F-statistic (P-value) 260.64 2594.05 4323.71

(0.0000) (0.0000) (0.0000)
Sanderson-Windmeijer First Stage 141.88
227.19 (0.0000) 289.73 (0.0000)
Chi-Square test: Statistic (P-value) (0.0000)
Weak identification test
(Kleibergen–Paap rk Wald F 32.141 51.020 64.496
statistic)
Stock and Yogo (2005) weak ID test
critical values at 5% maximal IV 16.85 16.85 16.85
relative bias
Hansen-J Statistic
(overidentification test of all 1.984
1.922 (0.5887) 1.839 (0.6064)
instruments) Chi-square Statistic (P- (0.5758)
value)
parenthesis. The variables "NAT" has been instrumented in the regressions by the
following variables "IDEOLOGY", "EXPEND", "REV", "CHECKS", and "POLARIZ". Standard
errors are corrected for cross-sectional dependence, using Driscoll and Kraay (1998)
technique.
Gnangnon 131
Table 3: Does the impact of Net Aid Transfers on Remittances Paid

depend on the stock of migrants?
Estimator: Instrumental variables Fixed Effects with Driscoll-Kraay
(1998) standard errors
VARIABLES Log(REMIT)
(1)
Log(NAT) 3.453***
(0.495)
DUMOUTL 0.572***
(0.158)
DUMOUTL*Log(NAT) -0.0904
(0.352)
[Log(STOCKMGR)]*[Log(NAT)] -1.262***
(0.187)
[Log(STOCKMGR)]*DUMOUTL*[Log(NAT)] 0.176**
(0.0802)
Log(GDPC) 1.570***
(0.451)
Log(FINDEV) 0.354***
(0.0849)
Log(REER) 0.517*
(0.272)
Log(UR) 0.0913
(0.0660)
Log(STOCKMGR) -1.530***
(0.118)
Observations - Countries 240 - 21
Joint F-statistic (P-value) 7372.89 (0.0000)

Sanderson-Windmeijer First Stage Chi-Square test:
85.86 (0.0000)
Statistic (P-value)
Underidentification test (Kleibergen–Paap rk LM
2.247 (0.6905)
statistic) [P-value]
Weak identification test (Kleibergen–Paap rk Wald F
19.112
statistic)
Stock and Yogo (2005) weak ID test critical values at
16.85
5% maximal IV relative bias
Stock and Yogo (2005) weak ID test critical values at
10.27
10% maximal IV relative bias
Hansen-J Statistic (overidentification test of all
2.174 (0.5370)
instruments) Chi-square Statistic (P-value)
parenthesis. The variables "NAT" has been instrumented in the regressions by the
following variables "IDEOLOGY", "EXPEND", "REV", "CHECKS", and "POLARIZ". Standard
errors are corrected for cross-sectional dependence, using Driscoll and Kraay (1998)
technique.


Estimator: Error component EC2SLS random effects
VARIABLES Log(REMIT) Log(REMIT)
(1) (2)
Log(NAT) -0.673*** -1.346***
(0.244) (0.426)
DUMOUTL*Log(NAT) 1.747**
(0.739)
DUMOUTL 1.011
(0.684)
Log(GDPC) 2.341*** 3.128***
(0.527) (0.715)
Log(FINDEV) 0.514*** 0.159
(0.125) (0.196)
Log(REER) -0.978*** -1.133***
(0.304) (0.408)
Log(UR) -0.242*** -0.0574
(0.0909) (0.149)
Log(STOCKMGR) 0.0704 -0.249
(0.157) (0.224)
Trend 0.0346*** 0.0460***
(0.00789) (0.0124)
Constant -93.63*** -122.8***
(15.78) (26.83)
Observations - Countries 240 - 21 240 - 21

Within R2 0.3124 0.2160
Between R2 0.2513 0.3343
Overall R2 0.2450 0.3284
Robust Sargan-Hansen
(overidentifying restrictions 13.962 (0.3765) 9.568 (0.5696)
test) Statistic (P-value)
parenthesis. The variable "NAT" has been instrumented in the regressions by the
following variables "IDEOLOGY", "EXPEND", "REV", "CHECKS", and "POLARIZ". As the
Driscoll and Kraay (1998) technique could not apply here, we have introduced a
trend variable in the regressions based on the EC2SLS approach.
Gnangnon 133
APPENDICES
Appendix 1: Definitions and sources of variables
Variable Definition Source
Authors' calculation based on
data on Personal remittances,
Personal remittances, paid (% paid (current US$) and GDP
REMIT
of GDP) (current US$) extracted from the
World Development Indicators
(WDI) of the World Bank
Net Aid Transfers (NAT), in % of GDP data (in current prices) are
GDP. This is the net Official extracted from the World
Development Assistance Development Indicators of the
(ODA), from which are World Bank (WDI). NAT data (in
NAT
subtracted principal payments current prices) are extracted from
are received on ODA loans, the database compiled by David
interest received on such loans Roodman (see online:
and debt relief. http://davidroodman.com/data/)
GDP per capita (constant 2010
GDPC WDI
US$)
Total Government Revenue of Author's computation based on
REV a Donor-Country, in % GDP. data sourced from different IMF
Public Finance Databases
Author's computation based on
Difference between total
data sourced from different IMF
Government Expenditure and
EXPEND Public Finance Databases for
the donor's Net Aid Transfers
Government Expenditure.
("NAT"), in % GDP.
Domestic credit to private
FINDEV WDI
sector (% of GDP)
Real effective exchange rate
REER WDI
index (2010 = 100)
Share (%) of stock of migrants in OECD International Migration
STOCKMGR
the total population Database
UR Unemployment Rate (%) OECD Economic Outlook
Party orientation with respect
to economic policy; Right (1);
Database on Political Institutions
IDEOLOGY Left (3); Center (2); No
(DPI) 2015 (Beck et al., 2001)
information (0); No executive
(NA) (code: EXECRLC).
This variable refers to "Checks
and Balances". It is the Extent of
CHECKS formal political control on DPI 2015 (Beck et al., 2001)
political decision makers
(code: CHECKS).
This variable captures the
maximum difference between
the chief executive’s party’s
value and the
POLARIZ DPI 2015 (Beck et al., 2001)
values of the three largest
government parties and the
largest opposition party (code:
POLARIZ).

Appendix 2: Descriptive statistics on variables used in the analysis

Variable Observations Mean Standard Minimum Maximum
Deviation
REMIT 336 0.584 0.592 0.024 3.807
NAT 335 0.388 0.253 0.091 1.439
GDPC 336 45705.930 14487.110 21228.090 91617.280
FINDEV 312 119.857 39.892 50.081 312.118
REER 336 99.673 10.444 66.851 152.607
UR 336 7.306 4.008 2.248 27.466
STOCKMGR 251 12.227 5.528 2.533 28.432
IDEOLOGY 336 1.711 1.035 0.000 3.000
EXPEND 336 44.734 7.234 28.646 66.419
REV 336 43.593 7.819 27.556 59.675
CHECKS 336 4.146 1.068 2.000 7.000
Appendix 3: Results of the First Stage Regression for the estimation of the baseline model
whose results are reported in column [1] of Table 2
Dependent variable: Log(NAT)
IDEOLOGY 0.021***
(0.0071)
Log(EXPEND) 0.7818***
(0.2257)
Log(REV) -1.1687***
(0.1466)
CHECKS 0.0594***
(0.0099)
Log(GDPC) 1.937***
(0.409)
Log(FINDEV) 0.1948***
(0.0239)
Log(REER) -0.574***
(0.174)
Log(UR) -0.144***
(0.046)
Log(STOCKMGR) 0.2022**
(0.09625)
parenthesis.

Impact of Development Aid On Remittances Sent From Donor-Countries

Enviado por

Dados do documento

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Impact of Development Aid On Remittances Sent From Donor-Countries

Enviado por

Direitos autorais:

Formatos disponíveis

October 2018

Volume: 3, No: 2, pp. 109 – 134

Article history: Received 8 June 2018; accepted 12 October 2018

Impact of development aid

Copyright @ 2018 REMITTANCES REVIEW © Transnational Press London

quality as well as some demographic and financial/political

percentage of donor-countries' Gross Domestic Product (GDP)

of migrants flows from developing countries towards donor-countries,

would discourage people from emigrating and hence, reduce

Copyright @ 2018 REMITTANCES REVIEW © Transnational Press London

development aid supplied by donors on the remittances sent from

where i is the subscript associated with a country; t denotes the time-

parameters to be estimated. 𝜇𝑖 are countries' fixed effects; 𝜔𝑖𝑡 is a

Copyright @ 2018 REMITTANCES REVIEW © Transnational Press London

government expenditure and the net aid transfers, in % GDP. The

do so, we construct a dummy variable (denoted "DUMOUTL"), which

Interpretation of empirical results

sign of the estimated coefficient) across countries (non-outliers) in the

remittances paid, whatever the share of migrants' stock in the

than doubles that of column [1]. Specifically, for non-outliers, a 1

constraint channel in the relationship between aid and migration),

Copyright @ 2018 REMITTANCES REVIEW © Transnational Press London

Tables and Appendices

Figure 2: Correlation pattern between Net Aid Transfers (%) and

REMIT Fitted values LogR EM IT Fitted values

Copyright @ 2018 REMITTANCES REVIEW © Transnational Press London

Predictive Margins with 95% CIs

1 1.5 2 2.5 3 3.5

1 1.5 2 2.5 3 3.5

POLS-DK FE-DK FE-DK

Observations - 240 - 21 240 - 21 240 - 21

Copyright @ 2018 REMITTANCES REVIEW © Transnational Press London

Table 2: Impact of Net Aid Transfers on Remittances Paid

Joint F-statistic (P-value) 260.64 2594.05 4323.71

Table 3: Does the impact of Net Aid Transfers on Remittances Paid

Joint F-statistic (P-value) 7372.89 (0.0000)

Copyright @ 2018 REMITTANCES REVIEW © Transnational Press London

Table 4: Impact of Net Aid Transfers on Remittances Paid

Observations - Countries 240 - 21 240 - 21

Copyright @ 2018 REMITTANCES REVIEW © Transnational Press London

Appendix 2: Descriptive statistics on variables used in the analysis

Dependent variable: Log(NAT)

Você também pode gostar