A partnership with academia

Building knowledge for trade and development

Vi Digital Library - Text Preview

Do Remittances Reduce Vulnerability to Climate Variability in West African Countries? Evidence from Panel Vector Autoregression

Discussion paper by Couharde, Cécile, Davis, Junior, Generoso, Rémi, 2011

Download original document (English)

In this paper, we empirically examine the role of remittances in smoothing the GDP fluctuations induced by precipitation variability and both meteorological and natural shocks. To this end, we use a panel VAR to empirically study six West African countries from 1983 to 2009. Our evidence suggests that remittances are an important element of macroeconomic stability especially for those countries most vulnerable to precipitation variability. The estimated orthogonalized impulse responses show on one hand, that meteorological shocks and declining precipitation have both adverse consequences on GDP per capita. On the other hand, remittances are characterized by counter-cyclical patterns in cases of precipitation variability and climate shocks. Remittances inflows in the selected countries (countries of emigration) are also heavily dependent on economic shocks in host countries.




1


Do Remittances
reduce vulnerability
to climate variability
in West African
Countries? Evidence
from panel vector
autoregression
UNCTAD Special Unit for
Commodities working paper
series on commodities and
development

September 2011


Cécile Couharde, Junior Davis
and Rémi Generoso


Discussion paper 2


The papers in this series are the preliminary results
of research undertaken by SUC staff members and
do not necessarily reflect the views of UNCTAD. They
are published in this form to stimulate discussion
and comment on work that is generally still in
progress.







2



Do Remittances reduce Vulnerability to Climate Variability in


West African Countries? Evidence from Panel Vector


Autoregression







Cécile Couharde*, Junior Davis†, Rémi Generoso‡



Preliminary version





Abstract:


In this paper, we empirically examine the role of remittances in smoothing the GDP
fluctuations induced by precipitation variability and both meteorological and natural
shocks. To this end, we use a panel VAR to empirically study six West African countries
from 1983 to 2009. Our evidence suggests that remittances are an important element of
macroeconomic stability especially for those countries most vulnerable to precipitation
variability. The estimated orthogonalized impulse responses show on one hand, that
meteorological shocks and declining precipitation have both adverse consequences on
GDP per capita. On the other hand, remittances are characterized by counter-cyclical
patterns in cases of precipitation variability and climate shocks. Remittances inflows in
the selected countries (countries of emigration) are also heavily dependent on
economic shocks in host countries.




Keywords: Climatic variability, Panel VAR, Remittances, West Africa
JEL classification: E30; F24; O11



* EconomiX, University of Paris Ouest-Nanterre La Défense, France. E-mail:cecile.couharde@u-paris10.fr
† United Nations, UNCTAD, Geneva. E-mail: junior.davis @unctad.org
‡ CEMOTEV, University of Versailles Saint-Quentin. E-mail: rgeneros@ens.uvsq.fr (corresponding author)





3


1. Introduction

The least developed countries (LDCs) of West Africa are at tremendous risk from climatic
shocks such as shifting weather patterns and environmental degradation and suffer the
greatest burden of adjusting to threats of climate change because they are already challenged
by what is known as ‘multiple vulnerabilities’ on account of their low levels of economic and
human development (UN-DESA, 2009a: 71). Clearly, West Africans face a future where a lack
of social and physical infrastructure, missing institutions and a narrow economic base may be
‘exposed not just to potentially catastrophic large-scale disasters but also to a more
permanent state of economic stress as a result of higher average temperatures, reduced
availability of water sources, more frequent flooding and intensified windstorms’ (ibid).


Many studies have sought to analyse the impact of external shocks, especially climatic shocks,
on gross domestic product (GDP) volatility in developing countries. Most of these studies have
concluded that there is a relatively modest impact of external shocks, particularly climatic
shocks on developing countries’ GDP volatility. For example, Raddatz (2009) has argued that
the low impact of external shocks on GDP fluctuations may be explained by domestic factors
such as the level of inflation, overvalued real exchange rates, civil wars, corruption and/ or
large public deficits. Another potential explanation concerns the existence of counter-cyclical
factors1 that smooth GDP fluctuations to climatic shocks. Noy (2009) using a sample of 109
countries for the period 1970-2003 includes structural factors such as the literacy rate, level
of education, quality of institutions and trade openness in the analysis on the impact of shocks
on GDP fluctuations using a variant of Raddatz’s (2007) model. Noy (2009) concludes that
countries with a higher literacy rate, better institutions, higher per capita income, a greater
degree of openness to trade, higher levels of government spending, foreign exchange reserves,
and levels of domestic credit, but with less open capital accounts are better able to withstand
the initial shock of a climatic disaster and limit wider spill-over effects.




1 In theory, any economic quantity that is positively correlated with the overall state of the economy is
considered pro-cyclical. Thus, any quantity that tends to increase when the overall economy is growing is
classified as pro-cyclical. Quantities that tend to increase when the overall economy is slowing down are
classified as 'counter-cyclical'. Thus, in developing countries remittances tend to rise during periods of financial
crises.





4


Using a generalised linear regression model Hochrainer (2009) highlights the role of aid and
remittance inflows in reducing the adverse macroeconomic consequences of shocks. The
author uses an autoregressive integrated moving average model (ARIMA) to forecast GDP into
the medium term following a disaster event. In order to extrapolate GDP trends, the evolution
of GDP following a shock is compared with GDP trends in the absence of a disaster.
Hochrainer (2009) then tests several explanatory variables (termed vulnerability predictors)
to explain variations in projected and observed GDP 5 years after a disaster event. Finally the
limited impact of climatic shocks on GDP volatility may be explained by the nature of the
shock itself. Loaysa et al. (2009) focus on the analysis of climate shocks and their impact on 94
developing countries from 1961 to 2005 on the basis of a Generalized Method of Moment
(GMM) model. The authors identified different impacts according to the type of climate shock.
Certain disasters (in particular low-span disasters e.g. low covariant disasters such as floods)
can also have a positive economic impact, if they induce investment and the reconstruction of
an economic sector. In contrast, more substantial span disasters (with a stronger covariance
e.g. droughts) have negative consequences and an almost immediate impact on the economy
of the poorest countries.


The paper aims to analyze the role of remittances in explaining the relationship between GDP
instability and climate shocks in six West African countries. Thus, the paper seeks to examine
the extent to which remittance receipts, climatic and biological shocks as well as rainfall
patterns help to explain the volatility of GDP growth in West Africa after controlling for a
number of climatic variables that have been cited in the literature as potential determinants of
GDP volatility.

To do so, we elaborate a methodology which differs from the cited studies in several ways.
First, in addition to the notion of shocks, we focus on climate variability by assessing the
impact of shocks in precipitation patterns in relation to GDP fluctuations. Indeed, sub-Saharan
Africa (SSA) is one of the most vulnerable continents to climate variability and in particular to
both intra and inter-annual precipitation variability. This vulnerability to climate variability
may also be aggravated by the interaction and occurrence of multiple stresses coupled to a
relatively low adaptive capacity of SSA populations. The economic impact on the agricultural
sector – an activity which accounts for 70 percent of the SSA labor-force and 30 percent of





5


GDP – is particularly important because of the predominance of rain-fed agriculture and thus
an increased dependence on rainfall patterns.


Second, among the possible factors affecting the link between climate and biological shocks
and GDP fluctuations, we take into account the impact of remittances on macroeconomic
stability and their interactions with shocks and rainfalls. Remittances represent one of the
main and most stable sources of financial inflows to West Africa. Between 2001 and 2008,
official remittance inflows increased by 700% in Mali, Mauritania, Niger, Togo, Benin and
Senegal. These countries received (on average) US$ 1.25 billion of remittances in 2008. Since
1990 remittances in West African countries have been a more important financial flow than
Official Development Assistance (ODA). Third, we use a Panel Vector Autoregressive (PVAR)
model which combines the traditional VAR model with a panel-data approach based on the
PVAR routine written by I. Love (World Bank). To assess the role of the level of GDP in the
spillover effects exerted by climatic shocks and precipitation variability, we consider two
PVAR models on the six selected countries: the first PVAR model (PVAR1) assesses the
contemporaneous and lagged impacts of both precipitation shortage and climate shocks on
remittances per capita and GDP per capita while the second PVAR model (PVAR2) estimates
the contemporaneous and lagged effects of GDP per capita in region of immigration on
remittances inflows in the six West African countries. By comparing the PVAR1 and PVAR2
outputs from orthogonalized Impulse Response Functions (IRF) and Forecast Error
Decomposition Variance (FEDV) we are able to:


i) Isolate the effect of precipitation variability on remittances inflows and GDP per capita
and compare it to the impact of climate and biological shocks.


ii) To discuss the constraints on remittances inflows in order to smooth GDP fluctuations
by taking account of exterior economic constraints.


Results from orthogonalized IRFs bring strong evidences that climate shocks and
precipitation variability are strong explanatory factors amongst in GDP fluctuations in West
African countries. Remittances responses to shocks on precipitation variability and climate
shocks are significant and counter cycle whereas responses to GDP shocks remain non-
significant. This suggests that remittances contributes directly to a high resilience to
precipitation variability and climate shocks and then have indirect effects on GDP fluctuations.
Results from FEDV also indicate that precipitation explains a large fraction of the GDP and





6


remittances fluctuations compared to climate shocks and home GDP shocks. Whereas climate
shocks explain almost 2.5% of the GDP variance, precipitation variance contributes to 1.6% in
GDP fluctuations. Precipitation variability explains 1% of the remittances variances and 3.4%
and 1.4% is explained by climate and GDP shocks respectively. Results from IRFs and FEDV
indicate GDP shocks explain a low fraction of the remittances variance whereas GDP of
countries of immigration explain the larger fraction of the remittances inflows in countries of
emigration. A negative shock corresponding to -57%% to GDP of Western Europe correspond
to a variation of 0.94% in remittances inflows and results from FEDV indicates that Western
European GDP variance contributes to 6.1% of remittances variance.

The remainder of the paper is organized as follows: section 2 outlines the empirical and
theoretical literature dealing with the impact of remittances on GDP volatility in West African
countries. Section 3 presents the methodology and empirical results of the paper. We draw
the main conclusions from the study in section 4.





7




2. The role of remittances in smoothing GDP fluctuations


in West Africa



In West Africa, remittances are an integral part of household risk management strategies.
According to Findley (1994), during the great drought (1984-1985), 63% of households
depended on remittances sent by (non-resident) family members abroad. Of those, 47% were
receiving money from migrants established in France and only 16% from migrants
established in sub-Saharan Africa cities. These findings show first, the importance of
international migration as a source of remittances collected by households in the region of
origin and secondly, the importance of migration as a means of providing insurance against
climatic hazards. Several case studies have shown a positive relationship between the
occurrence of natural disasters and growth in remittances. In a study conducted in Jamaica,
Clarke and Wallsten (2004) show that remittances tend to increase with the occurrence of a
natural disaster. Similarly, Gupta (2004) shows the positive impact of droughts in India on the
cyclical component of such transfers. Ratha (2006) finds that transfers tend to increase after
the occurrence of a natural disaster in Bangladesh, the Dominican Republic, Haiti and the
Honduras. However, no study addresses the importance of the characteristics of the seasonal
distribution of precipitation at a particular location (i.e. the precipitation regime) in
explaining the volume of remittances in West Africa.

In theory, an increase in the variability of precipitation should negatively affect the income of
households predominantly occupied in the agricultural sector and therefore lead to an
increase of remittance inflows. However, several studies have questioned this unidirectional
relationship. First, as migration is an ex-ante and ex-post risk management process, an
anticipated shock (such as drought) may give rise to compensation well before the occurrence
of the shock. Some other shocks, which are less predictable, may instead, give rise to
compensation after their occurrence (e.g. floods). Second, remittances may be sent because of
seasonal variability (intra-annual) intrinsic to rainfall patterns in West Africa. Remittances





8


also constitute a form of insurance against all types of shocks, whether climatic or economic
(Ebeke, 2010).

Although it is now widely recognized that households benefit from remittance inflows (Gupta
et al., 2007), their empirical macroeconomic consequences remain less well known. Despite a
broad set of theoretical studies (see Rapoport and Docquier, 2006), “empirical studies still lag
behind and have mostly focused on growth, inequality and poverty, leaving issues of


macroeconomic stability largely uninvestigated” (Bugamelli and Paterno, 2008). Nevertheless,
according to the IMF (2005), “[…] the relatively stable and a-cyclical nature of remittances
suggests that countries with access to significant remittance inflows may be less prone to


damaging fluctuations, whether in output, consumption or investment”.

Recent empirical evidence has sought to demonstrate this hypothesis but the results are often
ambiguous regarding the pro or counter-cyclical patterns of remittances to GDP fluctuations.
Ebeke (2010) measures the role of insurance played by remittances against shocks using a
measure of the cyclical nature of the transfers vis-à-vis real GDP. Ebeke (2010) found that half
the sample examined was defined by a strong counter-cyclical relationship (particularly
during the mid-1990s). Similarly, Gupta et al. (2009) test the cyclical nature of remittances on
real GDP of sub-Saharan African countries identifying a counter-cyclical relationship during
the period 1996-2006 and a pro-cyclical association during the previous period (1980-1995).
These studies offer contrasting results according to the particular country or region surveyed.
For example, Neagu and Schiff (2009) find a pro-cyclical relationship for 70% of their sample
including 116 countries. However, Sayan (2006) estimated correlations between the cyclical
components of remittances and GDP for 12 countries suggest that there is a counter-cyclical
relationship between the two variables for the sample as a whole. However, Sayan country-
by-country analysis also suggests a greater heterogeneity of cases as only India and
Bangladesh had counter-cycles whilst the rest of the sample revealed a pro-cyclical
relationship (especially in Senegal and the Côte d’Ivoire).

In theory, the smoothing role of remittances is justified by the assumption that individuals are
guided by altruism. Under this assumption, remittances should be counter-cyclical because
the migrants tend to remit more funds when the economy in the country of origin undergoes a
shock. A portfolio approach would consider the transfers a form of investment which should





9


be pro-cyclical. According to Elbadwi (1992) and Agarwal and Horowitz (2002), most
empirical evidence suggests that remittance flows are based on altruistic motivations.
Nevertheless, studies testing for a correlation between cyclical components of the transfers
and GDP suffer from several drawbacks. First, descriptive statistics (e.g. correlation
coefficients) cannot take into account multiple causality which requires sensitivity analyses
based on econometric models (e.g. Forbes and Rigobon, 2002). Second, the pro-cyclical nature
of transfers in relation to real GDP might have a more positive impact on the level of GDP
given a country’s relative dependence on remittances. In order to overcome these
shortcomings, we use a Panel VAR model in order to analyse the dynamic patterns of shocks’
impacts on GDP fluctuations and the dynamic responses of official remittance inflows to both
GDP instability and the occurrence of shocks. The panel VAR model also enables an empirical
exploration of the underlying incentives to remit not only as a consequence of significant
fluctuations in GDP but also as a direct consequence of shocks, whether climatic or economic
in West African countries.

The vector autoregressive model differs from the classical model because it exploits all the
causal links between the components of a phenomenon, and within a time space (Meuriot,
2008). There is no distinction between endogenous and exogenous variables and each
variable is expressed in terms of its own past values and past and current values of all other
variables. In addition, the VAR model allows us to study the consequences of economic
variables through the reaction functions of shock according to Granger causality tests
(Grenne, 2005). It identifies different types of shocks in the analysis of transmission
mechanisms of economic variables, through the study of the propagation of shocks or
impulses. Also, since the contribution of Sims (1980), researchers believe VAR models as the
most suitable methodological framework for analysis of fluctuations in terms of innovations
(Ziky, 2005). Recent studies use VAR model to determine the impact of remittances on mid
and long term growth (A. Tchokpon Medenou, 2010) and on the main determinants of
remittances (Coulibaly, 2009)











10



3. GDP fluctuations, climate variability and remittances in


West Africa: an investigation using a PVAR model

In this section, we assess the share of remittances received by six West African countries
experiencing climatic shocks using a simple Panel VAR model. The model will also be used to
explore the underlying incentives to remit not only as a consequence of significant
fluctuations in GDP but also as a direct result of shocks and precipitation variability in West
African countries and economic constraint in host countries.

3.1. Empirical methodology and data


We present an annual dataset comprising six West African countries (Mali, Niger, Mauritania,
Togo, Benin, Senegal), and covering the period 1980 to 2009. In order to analyze the impact of
climate shocks on GDP fluctuations and the potential role of remittances, we use a panel data
Vector Autoregression (VAR) methodology based on I. Love and L. Zicchino (2006). This
combines the traditional VAR model, which treats all the variables in the system as
endogenous, with a panel-data approach, which allows for unobserved individual
heterogeneity.

We specify a first-order four -variable VAR model as follows:

(1) ( ) fiuYLY iitit +++Γ= ε

where


it
Y is a vector of stationary variables, ( )LΓ is a matrix polynomial in the lag operator,


i
u is a vector of country specific effects and


it
ε is a vector of idiosyncratic errors.


it
Y is the


four variables vectors: {PREC, CLIMAFF, REMIT and GDP} where GDP represents real GDP
(purchasing power parity PPP adjusted) per capita2, REMIT reflects remittances and the
compensation of employees (earnings) per capita3, CLIMAFF is a variable reporting the




2 Real GDP per capita (PPP adjusted) come from Penn World tables. Real GDP per capita (GDPPC) is measured as
the log of the Real GDP divided by the total population in each country.
3 Remittances and compensation of employee received per capita (REMITPC) obtained from WDI, World Bank
(2010). Worker remittances and the compensation of employees is comprised of current transfers by migrant





11


number of affected people climate shocks experienced in each country between 1983 and
2009. We then specify another four variables PVAR model where


itY is the four variables
vectors: {GDPeur, GDPCI, REMIT and GDP}. GDPCI is the GDP (purchasing power parity PPP
adjusted) per capita of the Côte d’Ivoire and GDPeur is the GDP (purchasing power parity PPP
adjusted) per capita of Western Europe. These two variables aims at reflecting shocks on GDP
per capita in the main countries capturing the bulk of West African migration.

We use two proxies for climatic factors from the CRED Emergency Disasters Database4 (EM-
DAT) data: (i) the number people affected5 by droughts episodes experienced by countries;
and (ii) the number of people affected by extreme temperature. We choose to not include
floods event to approximate climate shocks since floods are well known to be low span events
and because of their ambiguous effects on local economies6. A number of studies (Raddatz,
2007; Raddatz, 2009; Skidmore and Toya, 2002) have assessed the impact of climatic and
natural disasters’ on macroeconomic stability using the EM-DAT data. The EM-DAT database
includes data on the occurrence and effects of over 12,800 mass- disasters in the world since
1900, compiled from a wide range of sources. In EM-DAT, disasters are divided into two main
categories (technological and natural corresponding to human and natural determinants of
disasters). The natural disaster category is divided into 5 sub-groups, which in turn covers 12
disaster types and more than 30 sub-types. The climatic disaster categories include floods,
droughts, extreme temperatures, and wind storms. The occurrence of disasters is proxy by a
dummy with a value equal to 1 in the case of a shock and 0 otherwise. In our article, we
choose to approximate shocks by the number of people affected rather than using a dummy
variable. We justify this choice by the following assumption: the more a shock is covariant and
the individuals have incentives to remit. Thus, a variable including the number of affected
people by a shock is more fitted to translate the dynamic properties of the model. However,
the production of this type of data does have some limitations (UNDP, 2004). First, the data
does not take into account the possibility that the occurrence of a particular shock can derive




workers, plus wages and salaries earned by non-resident workers. Per capita remittances are obtained by
dividing the total amount of remittance inflows by the total population of each country.
4 Center for Research on the Epidemiology of Disasters Database, CRED (see http://www.emdat.be/).
5 The number of affected people are approximate as the sum of People suffering from physical injuries, trauma,
people needing immediate assistance for shelter and people requiring immediate assistance during a period of
emergency; it can also include displaced or evacuated people as a direct result of the disaster.
6 We performed multiples PVAR incorporing floods as a proxy of climate events. Results from IRFs function and
FEDV showed no statistical differences.





12


from another shock7. A further limitation of the EM-DAT data concerns drought, which is a
relatively slow onset of some natural hazards. For example, a drought may develop gradually
over time and space and therefore its occurrence and economic impact cannot be simply
translated into a single dummy variable. These limitations are inherent to a conception of
disaster which places little emphasis on the temporal dimension. We propose to focus on this
temporal dimension using precipitation (PREC) data derived directly from multiple weather
stations8 throughout sampled countries. Therefore, for each country in our sample, we have
estimated the average of annual precipitations from a database of at least ten (uniformly
distributed) meteorological stations per country.

The presence of fixed effects which are correlated with the regressors, due to lags of the
dependent variable is a concern we address in estimating this model. In order to allow a fixed
effect without biased coefficients, we follow Love and Zicchino (2006) procedure by Helmert
transformed all variables. The Helmert procedure or forward mean-difference procedure
allow us to estimate the coefficient by system GMM preserving the orthogonality between
Helmert transformed variables and regressors (Arellano and Bover, 1995; Love and Zicchino,
2006). We focus our analysis on the orthogonalized impulse-response functions (IRFs) and
Forecast Error Decomposition Variance (FEVD). IRFs describe the reaction of one variable in
the system to the innovations in another variable in the system, while holding all other shocks
at zero.

Standard errors of IRFs and confidence intervals (5th and 95th percentiles) are generated with
Monte Carlo simulations with 90 repetitions. However, the actual variance-covariance matrix
of the errors is unlikely to be diagonal; it is therefore necessary to decompose the residuals in
such a way that they become orthogonal. The usual convention used is the Choleski
decomposition which assumes that the variables which come earlier in the ordering affect the
subsequent variables contemporaneously, and with a lag; whilst the variables that come later




7 For example, the EM DAT database has identified two main droughts in Mali during 1978 and 1980 (which is
regarded as the period of the great drought). However, many secondary sources indicate that Mali experienced
the most dramatic consequences of drought in 1973 and 1984 because of famine. Moreover, the rainfall deficits
between 1970 and 1985 negatively impacted local ecosystems resulting in a significant decline in cereal yields.
The human consequences of this were serious as Mali experienced two major famines during this period.
8 NOAA NCEP CPC EVE: Monthly station precipitation and temperature data from the Climate Prediction Center
Resolution: 13701 stations; Longitude: global; Latitude: global; Time: [Sep 1982, Sep 2010]; monthly data.





13


only affect the previous variables with a lag9. For the purposes of this paper, we also order the
variables according to the results of our Granger causality / bloc exogeneity Wald Tests. In
each model, the Wald statistic for joint significance of all other lagged endogenous variables in
the equation indicated under the non causality hypothesis has a probability < 5% for both
remittances per capita and the log Real GDP per capita. These two variables are treated as
endogenous to our model. We assume that disaster measures and precipitation are exogenous
to the system since climate shocks and precipitation are not subject to any reverse causality
from GDP and remittances10. The resulting Cholesky order for our variable ordering is as
follows: PREC, CLIMAFF, REMITPC and GDPPC for the first PVAR model and GDPeur, GDPCI,
REMIT and GDPPC for the second PVAR model.

3.2. Empirical results


In this section we estimate systems of equations using the GMM estimator to assess the
dynamic response of remittances inflows and GDP to both climatic disasters and precipitation.
As mentioned in the previous section, using a panel-modeling framework allow us to increase
the power of the analysis given the data limitations. We, firstly concentrate our analysis on the
response of GDP per capita to climatic disasters and precipitation variability. Secondly, we
focus on the response of remittances inflows response to climate shocks and precipitation. In
a first time we discuss the dynamic relationship between GDP per capita and remittances
inflows taking into account the GDP fluctuations in regions of immigration.

Before estimating the structural VAR model, we tested for the stationarity of variables. To do
this, we used several panel unit root tests: Levin, Lin and Chu (2002), Im, Pesaran and Shin
(2003), Fisher-type tests using augmented dickey-fuller (ADF) and Phillips–Perron (PP) tests
(Maddala and Wu (1999) and Choi (2001), and Hadri (1999). The tests results (see Table A.1
in Appendix A) show that all the variables are non-stationary in levels but stationary in first-
differences for all countries. We set all endogenous variables in the model, fixed the sample
period (1980 to 2009), and conducted diagnostic tests in order to verify the stability of the
PVAR. The inverse roots of the characteristic autoregressive (AR) polynomial tables




9 Variables that appear earlier in the system are exogenous and those which appear later are largely endogenous.
10 For further details see the Granger causality / bloc exogeneity Wald Tests tables (tables XXXX) in Appendix.





14


(Lütkepohl (1991) is shown in table A2 (see Appendix A). The test has (p = 1, 2) as proposed
in the Akaike Information Criterion (AIC) and Schwarz Information Criterion (SC)11.


3.2.1. GDP fluctuations and climate patterns in West African countries

IRFs are estimated on the basis of PVAR 1 model (see methodological description in section
3.1) of two lags because of the limited time dimension and the small number of countries in
the sample. Figure B.1 and Table B.1 show results obtained from impulse response functions
for GDP per capita to one standard deviation shocks to climatic and precipitation variables.
Results obtained from IRFs allow us to identify the dynamics impacts of a negative shock to
climate variable (resulting in a positive shock i.e. fewer people affected by a shock) and on
precipitation (resulting in a decline in precipitation) on GDP. Our results strongly indicates
that a positive shock corresponding to a decrease by 90% (one standard deviation innovation
shock) in the number of people affected by a climate disaster results in a statistical significant
increase in the GDP per capita immediately after the shock occurrence (a variation of 31.1
between year 0 and year 1). A one standard deviation shock on precipitation (by 73%) results
in a significant decline of GDP per capita and a negative variation of -2.91. between year 0 and
1. Results from IRFs also indicate that the effect of a climatic disaster as well as a precipitation
shock dies after 3 years.

The Forecast Error Variance Decomposition analysis presented in Table B.3 Shows that
climatic disaster shocks are responsible for 2.4% percent of variance of GDP at the 10 and 20
year horizon and precipitation variable account for 1.62% of the variance of GDP. As
mentioned by A.C. David (2010), climate shocks can be considered as weakly exogenous
variable. Since they are approximate by their impacts on individuals, they take into account
the capacity (or non capacity) of individuals to cope with those types of events. In terms of
vulnerability conceptual framework, the measure of climate shocks take into account the
socioeconomic capacity of individuals and cannot translate their real exposure to physical
events. On the contrary, precipitation variable can be treated as a purely exogenous variable
(see Granger causality test in Table A.2 in appendix A). These findings then suggest that the




11 See Burnham, K. P., and Anderson, D.R., 2002. Model Selection and Multimodel Inference: A Practical
Information-Theoretic Approach, 2nd ed. Springer-Verlag. ISBN 0-387-95364-7.





15


selected West African countries are particularly exposed to physical factors such as
precipitation variability.

According to Raddatz (2009), external shocks account only for a small fraction of the overall
variance of real GDP in low-income countries. Even in the long run, they cannot explain more
than 11 percent of this variance. The remaining 89 percent is accounted for by factors
associated with endogenous shocks. Among the external shocks, climatic disasters explain
14% of the GDP variance. Without taking account of the remittances effect on such economies,
our results are in line with those find by Raddatz (2009). Both climate disasters and
precipitation variables account for 4% of the whole GDP variance. When taking account of the
remittances effect, all variables included in the model explains 5.5% of the GDP variance.
Thus, 94.5% of the GDP variance is explained by other factors (external i.e. such as commodity
price, interest rate, or internal factors i.e. civil wars, corruption…) non-included in the model.

In the next section we give more details about the linkages between remittances inflows,
climate disasters and precipitation. We show that remittances inflows have considerable
indirect effects on GDP fluctuations acting as an external “stabilization” source of income.

3.2.2. Remittances and climate nexus


Emigrants may send money to home country in order to sustain consumption of their families.
These remittances may be independent of economic conditions of home country or they may
be countercyclical. At the microeconomic scales, research on the motives to migrate and remit
impulse in the 80’s by the New Economics of labor Migration (NELM) make distinction
between remittances being primarily driven by altruistic motives versus remittances being
driven by investment motives. According to Docquier and Rapoport (2006), «it is not only that
different individuals may be heterogeneous in their motivations to remit, but also that
different motivations to remit may coexist within the same individual." One of the positive
effects of remittances by unanimity is the impact on poverty. This effect is through increasing
the resources of households and by smoothing their consumption expenditure over time
(Gupta, Pattillo and Wagh, 2007). The impact of remittances on household living conditions is
reinforced by their counter-cyclical effect on the economy: increasing during periods of
recession and decreasing during expansions. Remittances help households’ recipients to
maintain their level of well-being and a better allocation in their consumption spending over





16


time, especially for those engaged in seasonal activities, such as farmers (Daffé, 2009).
However, as mentioned in Section 2, empirical assessments do not agree on the counter
cyclical nature of remittances.

Macroeconomic research assumes that pro cycle and counter cycle patterns between
remittances and GDP is evidence of either investment or altruistic motives, respectively.
Empirical research dealing with the compensatory or opportunistic nature of remittances
finds different results depending on time scale and selected region. Chami, Fullenkamp and
Jahjah (2005) have found some of the strongest evidence to date that remittances are better
described as compensatory transfers than as opportunistic ones. Work by Frankel (2009) also
finds them counter-cyclical. Neagu and Schiff (2009) find that remittances are pro-cyclical.
Durdu and Sayan (2008) find them counter-cyclical in Mexico but pro-cyclical in Turkey.

Our results suggest remittances have countercyclical patterns with climate shocks and
precipitation variable in WA countries whereas the direct link between remittances and home
GDP seems more multifaceted. Results from IRFs functions in PVAR 1 show that remittances
increase when climate and precipitation shocks occur in West African countries. A positive
shock corresponding to a decrease by 90% (one standard deviation innovation shock) in the
number of people affected by a climate disaster results in a statistical significant decrease in
remittances per capita inflows the year after the shock (a variation of -11.24 between year 1
and year 2). A one standard deviation shock on precipitation (by 73%) results in a significant
increase in remittances per capita and a positive variation of 10.16 between year 0 and 1. We
can notice a difference in terms of time variation between remittances sent after a climate
shock and remittances sent in case of precipitation shock. In the first case remittances tend to
increase one year after the shock occurred whereas in the second case remittances increase in
the same time.

Remittances variance is explained by 5.85% by all the variables included in the model and by
4.45% regarding both climate shocks and precipitation. Precipitation variable account for a
smaller share in the variance of remittances (1%) than climate shocks and 94% of the
remittances variance may be explained by other variables omitted in the PVAR model. These
results may be explained by the fact that:





17


i) Precipitation variability has multiple effects on local economies and a decline in
precipitation cannot be translated into a shock.


ii) In case of altruism or counter cyclical patterns, remittances are sent in case of different
type of shocks. In our article dealing mainly with precipitation shortage and drought, we
cannot take into account the effects of large natural disasters, geological disasters or
internal shocks on remittances inflows. This could explain the high fraction of the
remittances variance unexplained by the model.


iii) Incentives to remit are not only driven by altruism, in case of remittances driven by
investment, the share of remittances sent after a shock may decrease.


iv) Remittances inflows strongly depend on foreign economies. In case of economic or
political crisis in host countries, remittances inflows may decrease as was the case in
West African countries right after the recent economic crisis.



3.2.3. Is the cyclical nature of remittances may be questioned by the economy of host
countries?



In this section we focus on the relationship between GDP and remittances. As mentioned in
the previous section, precipitation and climate shocks have a significant impact on GDP
fluctuations. However, the response of remittances to GDP shocks seems less direct and more
complex to analyze. PVAR 2 model aims at assessing the exterior economic constraints
weighing on remittances inflows in sampled countries. Results from IRFs shows that the
response of remittances to one standard deviation innovation on GDP per capita remains non
significant over the time even if standard error bands (obtained by Monte Carlo simulation
with 90 reps) are closed to the zero line at period 3 and 4 (s=3; 4). This suggest that
remittances tend to increase after a shock on GDP if all other variables are maintain to zero 3
years after the initial shock.

The question we ask then is: if remittances increase in case of climate and precipitation shock
which have a significant impact on GDP, we would be able to expect a significant response of
remittances (upward) in case of negative shock to GDP. This latter relationship remains
however unverified according to results obtained from IRFs in PVAR 2.

The lack of significant response of remittances to shock on home GDP can be explained
theoretically by an arm's length of remittances to economic fluctuations (as policies) in





18


countries (or regions) of immigration (search authors on the theory and empirical). This
dependency may result in a disability situation of migrants living abroad to send money to
their families remaining in countries of origin when they undergoes a shock. The theoretical
literature identifies several potentially negative effects of remittances limiting their
effectiveness. According to Azam and Gubert (2002) and Ndione and Lalou (2005), the
presence of remittances may lead to a liquidity trap. Because of anticipations of recipients,
flows of remittances are used to maintain the dissipation of resources. The relationship
Poverty - Migration - Remittances – may be enhanced when remittances are no longer used
only to smooth consumption expenditure, but when they participate to transform and
diversify consumption needs.

Thus, a dependency to remittances can be favored in developing countries (Lipton 1980,
Binford, 2003; Chami and al. 2003), encouraging the emigration of individuals of working age,
so a massive outflow of labor, leading eventually to reduction in labor supply. In this case,
risks of dependence on external economic conditions (economic cycles of host countries) and
transmission of economic shocks may increase. According to the IMF (2009), the countries
receiving remittances may see their economic condition deteriorated depending on the
magnitude of the crisis in countries sending remittances. The probability of deteriorating
economic conditions in these countries then depends on the likelihood of such countries to be
affected by a crisis (the degree of exposure to economic crisis).

Our results strongly confirm the assumption that remittances inflows heavily depends on GDP
fluctuations of sending countries. Results from IRFs in PVAR 2 model indicate that the
economic conditions of the main host countries have a direct impact on the volume of
remittances inflows in countries of origin. The response of the remittances variable
significantly follows the simulated shock (one standard deviation innovation) on GDP of
European countries with a lag of one year. Results from Error Forecast Variance
Decomposition in PVAR 2 show that European GDP variance and GDP of Côte d’Ivoire
contributes, respectively, to 6.1% and 7.2% of remittances inflows variance whereas Home
GDP variance only contributes to a small fraction of the remittances inflows fluctuations
(1.7%).





19


4. Conclusions

By comparing results from FEDV (with 10 reps) in PVAR 1 and PVAR 2 models, we assess the
contribution of external economic shocks variance (approximated by the GDP of main
countries of immigration) vis-à-vis climatic and precipitation shocks on home GDP and
remittances inflows fluctuations. Results from FEDV are summarized in the following table.


Table 1 Results from FEVD: climate and precipitation Vs. Countries of immigration GDP fluctuations


Prec Climaff Prec and
ClimShock


GDPeur GDPCI
GDPeur and


GDPCI
Total


GDP 1,60% 2,40% 4,00% 5% 3,50% 8,50% 12,50%


Remit 1% 3,40% 4,40% 6,10% 7,20% 13,30% 17,70%

Both GDPeur and GDPCI variables contributes to a high fraction of the home GDP variance
(8.50%) and remittances inflows variance (13.50%) if compared to the climate component
(climate shocks and precipitation variable). Climate shocks and precipitation account for 4%
of GDP variance and 4.40% of remittances variance. If the fraction of the variance of European
GDP countries contributes to a higher volatility than GDP of Côte d’Ivoire in GDP of West
Africa countries (respectively 5% and 3.50%), fluctuations of GDP in Côte d’Ivoire explain a
larger fraction of remittances inflows to West Africa countries. 7.20% of the remittances
variance is explained by the variable GDPCI after 10 lags whereas 6.10% is explained by GDP
fluctuations in European countries. Results from Error Forecast Variance Decomposition in
PVAR 1 show that 4% of GDP variance and 4.40% of remittances are explains by precipitation
and climate shocks. The fraction of the GDP and remittances variances explained by
precipitation only remains lower with respectively 1.60% and 1%.





20


References


Azam, Jean-Paul & Gubert, Flore. 2005. "Migrant Remittances and Economic Development in Africa: A
Review of Evidence," IDEI Working Papers 354, Institut d'Économie Industrielle (IDEI), Toulouse.


Baltagi, B.H. (2005) “Econometric Analysis of Panel Data” Third Edition, John Wiley & Sons, Chichester.
Binford L. 2003. “Migrant remittances and (under) development in Mexico”. Critique of Anthropology


23:305–336.
Breitung, J. (2000) “The local power of some unit root tests for panel data” Advances in Econometrics


B. H. Baltagi (Ed.), Volume 15: Nonstationary Panels, Panel Cointegration, and Dynamic Panels,
Pages 161-178. Amsterdam: JAI Press.


Cavallo, E., Galiani, S., Noy, I. and J. Pantano (2010). “Catastrophic Natural Disasters and Economic
Growth”, Mimeos, Inter-American Development Bank: Washington, D.C.


Chami, R. et al. (2008) “Macroeconomic Consequences of Remittances” IMF Occasional Paper 259, IMF,
Washington, DC.


Chami, R., C. Fullenkamp, et S. Jahjah. 2003. “Are Immigrant Remittance Flows a Source of Capital for
Development?”, International Monetary Fund, Working Paper 03/189


Chami, R., Hakura, D. & Montiel, P. (2009) “Remittances: An Automatic Output Stabilizer?” IMF
Working Paper WP/09/91, IMF, Washington, DC.


Christian EBEKE, 2010. "Transferts des migrants, ouverture sur l'extérieur et dépenses publiques dans
les pays en développement," Working Papers 201011, CERDI.


Coulibaly Dramane. 2009. “Macroeconomic Determinants of Migrants’ Remittances : New Evidence
from a panel VAR, Document de Travail du Centre d'Economie de la Sorbonne -2009.07.


Dacy, D.C., and H.C. Kunreuther. 1969. The Economics of Natural Disasters. New York, United States:
Free Press.


Dercon, S. 2004. “Growth and Shocks: Evidence from Rural Ethiopia.” Journal of Development
Economics 74(2): 309-329.


EM-DAT: The OFDA/CRED International Disaster Database - www.em-dat.net – Université Catholique
de Louvain - Brussels – Belgium.


Findley S. E. (1994). "Does drought increase migration ? A study of migration from rural
Fomby, T., Y. Ikeda and N. Loayza. 2009. “The Growth Aftermath of Natural Disasters.” World Bank


Policy Research Working Paper 5002. Washington, DC, United States: The WorldBank.
Giuliano, P. & Ruiz-Aranz, M. (2009) “Remittances, Financial Development and Growth” Journal of


Development Economics, Vol. 90, pp. 144-152.
Gupta, S., C. Pattillo et S. Wagh. 2007. “Impact of Remittances on Poverty and Financial Development in


Sub-Saharan Africa”, International Monetary Fund, Working Paper 07/38
Hochrainer, S. 2009. “Assessing the Macroeconomic Impacts of Natural Disasters – Are there Any?”


World Bank Policy Research Working Paper 4968. Washington, DC, United
Im, K. S., M. H. Pesaran, and Y. Shin (2003) “Testing for unit roots in heterogeneous panels” Journal of


Econometrics 115: 53-74.
Levin, A., C.-F. Lin, and C.-S. J. Chu (2002) “Unit root tests in panel data: Asymptotic and finite-sample


properties”. Journal of Econometrics 108: 1-24.
Lipton, M.. 1980. “Migration from Rural Areas of Poor Countries: The Impact of Rural Productivity and


Income Distribution,” World Development, Vol. 8. pp. 1–24.
Loayza, N., E. Olaberría, J. Rigolini, and L. Christiansen. 2009. “Natural Disasters and Growth-Going


Beyond the Averages.” World Bank Policy Research Working Paper 4980. Washington, DC, United
States: The World Bank.





21


Love, I. & Zicchino, L. (2006) “Financial Development and Dynamic Investment Behavior: Evidence for
a Panel VAR” The Quarterly Review of Economics and Finance, vol.46, pp.190-210.


Lutkepohl, H. (1991). Introduction to Multiple Time Series Anal-
Mali during the 1983-85 drought." International Migration Review 28:539-553.
Matteo Bugamelli & Francesco Paternò, 2008. "Output growth volatility and remittances," Temi di


discussione (Economic working papers) 673, Bank of Italy, Economic Research Department.
Mohapatra, S., Joseph, G. & Ratha, D. (2009) “Remittances and Natural Disasters: Expost Response and


Contribution to Ex-ante Preparedness” World Bank Policy Research
Neagu, I. & Schiff, M. (2009) “Remittance Stability, Cyclicality and Stabilizing Impact in Developing


Countries” World Bank Policy Research Working Paper 5077,
Noy, I. 2009. “The Macroeconomic Consequences of Disasters.” Journal of Development Economics


88(2): 221-231.
Noy, I. and A. Nualsri. 2007. “What do Exogenous Shocks tell us about Growth Theories?” University of


Hawaii Working Paper 07-28.
Raddatz, C. (2007). “Are external shocks responsible for the instability of output in low income


countries?” Journal of Development Economics, Vol. 84(1):155-187.
Raddatz, C. (2009) “The Wrath of God: Macroeconomic Costs of Natural Disasters” World Bank Policy


Research Working Paper, World Bank, Washington, DC.
Ratha, D. 2003. « Worker’s Remittances: An Important and Stable Source of External Development


Finance », Global Developing Finance 2003, World Bank, pp. 157-175.
Skidmore M, Toya H., 2007. Economic development and the impacts of natural disasters. Economic


Letters 94; 20-25.
Skidmore, M. and H. Toya. 2002. Do Natural Disasters Promote Long-run Growth? Economic Inquiry


40(4): 664-687.
Strobl, E. 2008. “The Economic Growth Impact of Hurricanes: Evidence from U.S. Coastal Counties.” IZA


Discussion Papers Series.
World Bank (2005) “Global Economic Prospects 2006: “Economic Implications of Remittances and


Migration” World Bank, Washington, DC.





22


Appendix A. Unit roots tests, estimation by GMM system and
diagnostics tests


Table A1: Unit root tests


Levin, Lin & Chu t* Im, Pesaran and Shin W-
stat


ADF Ficher Chi square Hadri Z-stat


Variables Statistique de test Prob.**
Statistique de


test
Prob.** Statistique de


test
Prob.** Statistique de


test
Prob.**


In levels
GDP 1.19209 0.8834 1.12420 0.8695 12.2909 0.2661 6.16208 0.0000
Prec -1.79306 0.0365 -4.12215 0.0000 36.4171 0.0001 8.00620 0.0000
GDPeur 7.73988 0.7056 9.55211 0.7642 0.10895 0.0908 8.26733 0.0000
GDPCI 6.46570 1.0000 5.92817 1.0000 0.67792 1.0000 6.47312 0.0000


1st difference
GDP -4.01785 0.0000 -6.48380 0.0000 57.9795 0.0000 3.46774 0.0003
Remit -1.76963 0.0384 -3.97804 0.0000 34.3585 0.0002 2.07497 0.0190
Prec -4.29408 0.0000 -7.98664 0.0000 72.6967 0.0000 -0.47231 0.6816
GDPeur -1.60460 0.0143 -3.59515 0.0002 33.8442 0.0007 6.25741 0.1203
GDPCI -3.19640 0.0007 -3.62071 0.0001 34.3823 0.0006 3.91757 0.0000
Climaff -0.88329 0.1885 -13.8539 0.0000 130.950 0.0000 3.93529 0.0000


2nd difference
Remit -3.87267 0.0001 -7.94819 0.0000 73.3555 0.0000 1.15061 0.1249
Climaff -16.1837 0.0000 -19.0482 0.0000 144.862 0.0000 0.23544 0.4069




Table A.2: VAR Granger Causality / Block Wald Exogeneity Test


Dependent variable: D(GDPCI) Dependent variable: D(GDPPC)


Excluded Chi-sq df Prob. Excluded Chi-sq df Prob.


GDPPC 1.320597 2 0.5167 GDPCI 0.118341 2 0.9425


PIBEUR 22.90564 2 0.0000 PIBEUR 3.137121 2 0.2083


REMITPC 0.244792 2 0.8848 REMITPC 0.277574 2 0.8704






All 26.93260 8 0.0007 All 3.716872 8 0.8817


Dependent variable: D(PIBEUR) Dependent variable: D(REMITPC)


Excluded Chi-sq df Prob. Excluded Chi-sq df Prob.


GDPCI 8.101298 2 0.0174 GDPCI 3.702014 2 0.1571


GDPPC 1.090284 2 0.5798 GDPPC 0.306556 2 0.8579


REMITPC 7.832738 2 0.0199 PIBEUR 11.55238 2 0.0031










All 19.70006 8 0.0115 All 19.15764 8 0.0140






23


Table A.3: Results of the Estimation by system GMM for PVAR 1***




Equation 1: dep.var h_prec


Equation 2: dep.var h_climaff


b_GMM se_GMM t_GMM


b_GMM se_GMM t_GMM
L.h_prec .17243427 .07831066 2.2019258 L.h_prec -454.78964 1039.3224 .30245155
L.h_climaff -9.147e-06 9.362e-06 -.97706655 L.h_climaff .08364657 .27656188 .30245155
L.h_remit .04918788 .10501001 .46841135 L.h_remit -920.55611 1377.766 -.66815128
L.h_gdp .07088622 .08087138 .87653032 L.h_gdp 157.96685 649.8533 .24308078
L2.h_prec -.14563746 .07772399 -1.8737774 L2.h_prec 1519.2628 1350.2165 1.1251994
L2.h_climaff 2.604e-06 .00001431 .18195575 L2.h_climaff -.01262607 .30133079 -.04190101
L2.h_remit .01457851 .09624597 .15147139 L2.h_remit -2506.3969 1464.5424 -1.7113857
L2.h_gdp -.10837987 .08611828 -1.2585002 L2.h_gdp 573.78514 792.24303 .72425394




Equation 4: dep.var h_gdp


Equation 3: dep.var h_remit


b_GMM se_GMM t_GMM b_GMM se_GMM t_GMM
L.h_prec -.13084385 .0769516 -1.7003396 L.h_prec .08786628 .07237914 1.2139723
L.h_climaff .00001346 6.768e-06 1.9887487 L.h_climaff -9.155e-07 8.390e-06 -.10910943
L.h_remit .07919874 .09360953 .84605427 L.h_remit .18592129 .08748801 2.125106
L.h_gdp .39245217 .08193632 4.789722 L.h_gdp -.03547135 .0783626 -.45265664
L2.h_prec .01269925 .0779736 .16286602 L2.h_prec -.00680926 .0754834 -.09020873
L2.h_climaff -1.013e-06 8.213e-06 -.123358 L2.h_climaff -.00001404 7.863e-06 -1.7856739
L2.h_remit -.00448971 .09632676 -.04660918 L2.h_remit -.14849739 .08818535 -1.6839235
L2.h_gdp -.02404409 .07534579 -.3191166 L2.h_gdp .11663632 .07916923 1.4732531


* Number of observations used: 144.
** estimation sample from 1983 to 2009, strongly balanced





24




Table A.4: Results of the Estimation by system GMM for PVAR 2***




Equation 1: dep.var h_GDPeur


Equation 2: dep.var h_GDPCI




b_GMM se_GMM t_GMM


b_GMM se_GMM t_GMM
L.h_prec 1.1272458 .06774685 16.639089 L.h_prec .00012375 .00034885 .35473799
L.h_climaff 9.3475813 6.8508478 1.3644415 L.h_climaff 1.3545711 .06911239 19.599539
L.h_remit


-204.08517 127.51405 -1.6004916 L.h_remit -.13052109 .7938822 -.16440864
L.h_gdp 44.050125 127.51405 -1.6004916 L.h_gdp -.58580031 .73599065 -.79593445
L2.h_prec


-.79849975 .06541692 -12.206319 L2.h_prec -.00055062 .00037558 -1.4660232
L2.h_climaff 2.9285851 8.0258221 .36489535 L2.h_climaff -.56234947 .07583894 -7.4150494
L2.h_remit


-143.89566 121.82946 -1.1811237 L2.h_remit .09522354 .77829639 .12234869
L2.h_gdp 141.60404 107.58767 1.3161736 L2.h_gdp -.68142273 .66337697 -1.0272029




Equation 4: dep.var h_gdp


Equation 3: dep.var h_remit


b_GMM se_GMM t_GMM b_GMM se_GMM t_GMM
L.h_prec .00002042 .00004897 .41707146 L.h_prec .00009362 .00004378 2.1381741
L.h_climaff .00880181 .00731318 1.3941333 L.h_climaff .00949287 .00680916 1.3941333
L.h_remit .05452637 .09920928 .54960957 L.h_remit .20317607 .08689615 2.3381482
L.h_gdp .34031169 .0842326 4.0401423 L.h_gdp -.06592155 .07678715 -.85849714
L2.h_prec


-.00007808 .00005695 -1.3709705 L2.h_prec -.0000244 .00004693 -.51985786
L2.h_climaff


-.00858085 .00803231 -1.0682924 L2.h_climaff -.01719763 .00729561 -2.357258
L2.h_remit


-.02566304 .0958797 -.26765876 L2.h_remit -.12473479 .08276789 -1.5070432
L2.h_gdp


-.01216905 .08005775 -.15200335 L2.h_gdp .12049802 .0766133 1.5728081


* Number of observation used: 144.
** estimation sample from 1983 to 2009, strongly balanced






25




Appendix B. Impulse responses and decomposition variances




Table B.1 : Results from Orthogonalized Impulse-Response Functions in PVAR 1**


Response of Precipitation to Precipitation Shock Response of Remittances to Precipitation Shock
Varname S Prec_5 Prec Prec_95 S Prec_5 Prec Prec_95
Prec 0 52,64 59,62 64,32 Remit 0 -8,2 -0,55 8,47
Prec 1 5,12 10,71 18,39 Remit 1 -1,19 5,04 11,4
Prec 2 -16,31 -7,24 2,28 Remit 2 -4,68 2,55 11,13
Prec 3 -6 ,7 -2,91 2,18 Remit 3 -6,65 -1,34 3,34
Prec 4 -2,1 1,12 5,52 Remit 4 -5,3 -2,42 0,47
Prec 5 -2,13 0,57 2,83 Remit 5 -2,43 -0,1 2,03
Prec 6 -2,31 -0,25 1,18 Remit 6 -0,91 0,72 1,85


Response of GDP to Precipitation Shock Response of Climate to Climate Shock
S Prec_5 Prec Prec_95 S Climaff_5 Climaff Climaff_95
GDP 0 -6,12 3,28 11,5 Climaff 0 6.1e+05 7.1e+05 7.8e+05
GDP 1 -17,12 -6,88 -0,08 Climaff 1 -2.6e+05 6.1e+04 4.0e+05
GDP 2 -13,65 -3,37 4,38 Climaff 2 -3.4e+05 4.7e+03 4.3e+05
GDP 3 -4,22 1,22 5,25 Climaff 3 -1.9e+05 7.7e+03 2.3e+05
GDP 4 -1,95 0,74 4,41 Climaff 4 -1.6e+04 3.4e+04 3.0e+05
GDP 5 -2,19 -0,34 2,04 Climaff 5 -1.9e+05 6.1e+03 2.2e+05
GDP 6 -1,49 -0,19 1,01 Climaff 6 -9.4e+04 -5.6e+03 2.2e+05


Response of Remittances to Climate Shock Response of GDP to Climate Shock
S Climaff_5 Climaff Climaff_95 S Climaff_5 Climaff Climaff_95
Remit 0 -9.2812 -1.3851 6.8959 GDP 0 -7.5779 -0.3129 7.0075
Remit 1 -9.9294 -0.8979 7.9957 GDP 1 0.4745 9.3463 18.7860
Remit 2 -20.7483 -10.9566 -0.2356 GDP 2 -5.9958 4.5709 14.9792





26


Remit 3 -10.5927 -1.7185 4.5859 GDP 3 -8.7110 0.5197 9.7765
Remit 4 -6.0633 1.7415 8.1597 GDP 4 -7.8650 0.1192 6.3685
Remit 5 -5.2548 0.4199 5.0910 GDP 5 -2.6999 0.7386 6.9107
Remit 6 -6.5232 -0.7005 0.4978 GDP 6 -3.1371 -0.4044 5.4388


Response of GDP to GDP Shock Response of Remittances to GDP
S GDP_5 GDP GDP_95 S GDP_5 GDP GDP_95
GDP 0 52.6724 59.6557 64.4800 Remit 0 0 0 0
GDP 1 14.2269 23.4120 32.0239 Remit 1 -9.9992 -2.1161 5.1046
GDP 2 -0.9894 7.1597 14.5995 Remit 2 -1.1074 6.0971 13.1519
GDP 3 -0.8127 3.8632 10.1147 Remit 3 0.2740 3.3530 8.7161
GDP 4 -0.7615 2.2856 6.9969 Remit 4 -2.2060 -0.4249 2.4901
GDP 5 -1.6425 0.5045 4.1423 Remit 5 -2.4195 -0.5611 1.7746
GDP 6 -1.3985 -0.1083 2.0223 Remit 6 -1.2711 0.4965 3.0265


Response of Remittances to Remittances Shock Response of GDP to Remittances Shock
S Remit_5 Remit Remit_95 S Remit_5 Remit Remit_95
Remit 0 51.6455 58.0071 62.4299 GDP 0 -4.5778 4.1158 14.4701
Remit 1 2.2263 10.6388 18.7594 GDP 1 0.1764 6.2093 14.4980
Remit 2 -12.9999 -6.0515 2.7514 GDP 2 -9.4372 1.7985 13.8833
Remit 3 -4.8942 -0.9711 5.0029 GDP 3 -10.0247 -2.3120 7.2354
Remit 4 -0.3165 3.2816 9.2398 GDP 4 -5.0521 -1.2360 4.3014
Remit 5 -3.1106 0.7920 4.5785 GDP 5 -1.9682 0.2563 3.9849
Remit 6 -3.4800 -0.7835 1.5787 GDP 6 -1.9413 0.1859 4.0787


* Impulse-Responses for 2 lag Panel Vector AutoRegressive
** Errors are 5% on each side generated by Monte Carlo with 90 reps






27




Figure B.1: Results from Orthogonalized Impulse-Responses Functions in PVAR 1
Impulse-responses for 2 lag VAR of prec climaff remit gdp


Errors are 5% on each side generated by Monte-Carlo with 90 reps


response of prec to prec shock
s


(p 5) prec prec
(p 95) prec


0 6
-16.3179


64.3233


response of prec to climaff shock
s


(p 5) climaff climaff
(p 95) climaff


0 6
-15.7994


20.7389


response of prec to remit shock
s


(p 5) remit remit
(p 95) remit


0 6
-9.4951


13.0591


response of prec to gdp shock
s


(p 5) gdp gdp
(p 95) gdp


0 6
-12.9034


12.2794


response of climaff to prec shock
s


(p 5) prec prec
(p 95) prec


0 6
-1.4e+05


1.9e+05


response of climaff to climaff shock
s


(p 5) climaff climaff
(p 95) climaff


0 6
-3.4e+05


7.8e+05


response of climaff to remit shock
s


(p 5) remit remit
(p 95) remit


0 6
-3.6e+05


1.4e+05


response of climaff to gdp shock
s


(p 5) gdp gdp
(p 95) gdp


0 6
-8.3e+04


1.1e+05


response of remit to prec shock
s


(p 5) prec prec
(p 95) prec


0 6
-8.2028


11.4010


response of remit to climaff shock
s


(p 5) climaff climaff
(p 95) climaff


0 6
-20.7483


8.1597


response of remit to remit shock
s


(p 5) remit remit
(p 95) remit


0 6
-12.9999


62.4299


response of remit to gdp shock
s


(p 5) gdp gdp
(p 95) gdp


0 6
-9.9992


13.1519


response of gdp to prec shock
s


(p 5) prec prec
(p 95) prec


0 6
-17.1455


11.5105


response of gdp to climaff shock
s


(p 5) climaff climaff
(p 95) climaff


0 6
-8.7110


18.7860


response of gdp to remit shock
s


(p 5) remit remit
(p 95) remit


0 6
-10.0247


14.4980


response of gdp to gdp shock
s


(p 5) gdp gdp
(p 95) gdp


0 6
-1.6425


64.4800






28




Table B.2 : Results from Orthogonalized Impulse-Response Functions in PVAR 2* **




Response of GDPeur to GDPeur Shock


Response of GDPCI to GDPeur
Varname S GDPeur_5 GDPeur GDPeur_95 S GDPeur_5 GDPeur GDPeur_95


GDPeur 0 6.5e+04 7.3e+04 7.9e+04 GDPCI 0 -1.4e+02 -66.5667 10.3324
GDPeur 1 7.0e+04 8.1e+04 9.3e+04 GDPCI 1 -2.0e+02 -86.1092 22.1605


GDPeur 2 1.7e+04 3.1e+04 3.1e+04 GDPCI 2 -2.4e+02 -1.2e+02 0.7574
GDPeur 3


-5.0e+04 -3.3e+04 -1.1e+04 GDPCI 3 -2.7e+02 -1.5e+02 -40.2377
GDPeur 4


-8.2e+04 -6.6e+04 -4.9e+04 GDPCI 4 -2.5e+02 -1.6e+02 -37.6650
GDPeur 5


-7.3e+04 -5.1e+04 -2.9e+04 GDPCI 5 -1.9e+02 -1.1e+02 23.7464
GDPeur 6


-3.4e+04 -5.6e+03 1.6e+04 GDPCI 6 -1.1e+02 -21.8549 98.8109




Response of Remit to GDPeur Shock


Response of GDP to GDPeur Shock
S GDPeur_5 GDPeur GDPeur_95 S GDPeur_5 GDPeur GDPeur_95


Remit 0
-2.1750 3.9183 12.1688 GDP 0 -0.9352 7.5555 15.5883


Remit 1 1.5491 6.4729 12.1819 GDP 1 -2.5515 3.6838 8.5651
Remit 2 7.6183 7.6183 12.1819 GDP 2 -8.2371 -2.7979 2.8850
Remit 3


-3.7721 2.7102 8.6184 GDP 3 -12.4989 -6.7105 -0.6246
Remit 4


-10.2231 -3.5398 2.0554 GDP 4 -11.3516 -5.7513 1.4622
Remit 5


-10.8078 -5.7231 -1.6182 GDP 5 -5.3088 -1.0043 5.0179
Remit 6


-8.2644 -2.8619 2.6655 GDP 6 -1.2624 4.0063 10.0077




Response of GDPCI to GDPCI Shock


Response of Remit to GDPCI Shock
S GDPCI_5 GDPCI GDPCI_95 S GDPCI_5 GDPCI GDPCI_95


GDPCI 0 476.0078 531.5793 582.6257 GDP 0 -19.9822 -13.1255 -5.4306
GDPCI 1 637.2252 718.0287 807.1901 GDP 1 -3.7244 1.9578 9.5571


GDPCI 2 536.9348 665.2090 785.1393 GDP 2 -6.1124 -3.9934 2.5850
GDPCI 3 327.4001 488.7159 662.8708 GDP 3 327.4001 488.7159 662.8708
GDPCI 4 90.0735 277.1693 481.1832 GDP 4 -10.9682 -5.3894 0.4783
GDPCI 5


-1.1e+02 90.4923 334.5506 GDP 5 -10.4010 -4.8226 0.1223





29


GDPCI 6
-2.3e+02 -39.0411 203.3766 GDP 6 -9.1678 -3.9197 -0.3306




Response of GDP to GDPCI Shock


Response of GDP to GDP
S GDPCI_5 GDPCI GDPCI_95 S GDP_5 GDP GDP_95


GDP 0
-2.7214 6.3952 11.7475 GDP 0 52.7954 59.6049 64.7789


GDP 1 0.3769 6.1395 13.4613 GDP 1 12.0844 20.2842 29.0821
GDP 2


-1.9852 4.3756 9.7976 GDP 2 -3.6153 5.7097 14.5436
GDP 3


-5.9456 0.8892 7.6981 GDP 3 -4.8836 1.5520 8.3468
GDP 4


-10.2782 -2.4368 4.6622 GDP 4 -3.7751 -0.4599 4.2301
GDP 5


-11.1986 -4.3254 2.3969 GDP 5 -4.1350 -1.2271 3.2901
GDP 6


-10.4083 -4.5515 1.5330 GDP 6 -3.5866 -0.7316 2.3815




Response of Remittances to GDP Shock


S GDP_5 GDP GDP_95


Remit 0 0.0000 0.0000 0.0000


Remit 1
-11.5978 -3.9292 1.8113


Remit 2
-1.8487 4.9611 11.7127


Remit 3
-0.1583 4.3584 10.1219


Remit 4
-0.3551 2.2909 6.1732


Remit 5
-1.4864 1.1112 4.3530


Remit 6
-1.8767 0.1639 3.4070


* Impulse-Responses for 2 lag Panel Vector AutoRegressive.
** Errors are 5% on each side generated by Monte Carlo with 90 reps.






30


Figure B.2: Results from Orthogonalized Impulse-Responses Functions in PVAR 1


Impulse-responses for 2 lag VAR of GDPeur GDPCI remit gdp


Errors are 5% on each side generated by Monte-Carlo with 90 reps


response of GDPeur to GDPeur shock
s


(p 5) GDPeur GDPeur
(p 95) GDPeur


0 6
-8.2e+04


9.3e+04


response of GDPeur to GDPCI shock
s


(p 5) GDPCI GDPCI
(p 95) GDPCI


0 6
-1.9e+04


3.9e+04


response of GDPeur to remit shock
s


(p 5) remit remit
(p 95) remit


0 6
-4.2e+04


3.6e+04


response of GDPeur to gdp shock
s


(p 5) gdp gdp
(p 95) gdp


0 6
-2.9e+04


2.8e+04


response of GDPCI to GDPeur shock
s


(p 5) GDPeur GDPeur
(p 95) GDPeur


0 6
-2.7e+02


98.8109


response of GDPCI to GDPCI shock
s


(p 5) GDPCI GDPCI
(p 95) GDPCI


0 6
-2.3e+02


807.1901


response of GDPCI to remit shock
s


(p 5) remit remit
(p 95) remit


0 6
-2.0e+02


192.7565


response of GDPCI to gdp shock
s


(p 5) gdp gdp
(p 95) gdp


0 6
-2.8e+02


37.2253


response of remit to GDPeur shock
s


(p 5) GDPeur GDPeur
(p 95) GDPeur


0 6
-10.8078


12.1819


response of remit to GDPCI shock
s


(p 5) GDPCI GDPCI
(p 95) GDPCI


0 6
-19.9822


9.5571


response of remit to remit shock
s


(p 5) remit remit
(p 95) remit


0 6
-13.0993


59.0281


response of remit to gdp shock
s


(p 5) gdp gdp
(p 95) gdp


0 6
-11.5978


11.7127


response of gdp to GDPeur shock
s


(p 5) GDPeur GDPeur
(p 95) GDPeur


0 6
-12.4989


15.5883


response of gdp to GDPCI shock
s


(p 5) GDPCI GDPCI
(p 95) GDPCI


0 6
-11.1986


13.4613


response of gdp to remit shock
s


(p 5) remit remit
(p 95) remit


0 6
-12.0547


14.0805


response of gdp to gdp shock
s


(p 5) gdp gdp
(p 95) gdp


0 6
-4.8836


64.7789





31


Table B.3: Results from Decomposition Variance Functions in
PVAR 1*




S Prec Climaff Remit GDP


Prec 10 .97125971 .01175851 .00441656 .01256523
Climaff 10 .01525844 .92750396 .05234303 .00489457
Remit 10 .01080319 .03452022 .9403333 .01434329
GDP 10 .01624611 .02470672 .01485959 .94418757
Prec 20 .97125913 .01175867 .00441692 .01256528
Climaff 20 .0152597 .92750076 .05234472 .00489483
Remit 20 .01080486 .0345213 .94032999 .01434385
GDP 20 .01624629 .02470697 .01485988 .94418686
* Decomposition Forecast Error Variance Functions are calculated with 10 and
20 reps






Table B.4: Results from Decomposition Variance Functions in
PVAR 2*


S GDPeur GDPCI Remit gdp


GDPeur 10 .85197443 .06140248 .05453555 .03208754
GDPCI 10 .05660771 .90347065 .00094864 .03897299
Remit 10 .06100112 .07244104 .84881688 .01774096
Gdp 10 .05094001 .03586547 .00687577 .90631874
GDPeur 20 .85008015 .05645895 .05725824 .03620266
GDPCI 20 .05906279 .89963176 .00131994 .03998551
Remit 20 .07253643 .07272446 .83596251 .0187766
Gdp 20 .06002879 .03597359 .00749534 .89650228


* Decomposition Forecast Error Variance Functions are calculated with 10 and
20 reps





Login