Are Bilateral Remittances Countercyclical? Implications for Dutch

Download Report

Transcript Are Bilateral Remittances Countercyclical? Implications for Dutch

Are Bilateral Remittances
Countercyclical?
Forthcoming, Open Economies Review
Jeffrey Frankel
Harpel Professor, Harvard Kennedy School
Immigration Seminar Series,
Harvard Migration & Integration Program,
Harvard Center for Population & Development Studies
9 Bow Street, 12-1:30 pm
September 23, 2010
Importance of remittances
• Total recorded workers’ remittances received by
developing countries increased 73% from 2001 to 2005,
– reaching a total of $167 billion.
• By 2008 they had further doubled, to $338 billion.
• Remittances have grown more rapidly than private
capital flows, or official development statistics.
• They now constitute more than 15% of GDP in:
Tajikistan, Tonga, Moldova, Kyrgyz, Lesotho, Samoa,
Lebanon, Guyana, Nepal, Honduras, Haiti, Jordan & El
Salvador.
• Until recently, macroeconomic aspects of remittances
were even more neglected than the economic study
2
Which countries are the biggest recipients?
Outlook for Remittance Flows 2010-11, D.Ratha, S.Mohapatra & A. Silwal,
Migration & Remittances Team, World Bank, April 23, 2010
3
To which countries are remittances most important?
Outlook for Remittance Flows 2010-11, D.Ratha, S.Mohapatra & A. Silwal,
Migration & Remittances Team, World Bank, April 23, 2010
4
Even in low-income countries overall, remittances
finance roughly half the current account deficit
Outlook for Remittance Flows 2010-11,
D.Ratha, S.Mohapatra & A. Silwal,
Migration & Remittances Team,
World Bank, April 23, 2010
5
What about the 2008-09 global recession?
• Estimated remittances to developing countries
fell 6 % in 2009,
– to $316 billion.
– = a much smaller decline than private capital flows.
• With improved global prospects,
they are forecast to increase 6.2 % in 2010
– and 7.1 % in 2011 -– World Bank, Migration & Remittances Team, April 23, 2010
6
Remittances fell in the recession, but held up
much better than private capital flows, even FDI.
7
Hypothesis: Remittances can play
the stabilizing role that capital flows
are in theory supposed to play
• In theory, capital flows should bring a
variety of benefits:
– smoothing short-term income disturbances,
– financing high-return investment opportunities
in low K/L countries
• and so substituting for labor flows to high K/L
countries or factor-based trade,
– and disciplining policies and institutions in the
recipient country.
8
• In practice, however, capital flows fail to
deliver on this promise:
– Rather, private capital flows are often procyclical:
• pouring in during boom times and disappearing in recessions.
– Rather than flowing on average from high K/L
countries to low, capital often “flows uphill.”
– Rather than rewarding countries that follow
sound economic policies, financial markets often
abet irresponsible budget deficits,
• including among autocratic and kleptocratic rulers.
9
3 cycles of net private capital flows
to emerging markets, by region
peaking in 1982, 1997 and
2008
Source: Capital Flows to Emerging Market Economies, IIF, 1/27/09.
10
Brief summary
of remittances literature
• (1) Theory
– Rapoport & Docquier (2005) review
the New Economics of Labor Migration.
– In theory, emigrants’ decisions to send
remittances should be based on intertemporal
optimization, usually with a household utility
function.
11
(2) Bilateral Data
– Ratha & Shaw (2005), in the absence of hard
bilateral data, allocate the totals across partners.
– Schiopu & Siegfried (2006) create bilateral data set
between some EU countries & neighbors.
– Jiménez-Martin, Jorgensen, & Labeaga (2007)
estimate bilateral workers’ remittance flows from
all 27 members of the EU, to recipient countries.
– Lueth & Ruiz-Arranz of IMF (2006, 2008) the largest
known bilateral data set to date.
– IDB has data on bilateral remittances from US to
countries, esp., in the Central American region.
12
(3) Evidence on cyclicality
– World Bank (2006): p.c. remittances respond significantly
to p.c. income in the home country.
– Clarke & Wallstein (2004) and Yang (2007):
remittance receipts rise in response to natural disaster.
– Kapur (2003): they rise in response to economic downturn.
– Lake (2006): remittances into Jamaica respond to the
difference between US and Jamaican income.
– Yang & Choi (2007):
they respond to rainfall-induced economic fluctuations.
– IMF finds less countercyclicality.
• Sayan (2006) : 12-developing-country study finds none.
• Lueth & Ruiz-Arranz (2006, 2008): similarly, procyclical.
13
(4) Why does the question of
remittance cyclicality matter?
• (i) It is especially important because governments
in remittance-receiving countries often reflexively
treat them as a source of foreign exchange to be
“harnessed” for national development,
– rather than letting recipients spend it on “unproductive”
uses such as imports of consumer goods.
– This thinking is common even among benevolent
governments, let alone kleptocracies.
14
Applicability, continued
• (ii) The Dutch Disease.
– On the one hand,
• Martin (1990): steady remittance flows can undermine incentives
for governments to create sound institutional frameworks,
– a sort of natural resource curse for remittances.
• Amuendo-Dorantes & Pozo (2004): a rise in remittances to LAC
countries leads to real appreciation, prime symptom of Dutch Disease.
• Acosta, Lartey & Mandelman (2009):
– For El Salvador, remittances again raise the relative price of nontraded goods,
• though ALM also find welfare-improving smoothing behavior .
– On the other hand, Rajan & Subramanian (2005) :
although the Dutch Disease analogy does extend to foreign aid,
it does not extend to remittances.
15
Applicability,
concluded
• (iii) Optimum Currency Area criterion
– The OCA question:
• When do benefits of a common currency outweigh costs?
– e.g., facilitating trade & other international transactions
– Vs. losing the freedom to run one’s own monetary policy.
– The textbook answer:
• A country that has cushions against any asymmetric shocks
(labor mobility, fiscal transfers, capital flows...), because it has less need
of a monetary policy different from that of the anchor country.
– My claim: remittances belong on the list if they are countercyclical.
• In a downturn, they can help substitute for monetary expansion & depreciation.
– Singer (2008): remittances are, and should be,
a determinant of the currency decision.
16
Not all senders are industrialized countries
• Roughly 10 per cent come from developing countries.
• South Africa, for example, receives many immigrants
from neighbors to work in its mines, farms, & factories,
and sends remittances back to the countries of origin.
• In many Gulf countries, immigrants (called ex-patriate
workers)
> than ½ of the private-sector labor
force.
– For example, outward remittances from Saudi Arabia
are about 7% of all remittances globally .
• (not included in the developing country statistic) .
17
The hypothesis is that remittances
respond not just inversely to income
in the receiving country, but also
positively to sending-country income.
• One would need to control for sender-country
income even if only the coefficient of recipientcountry income were of interest.
– It should be in the equation in theory.
– Omitting it in practice often produces the wrong sign.
• But cyclicality with respect to sender-country
income is also of interest in its own right:
18
South Africa and the Gulf are
two places where the Dutch Disease and
OCA motivations are particularly relevant.
• When mineral prices are low (e.g., 1990s), it is useful
to South Africa to have the “unilateral transfers”
deficit in the balance of payments automatically
moderate.
• When mineral prices are high (e.g. 2003-08), outward
remittances provide a brake on reserve inflows and
inflation – a particularly important point in these
two regions debating regional monetary unions.
19
My estimation of remittance cyclicality
i) Table 1: The Lueth & Ruiz-Arranz (IMF)
bilateral data set,
which includes 64 pairs of countries.
Cross-section covering 2005.
ii) Table 2: LRA bilateral data set,
Panel study, covering 1979 to 2005 .
iii) Table 3: Splicing of LRA data set with the
EU & Central American (IDB) data sets.
20
Country-pairs with high bilateral migration also,
of course, tend to show high bilateral remittances.
-10
-5
0
5
10
Scatter Remittances Migration
0
5
10
15
Migration
ln_remittances
Fitted values
Remittances between included country pairs are around US$113.6 billion.
Total of 540 observations: 266 for 2003 and 274 for 2004.
21
Remittances per lagged migrant
are positively correlated with cyclical differential
0
.02
.04
.06
.08
Remittances per Migrants and Difference in cyclical position
-.15
-.1
-.05
Difference in cyclical position
remitt_migrants
0
.05
Fitted values
Sources: Western Hemisphere data: FOMIN & the Central Banks, data from 2003-2004;
Jiménez-Martín, Jorgensen & Labeaga (2007). Data from 2003-2004;
Lueth & Ruiz-Arranz, IMF(2006); data from 2003-2004.
22
Table 1: Cross-section,
with LRA bilateral data set
• Cross-section includes 64 pairs of countries, 2005.
• Lagged stock of migrants (in 2000) has highly significant
effect on remittances, as in Freund & Spatafora (2005).
• We also control for sender-country income per cap.
• The variable of interest is the difference in cyclical position
between the sender country and the recipient country.
– In this table, cyclical position is computed as the (log) difference
between GDP in 2005 and the long run trend value of GDP.
– The estimated coefficient is positive and highly significant.
• The t-statistic is almost 4.
• Use of gravity IV for migrant stock makes little difference.
23
Table 1: Pure Cross-Section Data, 2005
Lueth & Ruiz-Arranz data set
64 observations
Dependent Variable:
Ln Remittances 2005 between Countries
Ln (Stock migrants 2000 )
Cyclical Difference
(Ln (Real GDP/ Trend GDP))
sender country vs. recipient
(1)
(2)
(3)
0.459***
0.449***
0.327**
(0.085)
(0.0818)
(0.126)
57.585***
(15.875)
Sender GDP per capita
67.909*** 65.300***
(16.998)
(17.335)
0.061***
0.062**
(0.023)
(0.024)
Estimation Method
OLS
OLS
2SLS
R2
0.373
0.448
0.4259
24
Table 2: panel study with the LRA data
• 64 pairs of countries, 2005. 1979-2005 panel.
• => 1200 or more observations
• Lagged stock of migrants replaced by its determinants:
• geographical, historical, & cultural.
• Cyclical difference now captured by unemployment.
• 2(a) The estimated coefficient on us-ur is
negative,
as now hypothesized, and highly significant.
• The t-statistic is now 9.
• 2(b) The same when applying fixed effects for
25
Table 2a: Panel Data, 1979-2005,
with time effects & random effects
1200 observations R2=.414
Lueth & Ruiz-Arranz data
Dependent variable: Ln Pair Remittances
Cyclical difference (unemployment rate)
sender country relative to recipient
-8.981***
(0.976)
Sender’s GDP per capita
0.026***
(0.007)
0.076
(1.007)
0.570**
(0.284)
-1.431***
(0.259)
Language (Dummy for common language)
0.575
(0.440)
Border (Dummy for land border)
-0.547
(0.523)
Ln (Distance between I and j)
-0.561***
(0.174)
Ln (Product Population i and j)
0.609***
26
(0.078)
Colonial relationship? (0=never, ½ =post-1945,
1=always)
Islands ?
(0 = neither, 1 = one of the pair, 2 = both)
Landlocked (0 = neither, 1 = one of the pair, 2 = both)
Table 2b: Panel Data, 1979-2005, Lueth & Ruiz-Arranz data
1228 observations
Dependent variable: Ln Remittances between country pairs
(3)
(4)
(5)
-11.607***
-11.483***
-12.918***
(1.125)
(1.947)
( 1.976)
0.049***
0.049***
0.023
(0.006)
(0.009)
(0.017)
2.963***
3.017***
1.209
(0.243)
(0.400)
(0.974)
Time Effects
no
no
Yes
Random Effects
no
no
No
Fixed Effects for Countries
yes
no
No
Fixed Effects for Country Pair
no
yes
Yes
0.696
0.304
0.331
Cyclical difference
(unemployment rate)
sender country relative to recipient
Sender’s GDP/cap
Ln (Product
Population i and j)
R2
Constants not reported.
Standard errors in parenthesis. Significance * 10% level, ** 5% level, & *** 1% level
27
Table 3: cross-section study (2003-04)
with extended composite data set
Sources: Western Hemisphere data: FOMIN & the Central Banks;
EU data: Jiménez-Martín, Jorgensen & Labeaga, EC (2007);
Lueth & Ruiz-Arranz, IMF(2006).
•
•
•
•
•
Approximately 330 bilateral observations.
Lagged stock of migrants (2000) .
Cyclical difference again captured by GDP/trend.
The estimated coefficient >0 & highly significant.
So is the coefficient on currency union dummy.
– under OLS, but not under IV.
– Causality between CU & remittances is unclear.
28
Table 3:
Cross-Section, 2003-04; Composite data set
Dependent Variable: Ln Remittances between Countries
Ln Stock migrants 2000
(1)
(2)
(3)
0.762***
0.741***
1.061***
(0.040)
(0.041)
(0.088)
16.199***
16.099***
14.723***
(2.905)
(2.765)
(3.390)
0.039***
0.028*
(0.015)
(0.016)
1.345***
0.087
(0.222)
(0.389)
OLS
2SLS
Cyclical Difference
Ln (Real GDP/ Trend)
Sender relative to recipient
Sender GDP per cap
Currency Union
Estimation Method
OLS
border/language/
islands/colonial
Instrumental variables
29
Observations
331
328
328
To summarize the findings,
• splicing together a larger bilateral data set from three
data sets used by others,
• has allowed a moderately strong verdict on the question
of cyclicality.
• It runs contrary to the analogy with
capital flows /the Dutch Disease:
• Remittances respond positively to the cyclical
position in the sending country and negatively to
the cyclical position in the receiving country.
30
Policy implications
• This counter-cyclical pattern is precisely what one wants.
– It suggests that emigrants’ remittances can play some of
the stabilizing role that capital flows often promise but
seldom deliver.
• If the finding holds up under further investigation, it
carries at least two specific policy implications.
– First, it suggests governments should not try to harness
remittances in the name of national development, but
rather should allow emigrants to transact freely.
– Second, it suggests that remittances belong on the list
of Optimum Currency Area criteria,
• along with trade, labor mobility, & transfers.
31
Acknowledgements
I wish to thank:
• Olga Romero for research assistance;
• Erik Lueth & Marta Ruiz-Arranz
for generously making data available,
• Maurice Kugler & Hillel Rapoport for comments;
and
• the MacArthur Foundation,
the Center for International Development,
and Robert Hildreth for support.
32
Jeffrey Frankel
James W. Harpel Professor of Capital Formation & Growth
Harvard Kennedy School
http://ksghome.harvard.edu/~jfrankel/index.htm
Blog: http://content.ksg.harvard.edu/blog/jeff_frankels_weblog/
APPENDICES
References on procyclicality of government
spending in developing countries
•
•
•
•
•
•
Gavin & Perotti (1997);
Lane and Tornell (1999),
Kaminsky, Reinhart, & Vegh (2004),
Talvi & Végh (2005),
Mendoza & Oviedo (2006),
Alesina, Campante & Tabellini (2008).
34
References on failures of capital flows
• Flows in practice fail to deliver on the textbook promises.[1]
• Rather than smoothing short-term disturbances,
private capital flows are often procyclical:
– pouring in during boom times, disappearing in recessions,[2]
• especially among commodity-producers
• Rather than flowing from high capital/labor countries (e.g., the US)
to low capital/labor countries, capital often “flows uphill.”[3]
• Rather than rewarding only countries that follow sound economic
policies and punishing those that follow bad policies, capital
flows sometimes aid & abet irresponsible budget deficits.[4]
[1]
[2]
Prasad, Rogoff, Wei, & Kose (2003, 04); Prasad & Rajan (2008).
Kaminsky, Reinhart, & Vegh (2005); Reinhart & Reinhart (2009); Perry (2009);
Gavin, Hausmann, Perotti & Talvi (1996); Mendoza & Terrones (2008).
[3] Lucas (1990); Alfaro, Kalemli-Ozcan & Volosovych (2005); Gourinchas & Jeanne (2007);
Prasad, Rajan & Subramanaian (2007); Kalemli-Ozcan, Reshef, Sorensen & Yosha (2009).
[4] Lane & Tornell (1998).
35
Outlook for Remittance Flows 2010-11, D.Ratha, S.Mohapatra & A. Silwal,
Migration & Remittances Team, World Bank, April 23, 2010
36
Remittances began to recover in 2009
in India
&
the Philippines
Source: Reserve Bank of India
37
Appendix Figure 1a :
Bilateral stock of migrants (normalized by populations),
versus remittances (normalized by GDPs)
0
2.00e-084.00e-086.00e-088.00e-081.00e-07
Migrants 1990 and Remittances
0
.0005
Migration1990/(Popi*Popj)
Average Ratio Remittances/Income
.001
Fitted values
Sources: Central America data: FOMIN & the Central Banks. data from 2000-2007;
Jiménez-Martín, Jorgensen, & Labeaga, (2007), data from 2000-2005;
38
Lueth & Ruiz-Arranz (2006). Data from 1979-2005.
Appendix Figure 1b :
Bilateral stock of migrants (normalized by populations),
versus remittances (normalized by GDPs)
2.00e-084.00e-086.00e-088.00e-081.00e-07
Migrants 1990 and Remittances
PRTMDA
PRTMDA
ESPECU
0
ITAMDA
ITAALB
GRCMDA
FRAMAR
IRLMDA
PRTUKRUSANIC
USASLV
BELMAR
GBRGHA
USAHND
IRLYUG
AUTYUG
ESPMAR
NLDMAR
ESPBOL
IRLMDA
USADOM
CHEMKD
CHEHRV
USAGTM
ESPDOM
ESPCOL
CHEYUG
TURMDA
ESPMDA
USABLZ
AUTHRV
FRADZA
HUNYUG
LUXHRV
FRATUN
USAPHL
AUTMKD
GBRPAK
GBRBGD
ESPROM
PRTGEO
DEUYUG
DEUMKD
GBRMDA
BELMKD
DEUHRV
AUTMDA
GRCPHL
ITAMKD
DEUMDA
FRAMDA
ESPPER
GRCLBN
AUSHRV
GBRNGA
GRCJOR
USAPAN
PRTBRA
USACRI
IRLBLR
NORHRV
GBRPHL
GBRHRV
GRCYUG
DNKMKD
SWEMKD
HUNHRV
SWEYUG
NLDHRV
IRLUKR
ITAHRV
NZLTHA
NORPHL
DEUMAR
AUSPHL
SWEHRV
CANHRV
IRLHRV
GBRYUG
PRK/KORTHA
FRAYUG
NLDPHL
ITAPHL
CHEPHL
BELHRV
AUSYUG
NLDMKD
AUTSVN
JPNPHL
USAMDA
GRCMKD
FRAHRV
ITAMAR
DEULBN
CHESVN
DEUUKR
PRK/KORIDN
DEUPHL
GBRMKD
AUSTHA
GBRBGD
DEUTUN
ESPBGR
USAHRV
NORMKD
PRTMAR
IRLPHL
BELECU
GBRTHA
DNKTHA
PRK/KORPHL
ITAECU
DNKHRV
CHETHA
ITAYUG
ESPARG
USAMKD
CANPHL
CANMKD
CANYUG
TURMKD
GBRCHN
JPNTHA
USABGD
ESPBRA
GRCBLR
AUTPHL
SWETHA
USAYUG
PRTVEN
GRCEGY
LUXSVN
GRCUKR
DEUSVN
HUNUKR
HUNISR
IRLBRA
ESPUSA
FRAKAZ
FRAPHL
JPNKAZ
JPNIDN
PRK/KORKAZ
ESPMKD
TURKAZ
GBRKAZ
BELCOL
BELPER
ITABRA
ITAUKR
POLSYR
BELPHL
DEUKAZ
USAKAZ
ITACOL
DEUTHA
ITAVEN
ESPHRV
ESPPHL
SWEPHL
GRCISR
ITAARG
ITAEGY
NLDTHA
NORTHA
GBRSVN
DNKPHL
USATHA
ITAPER
BELSVN
USASVN
BELISR
ITASVN
FRAMKD
FRASVN
BELTUN
SWESVN
NLDSVN
HUNSVN
CANSVN
AUSSVN
POLBLR
0
.0005
Migration1990/(Popi*Popj)
Average Ratio Remittances/Income
AUSMKD
.001
Fitted values
Sources: Central America data: FOMIN & the Central Banks. data from 2000-2007;
Jiménez-Martín, Jorgensen, & Labeaga, (2007), data from 2000-2005;
Lueth & Ruiz-Arranz (2006). Data from 1979-2005.
39
Appendix Table 2: IMF data set
5.00e-07 1.00e-06 1.50e-06 2.00e-06
Migrants and Remittances
LUXHRV
CHEHRV
AUTHRV
CHEMKD
DEUHRV
0
NORHRV
BELMKD
GBRHRV
USAPHL
SWEHRV
LUXSVN
ITAHRV
BELHRV
CANHRV
DNKHRV
USAHRV
DNKMKD
ITAMKD
GRCPHL
SWEMKD
CHEPHL
USAPHL
HUNHRV
NZLTHA
ITAPHL
DNKTHA
NORPHL
DEUYUG
FRAHRV
AUSSVN
USAMKD
DEUSVN
CHETHA
GBRMKD
AUSTHA
CANPHL
NORMKD
BELSVN
AUSPHL
DEUPHL
GBRTHA
SWETHA
SWESVN
PRK/KORTHA
FRAYUG
NORTHA
AUSPHL
CANMKD
GBRYUG
GBRPHL
FRAMKD
GRCMKD
IRLHRV
NLDSVN
PRK/KORPHL
NLDPHL
ESPHRV
CHEPHL
HUNYUG
ITASVN
JPNPHL
GBRSVN
NLDTHA
USATHA
GRCPHL
USAYUG
USASVN
FRASVN
GRCYUG
ITAYUG
HUNSVN
CANSVN
TURMKD
AUSYUG
FRAPHL
IRLPHL
BELPHL
USABGD
ESPPHL
SWEPHL
DEUTHA
AUTPHL
DNKPHL
GBRBGD
ITAPHL
AUTPHL
IRLYUG
IRLPHL
CANPHL
CANYUG
0
AUTMKD
AUSHRV
NLDHRV
DEUMKD
AUTYUG
NLDMKD
SWEYUG
AUSMKD
CHESVN
CHEYUG AUTSVN
.0005
.001
Migration1990/(Popi*Popj)
remittances_ad
.0015
Fitted values
40