International Remittances: Different Estimation Methodologies	 
Produce Different Results (28-MAR-06, GAO-06-210).		 
                                                                 
Remittances are the personal funds that the foreign born send to 
their home countries. In recent years, estimated remittances have
grown dramatically, and policy makers have increased their	 
attention to these flows. Organizations use various methodologies
to estimate remittance flows, which result in a range of	 
estimates. In 2004, the Group of Eight (G8) leaders emphasized	 
the need for improved statistical data on remittances. In light  
of the growing volume of remittances and the differences in	 
estimates, GAO examined (1) the methodology that the Bureau of	 
Economic Analysis (BEA) uses to develop the official U.S.	 
estimate, (2) methodologies that other countries and multilateral
organizations use to estimate remittances, and (3) international 
efforts to improve the collection and reporting of remittance	 
data.								 
-------------------------Indexing Terms------------------------- 
REPORTNUM:   GAO-06-210 					        
    ACCNO:   A50217						        
  TITLE:     International Remittances: Different Estimation	      
Methodologies Produce Different Results 			 
     DATE:   03/28/2006 
  SUBJECT:   Data collection					 
	     Data integrity					 
	     Immigrants 					 
	     International relations				 
	     Monetary policies					 
	     Surveys						 
	     Statistical data					 
	     Statistical methods				 
	     Remittances					 

******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO Product.                                                 **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
******************************************************************
GAO-06-210

     

     * Report to the Committee on Banking, Housing, and Urban Affairs, U.S.
       Senate
          * March 2006
     * INTERNATIONAL REMITTANCES
          * Different Estimation Methodologies Produce Different Results
     * Contents
          * Results in Brief
          * Background
          * BEA Uses a Model to Estimate Remittances Sent from the United
            States, but the Accuracy of BEA's Estimate Is Uncertain for
            Several Reasons
               * BEA's Methodology Relies on Existing Data and BEA's
                 Assumptions to Estimate Remittances
               * The Accuracy of BEA's Remittances Estimates Is Affected by
                 the Quality of the Data and the Assumptions Used in the
                 Methodology
               * BEA's Revised Methodology Includes Remittances Sent by Some
                 of the Foreign Born Who Have Been in the United States for
                 Less Than One Year
          * Other Entities Use Different Estimation Methodologies Resulting
            in a Range of Remittance Estimates
               * Central Banks of Mexico and the Philippines Track Remittance
                 Flows into Their Countries
               * The IDB Uses a Variety of Sources to Estimate Remittances
                 Flows to Latin America and the Caribbean
               * Remittance Estimates Made by These Entities Vary
          * BEA Is Involved in International Efforts to Improve the
            Collection and Reporting of Remittance Information
               * International Estimates of Remittances Are Incomplete and Do
                 Not Reconcile
               * International Working Group Was Established in 2005 to
                 Improve Remittance Data
          * Observations
          * Agency Comments and Our Evaluation
     * Objectives, Scope, and Methodology
     * BEA's Methodology for Estimating Remittances
          * BEA's Estimate of Remittances Includes Remittances from Some of
            the Foreign Born Who Have Been in the United States for Less Than
            One Year
     * Analysis of the Sensitivity of BEA's Estimate to Judgmentally
       Determined Variables on the Remitting Behavior of the Foreign Born
     * IDB Remittance Estimation Methodology
     * Comments from the Department of Commerce
          * GAO's Comments
     * Comments from the Department of the Treasury
     * GAO Contact and Staff Acknowledgments
     * PDF6-Ordering Information.pdf
          * Order by Mail or Phone

Report to the Committee on Banking, Housing, and Urban Affairs, U.S.
Senate

March 2006

INTERNATIONAL REMITTANCES

Different Estimation Methodologies Produce Different Results

Contents

Tables

Figures

March 28, 2006Letter

The Honorable Richard C. Shelby Chairman The Honorable Paul S. Sarbanes
Ranking Minority Member Committee on Banking, Housing,   and Urban Affairs
United States Senate

The United States is the largest remittance-sending country in the world,
with a majority of funds sent to Latin America and the Caribbean, and
substantial amounts sent to Asia and Africa. In recent years, remittances
have received growing attention from policy makers in both developed and
developing countries because these flows serve as an important financial
source for some countries. According to the World Bank, remittances
received by developing countries were estimated to have been $167 billion
in 2005, up 73 percent from 2001; however, given that the extent of
unrecorded flows through formal and informal channels is unknown, actual
remittance flows may be much higher. World Bank data show that remittance
growth has outpaced private capital flows and official development
assistance over the last decade. When combined with official U.S.
development assistance, these flows significantly increase the percentage
of U.S. gross national income sent to developing countries.

In 2004, the Group of Eight (G8) leaders emphasized the importance of
remittances and the need for improved statistical data on them.1 In the
United States, some agencies have also expressed a need for improved
remittance estimates. For example, the Department of the Treasury
(Treasury) conducts bilateral outreach programs, and Treasury officials
believe improved remittance statistics could help it better target its
program to improve the financial infrastructure in countries that receive
a large amount of remittances from the United States. In 2004, the Federal
Reserve established a mechanism to facilitate the provision of low-cost
remittances to Mexico through its automated clearinghouse; improved
remittance statistics could help it identify other countries that could
benefit from its low-cost remittance product.

Different organizations use various methods to estimate remittance flows,
which result in a range of estimates. In light of the volume of
remittances and the differences in estimates, you asked us to review the
methodologies used to estimate remittances from the United States.2
Specifically, we examined (1) the methodology that the Department of
Commerce's (Commerce) Bureau of Economic Analysis (BEA) uses to develop
the official U.S. estimates of remittances from the United States, (2)
methodologies other countries and multilateral institutions use to
estimate remittances from the United States, and (3) international efforts
to improve the collection and reporting of remittance data. In addition,
we recently issued a report that focused on remittance products, costs,
and consumer disclosures for remittances sent from the United States to
other countries.3

To address these objectives, we met with officials at BEA and the U.S.
Census Bureau. We also reviewed documentation that described BEA's
methodology prior to 2005 and obtained documentation from BEA describing
their revised methodology, which was implemented in July 2005. We met with
officials from the Inter-American Development Bank (IDB), the
Inter-American Dialogue, and the Mexican and Philippine central banks to
obtain their estimates on remittances from the United States. We also
obtained descriptions of the methodologies they used to estimate
remittances, the reasons for using these methodologies, and the potential
advantages and disadvantages of their use. We primarily report data for
2003 because that is a time period for which BEA has statistically
reliable data and because data for more recent time periods are
preliminary. As a matter of consistency we use this period to report on
the other entities as well. We do, however, report more recent data when
available. We met with officials from BEA, Treasury, the Department of
State, the U.S. Agency for International Development, the International
Monetary Fund (IMF), the World Bank, and IDB to obtain information on
international efforts to improve remittance estimates. Appendix I provides
additional details on our scope and methodology. We conducted our work
from December 2004 to March 2006 in accordance with generally accepted
government auditing standards.

Results in Brief

BEA uses a model to estimate remittances from the United States and,
although this methodology has some strengths, the accuracy of BEA's
remittance estimate is uncertain for several reasons. BEA estimated
remittances for 2003 at $28.2 billion; its model used data on the number
of foreign-born residents in the United States, their income, the
proportion of income that is remitted, and other demographic data.4
However, BEA's methodology was limited by the quality and timeliness of
data available to BEA, particularly the data on the portion of income that
is likely to be remitted. BEA revised its model in 2005 to use new data
sources from the Bureau of the Census on the demographics of the foreign
born and more recent studies on the remitting behavior of the foreign
born. It then revised its estimates back to 1991 using this new approach,
which resulted in an increase in estimated remittances for all years. Two
of the strengths of BEA's methodology are that, first, in theory, it
estimates remittances sent through both formal and informal systems; and,
second, it is low-cost to BEA because it uses existing data on the foreign
born.5 The accuracy of BEA's estimate, however, depends on the accuracy of
its assumptions. For example, BEA's revised model assumes that the
proportion of income remitted is higher for U.S. residents from developing
countries closer to the United States and that the percentage of the
foreign born that remit is the same for all migrants from all countries,
but varies depending on how long they have been in the United States.
Further, it is not possible to directly link the parameters BEA uses to
capture the remitting behavior of the foreign born to the sources cited.
Our analysis of BEA's estimates also found that they are particularly
affected by the assumptions BEA used on the percentage of income remitted
and the percent of foreign born that remit. We used a statistical
technique that repeatedly and randomly samples from underlying data to
obtain a range for 90 percent of possible estimates and determined that
estimated remittances from the United States could range in value from
$17.3 billion to $35.9 billion. Finally, BEA's remittance estimate
includes remittances sent by some of the foreign born who have been in the
United States for less than one year, who, according to BEA's definition
of remittances, should not be included.

Some central banks and the IDB use a variety of methodologies and data
sources to provide estimates of remittances from the United States that
vary significantly. For example, Mexico's central bank estimates
remittances by collecting data on the amount remitted through money
transmitters and by surveying Mexican nationals returning to the country
at the U.S.-Mexico border. The Philippine central bank estimates
remittances by tracking the income of its residents working abroad that is
channeled into banks in the Philippines and netting out living expenses to
estimate remittances. The primary advantage of these methodologies is that
they capture actual or projected estimates of remittance flows. Unlike
BEA's methodology, however, these methods are limited in their ability to
capture remittances made through the informal sector. The IDB, which
provides financing for economic, social, and institutional development
projects for Latin America and the Caribbean, estimates remittances on a
regional basis. The IDB developed its own estimate of remittances from the
United States to Latin America and the Caribbean using a survey of the
annual remittance amounts sent from Latin American residents of the United
States to their countries of origin. In addition, the IDB has conducted
surveys of residents in countries in the region who receive remittances
and compared their estimates with those of central banks in these
countries. These in-country surveys also have allowed IDB to estimate the
amount of remittances these countries receive from the United States. Such
survey efforts can provide statistically valid estimates but can be costly
to implement and rely on the willingness of respondents to share
information. The remittance estimates generated by the central banks, IDB,
and BEA vary significantly from one another. For example, in 2003, the IDB
estimated that $30.1 billion was remitted from the United States to Latin
America and the Caribbean. Although BEA does not publish remittance
estimates by region, we aggregated BEA's country-by-country tabulations to
estimate remittances to Latin America and the Caribbean and found this to
be $17.9 billion.

BEA is an active participant in international efforts to improve
remittance statistics, but these initiatives, which began in 2005, are in
the early stages and have not yet produced results that make it easier to
reconcile remittance estimates. Currently, countries and other entities
that estimate remittances use a variety of methods such as model
estimation, bank reporting systems, or surveys of remitters; each method
has strengths and limitations as we discussed earlier. Further, few
countries seek to reconcile their estimates with other countries, and
others are unable to devote significant resources to collecting data on
remittances. For these reasons, the heads of the G8 countries at the 2004
G8 summit called upon international financial institutions such as the
World Bank and IMF to lead a global effort to improve remittance
statistics. As part of this effort, the World Bank and IMF hosted a
meeting in January 2005 and proposed that two different groups undertake
an objective: (1) to clarify the definition of remittances, and (2) to
provide guidance on how to collect and estimate remittance flows. In the
first case, the United Nations Technical Subgroup on the Movement of
Natural Persons, of which BEA is a member, recommended that "personal
transfers" be defined to include personal transfers received by resident
households from nonresident households. This definition was discussed at a
June 2005 meeting of the IMF Balance of Payments Committee and is expected
to appear in the updated international statistical standards that are
scheduled to take effect in 2008. In the second case, Eurostat, the
statistical agency of the European Union, agreed to host and jointly
organize a meeting in June 2006 of a new group called the Luxembourg
Group, to develop more detailed guidance for compiling remittance data. As
of March 2006, BEA had not been formally asked to participate on this
group but expected that it would be. No date has been set for the group to
complete its work.

Although we make no recommendations at this time, we observe that
estimates of the amount of remittances from the United States differ based
on different methodologies. More accurate remittance estimates could help
certain U.S. agencies such as Treasury make better decisions on how much
(and what kind of) development assistance to provide, and U.S. companies
could make better decisions regarding foreign direct investment.
Therefore, policy makers may want to consider exploring options for
improving the accuracy of U.S. remittance statistics.

We provided a draft copy of this report to Treasury and Commerce for their
review and comment. Treasury concurred with all of our observations, while
Commerce concurred with most of them. Commerce also provided a number of
additional specific comments. Commerce's comments and our response are
discussed in appendix V.

Background

Remittances have become an important source of financial flows to
developing regions and have been resilient in the face of economic
downturns. These funds can be used for various purposes, including basic
consumption, housing, education, and small business formation; they can
also promote financial development in cash-based economies. Because of the
importance of these flows to many developing countries, in recent years,
countries that send remittances and receive remittances, along with
international organizations, have expressed increasing interest in
understanding immigrants' remittance practices.

According to the 2000 Census, the 1990s saw the largest increase in the
foreign-born population that entered the United States, compared with any
other 10-year period. IMF figures show that in 2004, immigrants in the
United States sent over $29.9 billion in remittances, more than any other
country. Saudi Arabia was the second largest remittance-sending country;
however, as shown in figure 1, the volume of remittances from Saudi Arabia
has been falling since 1994, while that from the United States has been
steadily increasing.

Figure 1: Largest Sources of Remittances, 1990-2004

For some countries, remittances constitute the single largest source of
foreign currency and can often rival direct foreign investment in amounts.
World Bank data show that for selected countries remittances exceed the
flows of official development assistance and foreign direct investment and
are relatively large compared to exports and gross national
income-particularly for the Dominican Republic and the Philippines (see
table 1).

Table 1: Remittances as a Percentage of Various Economic Indicators for
Selected Countries for 2003

                                        

     Country    Remittances Remittances   Remittances Remittances Remittances 
                       as a        as a          as a        as a        as a 
                 percentage  percentage percentage of  percentage  percentage 
                 of exports of official       foreign    of gross  of foreign 
               of goods and development        direct    national    reserves 
                   services  assistance    investment      income 
                                          net inflows             
Brazil              3.4%      953.1%         27.8%        0.6%        5.7% 
Colombia            18.3       383.5         176.2         3.8        28.5 
Dominican           27.1      3371.5         750.2        12.5       918.4 
Republic                                                       
Egypt               16.6       331.3        1247.3         3.2        21.8 
India               19.2      1847.3         407.7         3.0        17.6 
Mexico               8.2     14147.9         135.3         2.3        24.8 
Morocco             25.6       691.3         158.6         9.2        26.1 
Nigeria              5.9       528.0         139.8         3.5        23.5 
Philippines         20.2      1068.9        2470.2         9.1        57.7 
Poland               3.3       194.2          56.1         1.1         7.1 
World Total          1.8       223.5          30.2         0.5         5.6 

Source: GAO analysis of World Bank data.

Notes: Data are derived from remittance estimates reported in the World
Bank's Global Development Finance. "Reserves" are total foreign exchange
reserves, excluding gold. The countries selected are the top recipients of
remittances in their respective regions.

Remittances are also very important for those households that receive
them. Table 2 shows the minimum wage per month for several developing
countries as well as our computation of the 2003 per capita remittances
from the United States per month. As can be seen from this table,
remittances received by households on a monthly basis tend to
substantially exceed the monthly minimum wage for these countries. For
example, per capita, remittances to households in the Philippines are
almost five times the monthly minimum wage a Filipino worker would make in
the retail and service sector.

Table 2: Per Capita Remittances from the United States Compared to Minimum
Wages for Selected Countries, 2003

                                        

     Country   Minimum wage      Per capita    Percentage of  Month of income 
                  per month     remittances minimum wage per   at the minimum 
                   (in U.S. from the United     month to per    wage that the 
                   dollars)      States per           capita       per capita 
                             month (in U.S. remittances from  remittance from 
                                   dollars)       the United       the United 
                                            States per month     States would 
                                                                  replace (in 
                                                                      months) 
Bangladesh           $30            $137              457              4.6 
El Salvador           49             189              386              3.9 
Ghana                 26             177              681              6.8 
Philippines           38             178              468              4.7 
Romania               84             200              239              2.4 

Source: GAO calculations using BEA's underlying tabulations for
remittances from the United States in 2003 and data from the International
Labor Organization's minimum wages database.

Note: The minimum wage in developing countries generally applies to urban
workers.

The IMF collects and publishes official estimates of remittances sent from
its member countries, including the United States, as part of its balance
of payments statistics. The IMF currently reports the sums of "workers'
remittances" and "compensation of employees" as the best measure of total
personal remittances. According to IMF, "workers' remittances" are
transfers by migrants who are employed in countries other than their birth
countries and are considered residents there; "compensation of employees"
is made up of wages, salaries, and other benefits earned by individuals in
economies other than those in which they are residents, for work performed
or paid for by residents of those economies. As a result, compensation of
employees applies only to individuals away from their place of origin for
less than a year.6

In the United States, no U.S. government agency tracks the flow of
remittances through the payment system. Because of its role in compiling
balance of payments statistics, BEA provides to the IMF official estimates
of U.S. remittance inflows and outflows. BEA publishes remittance
estimates in a different manner than reported in the IMF's balance of
payment statistics. BEA includes estimates of remittances by the
foreign-born population residing in the United States to households abroad
in the published item called "private remittances and other transfers."
This category is broader than the international definition of remittances,
as it also includes payments or receipts of nongovernmental U.S. entities
and foreign entities. Also, BEA publishes its estimates of "private
remittances and other transfers" in its tables of international
transactions accounts, defining it as the difference between transfers to
and transfers from the United States. However, BEA provides to the IMF an
estimate of remittances that flow from the United States to the world
based on its underlying country-by-country tabulations. Until this year,
BEA only provided this estimate to the IMF. For the first time, BEA
published the estimate it provided to the IMF, as well as revised
estimates back to 1991, in the July 2005 Survey of Current Business.7

The majority of remittances from the United States flow to Latin America,
which includes Mexico, Central America, South America, and the Caribbean
(see fig. 2). A large amount also flows to Asia, including the
Philippines.

Figure 2: Regional Destination of Remittances Sent from the United States,
2003

Note: The shaded parts of the bar signify those countries, or in the case
of Europe, regions, that make up the large majority of remittances
received from the United States in each location.

There are many obstacles to accurately estimating remittances. First, many
transactions may go through unregulated informal channels from which
information cannot be garnered for inclusion in official estimates. While
there are no official estimates, some experts believe that a large amount
of remittances flow through this system, with market observers estimating
that informal flows can range from 50 percent to 250 percent of recorded
remittance flows. Second, countries do not always report remittance
estimates or do not report them according to commonly held IMF
definitions, which exclude transfers by the foreign born who have been
in-country for less than one year. Variations in data compilation
procedures occur partially due to different interpretations of definitions
and classifications. In most cases, however, data weaknesses and omissions
are due to difficulties in obtaining the necessary data. For example, the
World Bank and other international organizations have indicated that
developing countries with large remittance inflows often have a relatively
weak capacity and limited resources, even though remittances are a large
item in their balance of payments statistics. Countries with large
remittance outflows often give lower priority to improvements in
remittance statistics because they are a relatively small item in their
balance of payments statistics, according to the World Bank and other
international organizations.

BEA Uses a Model to Estimate Remittances Sent from the United States, but
the Accuracy of BEA's Estimate Is Uncertain for Several Reasons

BEA uses a model to estimate remittances (which it calls "personal
transfers") from the United States. Although BEA's methodology has some
strengths, the accuracy of BEA's estimate is uncertain for a number of
reasons. BEA estimated that remittances from the United States in 2003
were $28.2 billion. To arrive at this estimate, BEA used a model that
estimates remittances based on demographic information on the foreign
born, such as their total number, income, and the percentage of income
they remitted. In 2005, BEA revised its model for estimating remittances
and incorporated more current Census Bureau data on the size and
demographic characteristics of the foreign-born population of the United
States; however, the model is limited particularly by lack of current data
on the proportion of income immigrants were likely to remit and the
assumptions BEA makes about its data. In addition, BEA uses the more
current census data in a way that may double-count some immigrants.

BEA's Methodology Relies on Existing Data and BEA's Assumptions to
Estimate Remittances

Prior to 2005, to derive its annual estimate of remittances sent from the
United States, BEA developed a model comprised of three factors-the number
of the foreign born, their family income, and the proportion of income
remitted. The count of the foreign born, their income, and other
demographic characteristics were obtained from information aggregated
annually from U.S. Bureau of the Census surveys. These data were arrayed
by length of residency in the United States and family types linked to
marital status (e.g., married foreign head of households, native-born
married to foreign-born spouse, and unmarried individuals). The remitter
was assumed to be the household head. BEA extrapolated the foreign-born

population derived from the 1990 Decennial Census using indicators,
including the Census Bureau's annual Current Population Survey (CPS).8

To estimate the proportion of income immigrants were likely to remit, BEA
relied on the 1989 Legalized Population Survey (LPS1) and the 1992
Legalized Population Follow-Up Survey (LPS2), which were conducted as a
result of the Immigration Reform and Control Act of 1986 (IRCA).9 BEA then
combined the information obtained from LPS1 and LPS2 with demographic and
income information obtained from the CPS to arrive at the total amount of
remittances sent from the United States. For a more detailed description
of BEA's methodology for estimating remittances, see appendix II.

In 2005, BEA made several revisions to its methodology to include more
recent census data, and recent studies on the foreign born and their
remitting behaviors. First, BEA incorporated data on the foreign-born
population and their income from the 2000 Census and the American
Community Survey (ACS), which is available annually, unlike decennial
census data, and thus requires less extrapolation of population and income
trends.10 According to BEA, these data will enable a better breakdown of
the foreign-born population by all relevant characteristics on an annual
basis. The ACS data on the number and income of the adult foreign-born
population are arrayed by their gender, duration of stay, presence or
absence of children, and per capita income of recipient countries and
proximity to the United States. BEA then used its own judgment to
determine the percentage of the adult foreign-born population that remits
and the probability of remitting from information gathered from various
academic studies published between 1995 and 2004, as well as LPS1 and

LPS2, which BEA used in its earlier model.11 BEA revised its estimates
back to 1991 using this new approach, which resulted in an increase in
estimated remittances for all years. Figure 3 shows the data that are
included in BEA's model and how the remittance estimate is calculated.

Figure 3: BEA's Methodology for Estimating Remittances, 2005

In most cases, BEA provides only a global estimate of remittances and does
not publish remittance statistics about remittances from the United States
to individual countries. BEA stated that some data elements are not
available for some time periods or geographic areas, so it must undertake
a variety of methods to fill the data gaps in order to produce the
underlying tabulations needed for an aggregate estimate for the world. BEA
cautions that disaggregating its estimate for the world is error-prone and
expresses confidence only in its aggregate estimate. Further, according to
BEA, in moving from the global estimate to increasingly smaller geographic
areas or countries, the average errors in the underlying tabulations
increase. When it estimates remittances for selected regions, it publishes
them on a net (inflows minus outflows) basis.

The Accuracy of BEA's Remittances Estimates Is Affected by the Quality of
the Data and the Assumptions Used in the Methodology

BEA's approach has several strengths: in theory, it captures both formal
and informal channels of sending remittances. It is also low-cost because
it relies on available data and not on eliciting data from a foreign-born
population that may not have an incentive to provide accurate data.
However, the accuracy of BEA's estimate is affected by the quality of the
data available to BEA. A critical component of the methodology relies on
information about the remitting behavior (e.g., amount, frequency) of the
foreign born. Prior to 2005, the primary data available to BEA were the
1989 LPS1 and the 1992 LPS2; however, these surveys may not have been
appropriate for use in estimating remittances of all the foreign born
because they sampled a population participating in a special legalization
program primarily aimed at Latin American immigrants. The LPS1 and LPS2
excluded undocumented aliens, temporary residents who did not wish (or
were not eligible) for legal status, and legal immigrants who became
legalized through processes other than IRCA. The survey design did not
provide a way to more extensively sample immigrant groups more likely to
remit than others (e.g. the foreign born with less than 10 years of
residence in the United States). In addition, recent census data show that
some basic demographic characteristics of the foreign born have changed
significantly since the LPS1 and LPS2 surveys were done.

BEA's revisions to its methodology recognize these changes in the foreign
born population. In its revision, BEA reviewed a number of academic
studies to update the findings of the LPS1 and LPS2 and published the
sources in the July 2005 Survey of Current Business, however, the
estimates on the proportion of income remitted cannot be directly tracked
to these source documents. Although this approach is more transparent than
the prior approach of relying primarily on LPS1 and LPS2, BEA's estimate
is still affected by its "judgment" of how it incorporates information
from the academic studies it is now using, and the assumptions it makes in
its model. For example, two of BEA's assumptions are that the proportion
of income remitted is higher for U.S. residents from developing countries
than developed countries, and that the percentage of the foreign born that
remit is the same for all countries and only varies based on how long they
have been in the United States. Our analysis suggests that the final BEA
estimates of remittances are affected by these assumptions. We used a
statistical technique that repeatedly and randomly samples from underlying
data to obtain the range for 90 percent of possible estimates and
determined this to be between $17.3 billion and $35.9 billion. See
appendix III for the analysis we used to determine these ranges.

BEA's Revised Methodology Includes Remittances Sent by Some of the Foreign
Born Who Have Been in the United States for Less Than One Year

Remittance estimation in the balance of payments framework generally
separates remitters by their length of residency in host countries. All
remittances are presumably sent by the foreign born who have been in the
host country for greater than one year, while those that are in a country
for less than a year are presumed to be temporary, earning only
compensation. For this reason, some experts compile remittances as the sum
of (1) the remittances sent by those in country greater than a year and
(2) the compensation for those in-country for less than a year. In its
description of its revised methodology, BEA states that it excludes
transfers by the foreign born who have been in the United States for less
than 1 year from its measure of remittances; however, BEA uses a
U.S.-residency-duration grouping of 0-5 years in its personal remittances
calculation. It thus includes both employees who are in the United States
for less than or equal to 1 year, and migrants who are in the United
States for more than a year, in its estimates of personal remittances. Our
analysis determined that BEA's estimates of remittances are therefore
potentially overstated by up to $377 million because they include
estimates for approximately 467,000 foreign-born individuals who were in
their first year of residency in the United States, according to 2003 ACS
data.

Other Entities Use Different Estimation Methodologies Resulting in a Range
of Remittance Estimates

Some central banks and IDB use a variety of methodologies and data sources
to estimate remittances.12 The central banks of Mexico and the
Philippines, two of the major recipients of remittances from the United
States, track funds coming into their countries. The IDB, a multilateral
organization that provides financing for economic, social, and
institutional development projects for Latin America and the Caribbean,
estimates remittances on a regional basis-primarily through the use of
surveys. The remittance estimates produced by these methodologies vary
from each other and from BEA's estimates, thus further illustrating the
dependency of estimates on their methods and data.

Central Banks of Mexico and the Philippines Track Remittance Flows into
Their Countries

The Central Bank of Mexico, known as the Banco de Mexico (Banxico), tracks
remittance flows to Mexico with the help of a regulatory reporting
requirement on money transmitters. Since 2003, Mexico's methodology for
estimating remittances has required firms that receive remittances to
report the amount of money received and the number of transactions
conducted between the United States and Mexico on a monthly basis. A
Banxico official stated that the firms' systems that channel the
information to Banxico are designed to transfer money from person to
person and that the firms make the determination if a transaction is a
person-to-person transfer. He stated that these systems are not efficient
enough for commercial transactions; the likelihood that other types of
transactions may be getting into the systems is negligible because the
systems that have been developed are designed for personal remittances.
The Banxico official stated that Banxico is confident in its estimates
because it believes the vast majority of firms (about 90 percent) are
reporting and, while some transactions that are not personal remittances
may be getting through, this is a very small portion. To track remittances
through informal channels such as couriers, at the U.S.-Mexico border
Banxico conducts a survey of Mexicans returning to visit relatives. The
survey asks questions about funds and goods they are bringing to
relatives. However, these individuals, according to the Banxico official,
are often reluctant to answer these questions.

The Philippine government has established a formal program whereby it
registers and tracks its resident Overseas Filipino Workers (OFW). This
program provides data to the government on the type of employment these
workers obtain as well as their salaries. The Philippine central bank,
known as the Bangko Sentral ng Pilipinas (BSP), estimates remittances
channeled into banks, which are already net of living expenses of these
workers. However, BSP officials caution that the country source data are
not truly reflective of remittances coming from a country, particularly
from the United States, because most remittance centers for OFWs (e.g.,
Saudi Arabia, Japan, and Taiwan) send funds through correspondent banks in
the United States, which then send the funds to banks in the Philippines.
The BSP only captures the most immediate source of OFWs' funds coming into
the Philippines, primarily U.S. correspondent banks. Thus, this
methodology overstates the funds being remitted from the United States to
the Philippines because it includes funds from other countries, not just
from Filipino workers in the United States.

The BSP also recently revised its methodology to track remittances that
flow outside of banks using results of the Survey of Overseas Filipinos.
Specifically, these remittances are funds sent by OFWs through friends and
relatives, or amounts brought in by OFWs when they return home. This
revision caused the BSP to increase its overall estimate of remittances
into the Philippines by $1.7 billion (20 percent) in 2004. BSP officials
stated that they are in the process of updating prior years' figures.13

The primary advantages of these tracking methodologies are that they
capture actual or projected remittance flows, as well as rapid or sudden
changes in the characteristics of remitters-such as the average amount
remitted or the frequency of remitting. However, these methods are limited
in their ability to capture remittances sent through the informal sector
and to distinguish between personal remittances and other types of
personal business transactions when money transfer operators and banks do
not correctly code the remittance transactions.

The IDB Uses a Variety of Sources to Estimate Remittances Flows to Latin
America and the Caribbean

Since the year 2000, the Multilateral Investment Fund (MIF) of the IDB has
been studying the issue of remittances and their impact on the development
of the Latin American and Caribbean region. In addition to using its own
researchers, MIF's methodology uses remittance information collected by
other researchers. The IDB remittance estimates for selected Latin
American and Caribbean countries are obtained from a combination of
sources consisting of estimates from selected central banks of recipient
member countries judged to have reasonable remittance estimates,
transaction information from remittance transfer companies to selected
countries, and information obtained from surveys of remittance senders in
the United States and remittance recipients in Latin American and
Caribbean countries. IDB officials stated that they compare the remittance
estimates that they derive from their surveys of remittance recipients in
Latin America and the Caribbean with the estimates from the central banks
of these countries. These officials also stated that these surveys have
allowed them to estimate remittances these countries have received from
the United States. According to IDB officials, for countries for which
they have not conducted an in-country survey, they use data collected from
establishments that facilitate money transfers to each country. These
officials indicated that data were obtained from a sample of 45
money-transfer businesses involving approximately 14 countries. The amount
and frequency of the average remittance sent by residents from the survey
countries was used to estimate the total remittance outflow to each
country, according to IDB officials. They also indicated that MIF staff
work with the researchers to reconcile the various estimates and arrive at
country-specific estimates they believe are fairly accurate. For a more
detailed description of IDB's methodology, see appendix IV.

The advantage of using this method to estimate remittances is that the
information is obtained from establishments that have a vested interest in
maintaining accurate data on the amount and volume of remittances.
However, estimates relying on reporting of information from remittance
providers in the formal financial sector-such as money transfer
operators-cannot account for remittances sent through the informal sector
(e.g., by couriers or hawalas). In addition, they may not be able to
distinguish between personal remittances and other types of personal
business transactions if the money transfer operators and banks do not
code the remittance transactions correctly. Although the consumer surveys
IDB used to derive its estimates collect information directly from
remittance senders and receivers, such surveys are difficult to administer
because remittance senders may be reluctant to participate in the surveys
due to language barriers, legal status, and lack of experience with
institutions that administer surveys. IDB officials also stated that
surveys only reach individuals with telephones. In addition, with these
surveys there often is a discrepancy between the amount of funds
remittance senders claim to send and the amount remittance recipients
claim to receive. Finally, these surveys can be more costly due to the
need to hire experienced survey firms with bilingual staff.

Remittance Estimates Made by These Entities Vary

The central banks of Mexico and the Philippines, the IDB, and BEA use
different methodologies to estimate remittances, resulting in a range of
estimates. For example, in 2003, the Mexican central bank estimated that
Mexico received about $13.4 billion in remittances from the United States
and the IDB estimated that Mexico received almost $12.9 billion in
remittances from the United States. In 2003, BEA estimated the amount of
remittances from the United States to Mexico at $8.9 billion.14 In terms
of remittances from the United States to Latin America and the Caribbean,
in 2003, the IDB estimated this to be $30.1 billion. Although BEA does not
publish remittance estimates by region, we aggregated BEA's
country-by-country tabulations to estimate remittances to Latin America
and the Caribbean, and found this to be $17.9 billion.

We found that the reasons for the large discrepancies in the IDB and BEA's
estimates for Latin America and the Caribbean were primarily due to
differences in population size, the percentage of persons that remit, and
the average remittance amount per year each used. Our analysis of BEA
estimates of remittances from the United States in 2003 to 21 countries
for which IDB also makes estimates show that BEA assumes that 54 percent
of the foreign born population remits an average of $2,076 per year as
shown in table 3.15 BEA assumes that the percentage of adult foreign born
that remit varies by duration of stay and the absence or presence of
children in the household. To determine the $2,076 that is, on average,
remitted per year, we used information from BEA's underlying tabulations
and calculated the average remittance per person for the 21 countries. BEA
assumes that the percent of income remitted varies by the presence or
absence of children, the type of countries of birth (according to economic
development), and proximity to the United States. In contrast, based on
our analysis of IDB's survey results, 70 percent of percent of adult
foreign-born Hispanics remit and on average, they remit $3,024 per year,
as shown in table 3.

Table 3: Comparison of BEA and IDB Estimates of Remittances to 21 Latin
American and Caribbean Countries, 2003

                                        

       Number in the population (in                     
                 millions)                              
                                                    BEA                   IDB 
                                                   14.7                  16.7 
                                                                              
                                           (Total adult (Total adult Hispanic 
                                          foreign born)           population) 
Less adjustment to exclude U.S.-born             NAb                   2.0 
Hispanics who have been found not to                 
remita (in millions)                                 
Adjusted number of Hispanic                     14.7                  14.7 
foreign-born population                              
Number of Hispanic foreign born who              7.8                  10.2 
remit                                                
Implied percentage of Hispanic                   54%                  70%c 
foreign born who remit                               
Average annual remittance sent                $2,076                $3,024 
Total estimated remittances to 21              $16.3                 $30.8 
Latin American and Caribbean                         
countries (in billions)                              

Source GAO.

aThe IDB stated that U.S.-born adult Hispanics do not send remittances.

bNot applicable.

cThe IDB survey of remitters in the United States found that 61 percent of
all 16.7 million Hispanics (10.2 million) in the United States remit
funds. However, when the U.S.-born adult Hispanics are subtracted from
this population-because IDB officials state that they do not remit-then
10.2 million of the 14.7 million foreign-born Hispanics (70 percent) remit
funds to their home countries.

BEA Is Involved in International Efforts to Improve the Collection and
Reporting of Remittance Information

BEA is involved in international efforts that began in January 2005 to try
and improve upon the collection and reporting of remittance data; however,
it is too early to tell how successful these initiatives will be.
Currently, remittance data are incomplete and cannot be reconciled because
of inconsistency in the various institutions' methods of collecting and
reporting remittance data. Recognizing the importance of remittances and
the need for improved data, the governments of the G8 at the Sea Island
Summit in 2004 called for the establishment of a working group to improve
remittance statistics. BEA is an active member of an international group
supporting this effort, which recommended an agreed upon definition of
remittances. In June 2006, a new group will also start an effort to
improve guidance on collecting and reporting remittance data. BEA expects
to be invited to serve on this group.

International Estimates of Remittances Are Incomplete and Do Not Reconcile

The international estimates of remittances vary by the methods used and
the coverage, quality, and reliability of the data, making comparisons of
such estimates difficult. In principle, the combined inflows and outflows
for all countries should equal zero-as the outflows from one country or
international organization become the inflows of another. However, many
countries do not provide information on both remittance inflows and
outflows, resulting in global remittance figures that do not reconcile.
Figure 4 shows the remittance inflows (credits) and outflows (debits) from
1990 through 2003. If global remittance figures reconciled, the lines in
this figure would be the same. However, as can be seen from the figure,
while the lines were fairly close prior to 1998, since then they have
diverged with countries showing remittance inflows (primarily developing
countries) larger than remittance outflows (primarily developed
countries).

Figure 4: Differences between Global Remittance Credits and Debits,
1990-2004

The IMF accepts member countries' estimates of remittances at their face
value because, according to IMF officials, all methods of estimating
remittances have their weaknesses. According to IMF officials, the choice
of methodology is primarily related to the availability of resources. IMF
officials indicated that they were not aware of any country that has
institutionalized household surveys to generate remittance data.
Remittance estimates submitted by IMF member countries do not reveal the
methodologies used for the estimates. However, according to IMF officials,
most countries report their remittances as residuals of existing data;
others simply do not report remittances.

International Working Group Was Established in 2005 to Improve Remittance
Data

In 2004, at the annual G8 meeting in Sea Island, Georgia, leaders of the
G8 countries recognized the important role remittances play and called
upon international financial institutions such as the World Bank and the
IMF to lead a global effort to improve remittance statistics. As a result,
the World Bank, IMF, and the United Nations formed the International
Working Group on Improving Data on Remittances. This group delegated the
tasks of clarifying concepts and definitions on remittances and addressing
compilation issues to other groups. The working group met in January 2005
and included BEA and representatives from key remittance-sending
countries, one key remittance-receiving country, and the Organization for
Economic Cooperation and Development.

The working group's first objective was to clarify the definition of
remittances. The group agreed that the United Nations Technical Subgroup
on the Movement of Natural Persons (TSG), of which BEA is a member, should
be the forum to discuss improvements in concepts and definitions for
remittances. The TSG recommended, among other things, that the "workers'
remittances" item in the balance of payments be replaced with a new
component called "personal transfers," which would include all current
transfers (in cash or in kind) sent or received by resident households to
or from nonresident households. This new component would not be based on
employment or migration status and would resolve the inconsistencies
associated with "workers' remittances."16 This new definition was
discussed at the June 2005 meeting of the IMF Committee on

Balance of Payments Statistics.17 BEA officials stated that they have
begun using this new definition; however, it will be included in the
publication of the revised Balance of Payments Manual, which is scheduled
to be completed in 2008.

The second objective of the working group was to improve guidance on
collecting and compiling remittance statistics, including the use of
household surveys, if needed. The working group agreed that it would be
useful to form a core group of compilers to review methods and develop
more detailed guidance for compiling remittances data. Eurostat, the
statistical office of the European Communities, offered to host the first
meeting in June 2006 in Luxembourg, thereby creating the "Luxembourg
Group," which includes the World Bank and IMF's statistics department. The
Luxembourg Group will review, among other things, the extent to which
household survey data can be used to improve balance of payment
statistics. BEA expects to be invited to serve on this group. According to
the IMF, the prerequisite to the group's success is the commitment of
national compilers to share their methodologies. The progress of this
group will be reviewed by the IMF Committee on Balance of Payments
Statistics, of which BEA is a member. No date has been set for this group
to complete its work.

In the meantime, the international working group will coordinate with a
recent project conducted by the Center for Latin America Monetary Studies
to improve central bank remittance reporting and procedures. This project
is supported by the MIF. The final report of the working group is to be
presented by the end of September 2006, so that initial work of the
Luxembourg Group can be incorporated.

Observations

In recent years, remittances have received growing attention from policy
makers because major industrial countries began to understand the
magnitude and importance of these flows to developing countries. By their
nature, remittance flows are difficult to measure. Some remittances move
through informal channels that official data often cannot easily or
reliably measure. Countries define remittances differently and use various
methodologies to estimate them; it is therefore not surprising that
estimates vary widely.

Although there are international efforts in which BEA participates to
improve remittance statistics, two issues suggest the challenges facing
these efforts. First, current remittance data are incomplete globally and
cannot be easily reconciled because of the inconsistency in the methods of
collecting and reporting remittance data. Second, for source countries,
remittances constitute a small share of their overall economy-thus there
may not be enough incentives for these countries to improve their
remittance estimates. For recipient countries, remittances constitute a
larger share of the economy; but these countries lack the resources to
improve their statistics. International efforts to improve remittance
statistics have begun recently, and it is too soon to tell whether these
efforts will improve the accuracy of remittance statistics.

In the United States, remittance estimates are important for agencies such
as Treasury and the Federal Reserve; more accurate remittance estimates
could help them better target their financial infrastructure and
automated-clearinghouse remittances programs. With better data on
remittances, the U.S. government could make better decisions about how
much (and what kind) of development assistance to provide, and U.S.
companies could make better decisions regarding foreign direct investment.
As remittance flows from the United States continue to grow, U.S. policy
makers may want to explore options for improving the accuracy of U.S.
remittance statistics-such as conducting a new survey to determine the
remitting behavior of U.S. immigrants, or adding specific questions to
current government surveys to obtain better information.

Agency Comments and Our Evaluation

The Departments of Commerce and the Treasury provided written comments on
the draft report, which are reproduced in appendixes V and VI,
respectively. Commerce also provided technical comments, which we
incorporated into the report as appropriate.

Treasury concurred with our observations, especially on the need for more
accurate remittance data to provide policy makers with the information
necessary to improve their decision-making process. Commerce concurred
with most of our observations. Specifically, they concurred that estimates
of remittances from the United States derived by BEA and those of foreign
governments and international organizations differ substantially and that
there are several methodological reasons for these differences. Commerce
also concurred that more accurate estimates would enable users of
remittance data to make better informed decisions.

Commerce, however, stated its view that BEA's estimates are lower than
most of the others we discuss because we compare BEA's estimate of
personal gifts to foreign residents (personal transfers) with much broader
estimates of remittances, which include compensation paid to foreign
workers who are temporarily employed in the United States. Commerce
believes that a substantial portion of the differences between BEA's
estimates and those of other government or international organizations is
accounted for by this definitional difference. Contrary to Commerce's
view, compensation paid to foreign workers temporarily employed in the
United States was not included in the remittances estimates with which we
compared BEA's personal transfers estimates. We therefore do not believe
that the differences among the estimates we discuss in our report are due
to this definitional difference. Commerce further stated that some
countries may overestimate their receipts of remittances from the United
States because remittances may be channeled through banks in the United
States from remitters not living in the United States. Of the countries we
discuss in this report, we found this only to be true for the Philippines
and, for this reason, we do not compare BEA's remittances estimate to that
of the central bank of the Philippines. As we discuss in our report,
efforts are underway to improve remittance statistics, which may help make
estimates more comparable in the future.

We are sending copies of this report to the Department of Commerce,
Treasury, the Chairman and Ranking Minority Member of the House Committee
on Financial Services, and other interested congressional committees. We
will also make copies available to others on request. In addition, this
report will be available at no cost on our Web site at http://www.gao.gov
.

If you or your staff have any questions regarding this report, please
contact me at (202) 512-2717 or [email protected] . Contact points for our
offices of Congressional Relations and Public Affairs may be found on the
last page of this report. GAO staff who made major contributions to this
report are listed in appendix VII.

Yvonne D. Jones Director, Financial Markets   and Community Investment

Objectives, Scope, and Methodology Appendix I

Our reporting objectives were to examine (1) the methodology the Bureau of
Economic Analysis (BEA) uses to develop the official U.S. estimates on the
volume of remittances from the United States, (2) methodologies used by
other countries and multilateral institutions to estimate remittances from
the United States, and (3) international efforts to improve the collection
and reporting of remittance data.

To understand the methodology BEA used to derive its estimate of
remittances from the United States, we met several times with BEA
officials responsible for developing the estimate. They provided us with
the 2003 estimate on the total volume of remittances from the United
States to the rest of the world-and explained how they provide this number
to the International Monetary Fund (IMF)-so that the U.S. figures can be
presented in the IMF's balance of payments statistics. We also obtained
documentation describing BEA's methodology before 2005, including BEA's
Survey of Current Business and other written documentation. BEA officials
provided us with examples of the various data used in their model to
calculate their remittance estimate. In addition, we provided BEA with
numerous follow-up questions about their methodology, and they provided us
with written responses. To understand BEA's revised methodology, we
obtained relevant documentation from BEA and provided follow-up questions
to BEA. We also met with the U.S. Census Bureau to understand the data
underlying BEA's methodology for estimating remittances. To understand how
we evaluated the statistical reliability of BEA's estimate for 2003, see
appendix III. We interviewed remittance experts from the IMF, World Bank,
Inter-American Development Bank (IDB), and academia to obtain their views
on BEA's (and alternative) methodologies.

To understand the methodologies used by other countries and multilateral
institutions to estimate U.S. remittances to specific countries and
regions, we met with officials from the IDB and their external consultant,
the Asian Development Bank, the African Development Bank, as well as the
Mexican and Philippine Central Banks. The IDB provided remittance
estimates from the United States to specific countries in Latin America,
the Caribbean, and to the region as a whole. The Asian and African
Development Bank do not provide estimates for their respective regions.
The Central Bank of Mexico provided estimates of remittances received by
Mexico from the United States, while the Central Bank of the Philippines
provided estimates of remittances received by the Philippines from the
United States. In meetings with these entities, we obtained an
understanding of the methodologies used to estimate remittances, the
reasons for using these methodologies, and their strengths and potential
limitations. We also obtained a report that described IDB's methodology.
Further, we obtained government regulations from Mexico and the
Philippines to understand what financial institutions are required to
report to Central Banks so that they can estimate remittances. To compare
remittance estimates obtained from the Mexican Central Bank and IDB with
those of BEA, we obtained BEA's 2003 estimates of remittances to specific
countries. BEA officials cautioned us that the estimates to specific
countries are less reliable than their overall remittance estimate and
stated that these numbers should not be considered BEA estimates to
specific countries.

Given our understanding that remittance estimates vary for a number of
reasons and that international efforts are under way to improve remittance
statistics, it was not possible for us to cross check the estimates of
remittances from the United States against any accurate known amount.
Because of this, for the purposes of this report, we focused on
understanding the methodologies used by BEA, IDB, and the Central Banks of
Mexico and the Philippines, to estimate remittance from the United States.
We also focused on understanding the strengths and limitations of the
methodologies of the BEA and the other entities to obtain a better
understanding of the reasonableness of their approaches to estimating
remittances. We presented BEA's estimates and the estimates of IDB and the
Central Bank of Mexico to show the range of estimates generated from
different methodologies, rather than as a statement of their being precise
measurements of remittances. We chose not to present the Central Bank of
the Philippine's estimate of remittances because central bank officials
stated that their current methodology could not be used to report on
remittances solely received from the United States.

To obtain a global perspective on international efforts to improve the
collection and reporting of remittances, we met with officials from the
IMF, World Bank, IDB, Asian Development Bank, African Development Bank,
and experts in the field of remittances. We reviewed IMF documents on
remittances as they are discussed in the balance of payments framework and
reviewed IMF balance of payments statistics to get a sense for which
countries regularly report on remittances. We obtained limited
documentation (e.g., minutes from meetings) on international efforts to
improve the collection and reporting of remittances. BEA and the U.S.
Department of the Treasury (Treasury) also provided us with descriptions
of these international efforts and identified the U.S. government
officials that participate in these international bodies.

Our work was performed in San Francisco, California; and Washington, D.C.,
from December 2004 to March 2006 in accordance with generally accepted
government auditing standards.

BEA's Methodology for Estimating Remittances Appendix II

BEA's model to estimate remittances combines data on the number of the
adult foreign-born population living in the United States, the percentage
of the adult foreign-born population that remits, the income of the adult
foreign-born population, and the percentage of income that is remitted by
the adult foreign-born population.1 BEA first multiplies the foreign-born
population, arrayed by selected demographic characteristics, by the
percentage of the foreign-born population that remits to obtain the
population of remitters. BEA then multiplies the average per capita income
of the foreign-born population by the percentage of income remitted by
those who remit to obtain per capita remittances. Finally, BEA multiplies
per capita remittances by the population of remitters to obtain total
personal transfers.

BEA obtains estimates on the adult foreign-born population by place of
birth and their average income from the American Community Survey (ACS),
arranged by duration of stay in the United States, gender, and presence of
children in the household. BEA obtains estimates of the percentage of the
adult foreign-born population that send remittances to their country of
origin from various academic studies, in addition to the 1989 Legalized
Population Survey (LPS1) and the 1992 Legalized Population Follow-Up
Survey (LPS2); however, the estimates it uses cannot be directly tracked
to these source documents. BEA obtains these proportions by making
assumptions based on its judgment. BEA assumes that the place of birth of
the adult foreign-born population does not affect the likelihood of
remitting but that it does affect the percentage of income remitted. BEA
also assumes that, once the presence of children in the household and the
duration of stay are accounted for, men and women are equally likely to
remit. In effect, only the presence of children in the household and the
duration of stay determines the percentage of the adult foreign-born
population that remit to their countries of birth under these assumptions,
as shown in figure 5.

Figure 5: BEA Values for the Percentage of the Adult Foreign-Born
Population in the United States That Send Remittances

To determine the percentage of income that the adult foreign-born
population remits, BEA makes assumptions about the development status and
proximity of the country of origin of the adult foreign-born population,
along with the presence of children in the U.S. household. BEA groups
countries of origin into four categories indicating their propensity to
send remittances and representing highest-remitting, high-remitting,
medium-remitting, and low-remitting countries of birth. The
highest-remitting countries are closest to the United States, while other
developing countries are either high-remitting or middle-remitting,
depending on their development status. Low-remitting countries are
generally developed economies. Figure 6 shows that the percentage of
income remitted varies by the presence of children and country groupings.
Although average incomes are lower for women than for men, BEA assumes
that the percentage of income remitted does not vary by gender.
Furthermore, BEA assumes the duration of stay is negatively associated
with likelihood to remit-but that it has no effect on the percentage
remitted. Also, BEA assumes that there are no variations in the portion
remitted for countries designated as low remitting.

Figure 6: Percentage of Income Remitted by Category

Table 4 shows the application of BEA's methodology in estimating
remittances from the United States in 2003. As can be seen from table 4,
estimated total remittances are $28 billion. Also, as can be seen in table
4, in 2003, the Latin America and Caribbean region was the largest
recipient region of remittances from the United States. Remittances to
Asia and Africa represented approximately 24 percent and 4 percent of the
total for the United States, respectively.

Table 4: Regional and Sub-Regional Remittances in 2003 as Estimated Using
BEA's Underlying Country-by-Country Tabulations

                                        

        Region or sub-region          Estimate of remittances (in millions of 
                                                                U.S. dollars) 
Africa                                                              $1,003 
North Africa                                                           236 
Africa (other)                                                         768 
Latin America and the Caribbean                                     17,914 
Central America                                                     11,487 
Caribbean                                                            4,360 
South America                                                        2,068 
Europe                                                               2,202 
European Union                                                         550 
Eastern Europe & Transition                                          1,613 
Countries                       
Europe (other)                                                          39 
Asia                                                                 6,616 
Eastern Asia                                                         1,585 
Southern Asia                                                        1,613 
South-Eastern Asia                                                   3,055 
Near East                                                              313 
Asia (other)                                                            49 
Oceania                                                                103 
Australia & New Zealand                                                 26 
Oceania (other)                                                         77 
North America                                                          194 
Total                                                              $28,033 

Source: GAO calculations using BEA underlying country-by-country
tabulations.

Note: BEA officials stated that they would correct the total from $28.2
billion to $28.033 billion.

BEA's Estimate of Remittances Includes Remittances from Some of the
Foreign Born Who Have Been in the United States for Less Than One Year

BEA's revised methodology uses a U.S. residency duration of 0-5 years as
its first category. This means that it includes both the foreign-born
population, who are in the United States for less than or equal to 1 year,
and those who are in the United States for more than a year. The
definition of "remittances" is the portion of income sent as remittances
by those who have resided in the United States for more than one year,
thus excluding the foreign-born population residing in the United States
for less than one year. BEA's estimate of remittances is in effect
overstated, because it

Analysis of the Sensitivity of BEA's Estimate to Judgmentally Determined
Variables on the Remitting Behavior of the Foreign Born Appendix III

BEA publishes single-value estimates of remittances to the rest of the
world by foreign-born U.S. residents. To evaluate the statistical
reliability of the estimate for 2003, we derived the estimate's probable
range and its corresponding breakdown into regional estimates. To
accomplish this, we obtained details of the BEA's underlying tabulations
of remittances by country. We replicated the BEA methodology to obtain
BEA's estimate for the world and for each country in its underlying
tabulation. In particular, we used BEA's underlying tabulation and
included additional information (e.g., the standard deviation and the
shape of the distribution of each data series) from the sources that BEA
primarily used to arrive at its estimate. We calculated the respective
standard deviations of the values that BEA uses for the propensity to
remit and the percentage of the foreign born that remit.

BEA uses a variety of sources to estimate the propensity of the foreign
born to remit and the percentage of the foreign born that remit. However,
BEA stated that the values chosen cannot be linked to any specific source.
BEA primarily used the LPS, a survey mandated by the Immigration Reform
and Control Act of 1986 to estimate the portion of income that the foreign
born in the United States were likely to remit; thus, we also relied on
this data. We assumed that the distribution around the means of the
variables used in the BEA methodology were lognormal to satisfy (1) the
nonnegativity of the values used and (2) a desired bell-shaped
distribution for the estimates. We converted the BEA estimation process
from one that relied solely on the averages of the variables underlying
the BEA methodology to one that accounts for the variation around the mean
and its distribution. We used a Monte Carlo statistical technique-a
technique that repeatedly and randomly samples from the underlying data-to
obtain a range of possible values for each estimate due to the uncertainty
in BEA's judgmentally determined variables on the foreign born
propensities to remit and percentage of the foreign born that remit.

Table 6 shows the regional breakdown of BEA's 2003 estimate and the
statistically derived range for these estimates. In table 6, we show in
the column labeled "BEA point estimate"-the regional components of BEA's
global estimate in 2003-obtained by aggregating the underlying
country-by-country tabulations. We also show in the following two columns
the range of estimates obtained by our uncertainty analysis, assuming that
this uncertainty is only due to BEA's judgmentally determined variables.
In table 6, BEA reported $28 billion in total remittances from the United
States for 2003; however, we estimate that the range for 90 percent of the

IDB Remittance Estimation Methodology Appendix IV

To estimate remittances from the United States to Latin America in 2003,
the IDB contracted researchers to survey Latin Americans aged 18 years or
older and living in the United States. These researchers queried Latin
American immigrants living in various states of the United States about
their remittance experiences. The survey interviewed 3,802 households in
37 states and the District of Columbia from January through April 2004.1
The survey showed that 61 percent of Latin Americans send remittances to
their countries of origin, sending an average of $240 approximately 12.6
times per year. IDB extrapolated the results of the survey to the total
population of adult Latin American immigrants in the United
States-estimated at 16.9 million in 2003-and estimated remittances from
the United States to Latin America to be $30.1 billion for that year.
Figure 7 provides a diagram of the methodology IDB used to arrive at the
$30.1 billion estimate. According to IDB, the estimate captured remittance
flows through the formal and informal sectors. The IDB also used the
survey to estimate remittances from each of the 37 states and the District
of Columbia. To obtain the state-by-state remittance estimates, the IDB
obtained estimates for the average amount remitted and the number of times
sent in one year by the Latin American immigrant population in each state
and the percentage of the Latin American immigrant population in each
state that sends remittances.

Figure 7: IDB Methodology for Estimating Remittances from the United
States to Latin America, 2003

The IDB remittance estimates for selected Latin American and Caribbean
countries are obtained from a combination of sources consisting of
estimates from selected central banks of recipient member countries judged
to have reasonable remittance estimates, transaction information from
remittance transfer companies to selected countries, and from information
obtained from researchers' surveys of remittance senders in the United
States and remittance recipients in Latin American and Caribbean
countries. According to IDB officials, for countries where no in-country
survey has been conducted, data from establishments facilitating money
transfers to each country was used. These officials indicated that data
were obtained from a sample of 45 money transfer businesses to
approximately 14 countries. The amount and frequency of the average
remittance sent by residents from the survey countries was used to
estimate the total remittance outflow to each country, according to IDB
officials. They also indicated that Multilateral Investment Fund (MIF)
staff work with the researchers to reconcile the various estimates and
arrive at country-specific estimates. Table 7 shows the IDB estimate of
remittances that 21 Latin American and Caribbean countries received in
total in 2003, and from the United States the same year.

Table 7: IDB/MIF Estimates of Remittances to Latin American and Caribbean
Countries, 2003

                                        

        Country      IDB/MIF estimate IDB/MIF estimate   Remittances from the 
                            for total  for remittances   U.S. as a percentage 
                          remittances    from the U.S.           of the total 
                                                       
                      (in millions of  (in millions of 
                        U.S. dollars)    U.S. dollars) 
Argentina                     $225             $180                    80% 
Belize                          73               58                     79 
Bolivia                        340              240                     71 
Brazil                       5,200            2,600                     50 
Colombia                     3,067            2,147                     70 
Cost Rica                      306              245                     80 
Dominican Rep.               2,217            1,773                     80 
Ecuador                      1,656              994                     60 
El Salvador                  2,316            2,085                     90 
Guatemala                    2,106            1,685                     80 
Guyana                         137              109                     80 
Haiti                          977              879                     90 
Honduras                       862              775                     90 
Jamaica                      1,425            1,069                     75 
Mexico                      13,266           12,868                     97 
Nicaragua                      788              709                     90 
Panama                         220              176                     80 
Peru                         1,295              777                     60 
Trinidad & Tobago               88               71                     81 
Uruguay                         42               29                     69 
Venezuela                      247              173                     70 
Sub-Total                   36,853           29,642                     80 
Rest of Latin                1,240              992                     80 
America and                                         
Caribbean                                           
countries                                           
Latin America and          $38,093          $30,634                     80 
Caribbean Total                                     

Source: IDB/MIF, Sending Money Home: Remittances to Latin America and the
Caribbean, Washington, D.C.: (May 2004).

As indicated earlier, the IDB and BEA used different methodologies to
estimate remittances, resulting in a range of estimates. While, in most
cases, BEA provides only a global estimate of remittances and not
bilateral estimates, BEA provided us with country-by-country tabulations
that enabled us to construct estimates for the same 21 countries that IDB
provided estimates for in 2003. As shown in table 8, IDB and BEA's
estimates vary; IDB's estimates in general tend to be higher than
estimates from BEA's underlying country tables. However, for Guyana,
Panama, and Trinidad and Tobago, BEA's estimates are higher. The last
column computes the difference between the estimates for each country as a
percentage of the average of the estimates.2 The average percentage
difference is 72 percent, with a low of 7 percent for Jamaica and a high
of 168 percent for Brazil.

Table 8: Percentage Difference between BEA and IDB Estimates of
Remittances from the United States to Selected Latin American and
Caribbean Countries, 2003

                                        

    Millions of                                       
      dollars                                         
                   IDB estimate of Estimates from BEA   Percentage difference 
                  remittances from underlying country     between IDB and BEA 
                          the U.S.             tables               estimates 
Argentina                  $180                $99                     58% 
Belize                       58                 43                      30 
Bolivia                     240                 98                      84 
Brazil                    2,600                223                     168 
Colombia                  2,147                740                      98 
Cost Rica                   245                 99                      85 
Dominican Rep.            1,773                700                      87 
Ecuador                     994                478                      70 
El Salvador               2,085              1,013                      69 
Guatemala                 1,685                611                      94 
Guyana                      109                255                      80 
Haiti                       879                630                      33 
Honduras                    775                308                      86 
Jamaica                   1,069                992                       7 
Mexico                   12,868              8,905                      36 
Nicaragua                   709                290                      84 
Panama                      176                217                      21 
Peru                        777                290                      91 
Trinidad &                   71                205                      97 
Tobago                                             
Uruguay                      29                 22                      28 
Venezuela                   173                 61                      96 

Source: GAO analysis of information provided by IDB/MIF and BEA.

Comments from the Department of Commerce Appendix V

The following are GAO's comments on the Department of Commerce's March 10,
2006, letter.

GAO's Comments

1.BEA commented on the Highlights page that the IDB estimates differ from
BEA's estimates because the IDB estimate includes "net compensation" of
foreign workers and the BEA estimate does not. BEA also commented that
data provided by foreign central banks and financial establishments are
sometimes overstated because U.S. correspondent banks are used in
transmitting funds for senders not living in the United States. We
disagree with BEA on these points. This "net compensation" of foreign
workers is a new concept that was just proposed by the Technical Subgroup
on the Movement of Natural Persons (TSG) in June 2005, and we are not
aware of any remittances estimates for 2003 that use this definition.
Further, IDB never stated that any of the funds accounted for in their
estimates came through U.S. correspondent banks for workers who were not
located in the United States. This was true for the Philippines, which we
noted in the report. BEA also commented that IDB's estimates are
substantially derived from data reported from central banks and private
money transfer establishments. BEA is correct on the latter point and we
have corrected the Highlights page to be consistent with the letter and
reflect that IDB uses a variety of sources in making its remittances
estimates.

2.BEA suggested that we place Mexico in North America or create a separate
bar in our graphic in the Highlights page for Mexico. In this report, we
used the United Nations' Standard Country and Area Codes Classification,
which places Mexico in Central America.

3.BEA commented that to develop an estimate that corresponds to our
definition of remittances, we should have used BEA's estimates of personal
transfers and compensation of employees, net of their expenditures.
However, we make it clear in footnote 6 that we are focusing only on
personal transfers and that we call these remittances for the purpose of
this report.

4.BEA states that it has confirmed with the Bank of Mexico that Mexico's
estimates of remittances include net compensation of migrant Mexican
workers in the United States. BEA states that if we added BEA's net
compensation of employees figure to its estimate of personal transfers,
the two figures for 2003 would be closer. As stated above, this new
definition was proposed in June 2005, and, to our knowledge, the Mexican
central bank has not published 2003 figures for "net compensation" of
employees. The Mexican central bank figures for 2003 as reported by the
IMF in its balance of payments statistics are almost $13.4 billion for
workers' remittances, which we use in our report, and $1.5 billion in
compensation of employees. The $12.9 billion estimate BEA attributes in
its comments to the Mexican central bank is the IDB's estimate.

5.BEA commented that the data used in our analysis of the potential
effects of BEA's judgmentally determined values in its remittance
estimating methodology are unclear, as are the particulars of our modeling
technique. As we stated, we replicated BEA's methodology using its
underlying tabulation of remittances by country and included additional
information from the sources that BEA primarily used to arrive at its
estimate. BEA further stated that there is a very small probability that
the BEA estimate would be near the end points of the intervals and
suggested that we use the midpoint of the intervals instead. As explained
in appendix III, the purpose of our analysis was to show the effect of
BEA's judgmentally determined values on its estimate $28.03 billion in
2003. Using a range illustrates the uncertainty in BEA's estimate. BEA
also commented on our use of the lognormal distribution for the percentage
of income remitted and the percentage of the adult foreign born population
that remit. We chose the lognormal distribution because it satisfied the
requirements that both of these variables were nonnegative and distributed
in a bell-shaped curve.

6.BEA commented that we left the impression that BEA's estimates of
personal transfers contain a double count of $377 million and that any
double count that may exist probably involves the compensation of
employees, not the personal transfers account. We modified the text of our
report to reflect that BEA's personal transfers are therefore potentially
overstated by up to $377 million because BEA's estimate includes
remittances sent by some of the foreign born who have been in the United
States for less than one year.

7.Commerce reiterated its concerns about our comparison between BEA's
estimates and those of other organizations. Commerce restated its view
that the methods used by the Mexican central bank and others capture both
remittances and compensation of employees and further stated that BEA's
estimates for personal transfers and compensation of employees should be
summed when making these comparisons to other organizations. However, none
of the organizations with which we compare BEA's estimates indicated that
their methods captured compensation of employees, therefore, we believe
our comparisons are appropriate.

8.BEA states that the TSG now recommends that "personal transfers" also
include capital transfers. This is incorrect. The paper BOPCOM-05/9 states
that the TSG agreed to define "personal transfers" as consisting of all
current transfers in cash or in kind.

9.BEA disagreed with our statement that remittance data cannot be
reconciled and stated that, because reconciliation projects are resource
intensive and difficult, BEA must choose the statistical items it
reconciles with which trading partners. We concur that reconciliation
cannot be done easily. However, our observations were on reconciliation of
remittance data on a global level, not between individual countries, as
shown in figure 4. The global discrepancy has grown in recent years.

Comments from the Department of the Treasury Appendix VI

GAO Contact and Staff Acknowledgments Appendix VII

Yvonne D. Jones (202) 512-2717 or [email protected]

In addition to the contact named above, Barbara I. Keller, Assistant
Director; Gezu Bekele; Tania Calhoun; Lynn Cothern; William R. Chatlos;
Bruce L. Kutnick; James M. McDermott; Marc M. Molino; Jose R. Pena; and
Rachel Seid made key contributions to this report.

(250220)

www.gao.gov/cgi-bin/getrpt? GAO-06-210 .

To view the full product, including the scope

and methodology, click on the link above.

For more information, contact Yvonne Jones at (202) 512-2717 or
[email protected].

Highlights of GAO-06-210 , a report to Committee on Banking, Housing, and
Urban Affairs, U.S. Senate

March 2006

INTERNATIONAL REMITTANCES

Different Estimation Methodologies Produce Different Results

Remittances are the personal funds that the foreign born send to their
home countries. In recent years, estimated remittances have grown
dramatically, and policy makers have increased their attention to these
flows. Organizations use various methodologies to estimate remittance
flows, which result in a range of estimates. In 2004, the Group of Eight
(G8) leaders emphasized the need for improved statistical data on
remittances.

In light of the growing volume of remittances and the differences in
estimates, GAO examined (1) the methodology that the Bureau of Economic
Analysis (BEA) uses to develop the official U.S. estimate, (2)
methodologies that other countries and multilateral organizations use to
estimate remittances, and (3) international efforts to improve the
collection and reporting of remittance data.

What GAO Recommends

While GAO makes no recommendations at this time, GAO observes estimates of
the amount of remittances from the United States differ. More accurate
remittance estimates could help certain U.S. agencies make better
decisions. Therefore, policy makers may want to consider exploring options
for improving the accuracy of U.S. remittance statistics. We received
written comments on a draft of this report from the Departments of the
Treasury and Commerce. They both generally agreed with our observations.

BEA uses a model to estimate remittances from the United States and,
although the methodology has some strengths, the accuracy of BEA's
estimate is uncertain for several reasons. BEA estimated remittances for
2003 at $28.2 billion; its model used data on the number of foreign-born
residents, their income, the proportion of income that is remitted, and
other demographic data. The strengths of BEA's methodology are that, in
theory, it estimates remittances sent through formal and informal
channels. It also is low-cost because it uses existing data on the foreign
born. However, BEA's methodology was limited by the quality and timeliness
of the data, particularly on the portion of income likely to be remitted.
BEA revised its model in 2005 to use new data sources, but the accuracy of
its estimates depends on the accuracy of its assumptions regarding the
remitting behavior of the foreign born and other factors.

Some central banks and the Inter-American Development Bank (IDB) use
different methodologies to provide estimates of remittances from the
United States that vary significantly. For example, Mexico's central bank
estimates remittances primarily by collecting data from money
transmitters. The IDB used a variety of sources, such as surveys of
remittance senders and receivers, and information from remittance transfer
companies and central banks, to estimate remittances from the United
States to Latin America to be $30.6 billion in 2003. We aggregated BEA's
data to estimate remittances to this region to be $17.9 billion.

BEA is an active participant in recent international efforts to improve
remittance statistics. The World Bank and others established a remittances
working group in 2005, which delegated tasks to other international groups
to (1) clarify the definition of remittances and (2) provide guidance on
how to collect and estimate remittances. BEA participated in the first
group, which recommended a new definition of remittances. The second group
will have its first meeting in June 2006.

Regional Destination of Remittances Sent from the United States, 2003

GAO's Mission

The Government Accountability Office, the audit, evaluation and
investigative arm of Congress, exists to support Congress in meeting its
constitutional responsibilities and to help improve the performance and
accountability of the federal government for the American people. GAO
examines the use of public funds; evaluates federal programs and policies;
and provides analyses, recommendations, and other assistance to help
Congress make informed oversight, policy, and funding decisions. GAO's
commitment to good government is reflected in its core values of
accountability, integrity, and reliability.

Obtaining Copies of GAO Reports and Testimony

The fastest and easiest way to obtain copies of GAO documents at no cost
is through GAO's Web site ( www.gao.gov ). Each weekday, GAO posts newly
released reports, testimony, and correspondence on its Web site. To have
GAO e-mail you a list of newly posted products every afternoon, go to
www.gao.gov and select "Subscribe to Updates."

Order by Mail or Phone

The first copy of each printed report is free. Additional copies are $2
each. A check or money order should be made out to the Superintendent of
Documents. GAO also accepts VISA and Mastercard. Orders for 100 or more
copies mailed to a single address are discounted 25 percent. Orders should
be sent to:

U.S. Government Accountability Office 441 G Street NW, Room LM Washington,
D.C. 20548

To order by Phone: Voice: (202) 512-6000 TDD: (202) 512-2537 Fax: (202)
512-6061

To Report Fraud, Waste, and Abuse in Federal Programs

Contact:

Web site: www.gao.gov/fraudnet/fraudnet.htm E-mail: [email protected]
Automated answering system: (800) 424-5454 or (202) 512-7470

Congressional Relations

Gloria Jarmon, Managing Director, [email protected] (202) 512-4400 U.S.
Government Accountability Office, 441 G Street NW, Room 7125 Washington,
D.C. 20548

Public Affairs

Paul Anderson, Managing Director, [email protected] (202) 512-4800 U.S.
Government Accountability Office, 441 G Street NW, Room 7149 Washington,
D.C. 20548
*** End of document. ***