3
1
REMITTANCES AND MACROECONOMIC VARIABLES ON THE
DECLINE OF UNEMPLOYMENT RATE IN THE WORLD
INTRODUCTION
Unemployment is one of the most crucial problems in the world. When a person is
willing to work and able to look for a job, by which they can fulfill their daily needs and
other expenses in life, but they do not reach the point of getting a job or income, this is
known as unemployment. Every citizen will try to find employment opportunities, and it is
the responsibility of the government to provide jobs, including for workers with low
education.
Unemployment in a country indicates low economic growth and low economic
performance. Unemployment also indicates that existing resources cannot be fully utilized so
that the economy is working below its full capacity. According to the Bureau of Labor
Statistics, unemployed workers deprive their families of income. If unemployment is
widespread, it can have an impact on a nation, where a nation will lose its participation in the
economy to produce goods and services. In other words, unemployment not only affects
people who are not working, but also affects those who have jobs and can even broadly affect
a country's economy (Siddiqa 2021).
Unemployment is not only a problem for developing countries but also a problem in
developed countries, but it does have a more severe impact on developing countries.
According to Siddiqa (2021) the negative impact of unemployment in developing countries
causes social isolation, such as psychological problems, loss of identity and self-esteem,
increased stress from the family, and can result in social pressure that can lead to criminal
acts (such as theft, robbery, and so on), so that in overcoming this problem developing
countries must concentrate on the problem of unemployment in order to create jobs for the
people of their country in such a way that the economy can grow and people can also take
advantage of these opportunities to reduce unemployment.
Maqbool et al. (2013) said that the increasing population of a country can be a problem
where it can cause many socio-economic problems, especially in developing countries. The
increase in population not only increases unemployment, but can also hoard unemployment.
The increase in a country's population can make competition for jobs higher. This causes
employment opportunities in a country to be low. Often, the lack of jobs causes job seekers to
2
migrate to other countries to find work.
Based on the 2020 world migration report, the total number of world migration in 2020
is 281 million people or 4.98% of the total population in the world. Although the total
migration in the world is relatively low in percentage terms, which is only 4.98%, but in
migration and population figures this is high. The trend of migration itself has also continued
to increase every year since 1990, where the number of people migrating to other countries has
increased. migration or the number of people living in a country other than their country of
birth in 2020 increased by 128 million compared to 1990, and is more than three times the
amount of world migration in 1970 (IOM 2020). This can be seen in detail in Table 1.
The reasons why people around the world migrate vary. One of them is economic
factors such as to improve their quality of life. (Wodon et al. 2003) said that migration can
increase individual income up to 20-25% of the initial income. This is in accordance with the
IOM 2020 report where most people who migrate their main goal is the employment factor.
Based on IOM 2020 migration data, 74% of the total international migrants are those who are
in the labor force age, namely in the age range of 15-64 years.
The activity of remittances transferred to families by migrant workers who are paid
outside their home countries is known as remittances. Remittances have now become one of
the sources of foreign income (foreign capital inflows) in addition to portfolio investment,
foreign direct investment (FDI) and Official Development Assistance (ODA) that can help
overcome unemployment problems. According to Maimbo and Ratha (2003) in developing
economies, remittances are one of the country's sources of inflows which are often larger than
ODA and also more stable than private capital flows. This is corroborated by data from the
World Bank (2020) which shows that the amount of remittances in the world continues to
increase from year to year which can be seen in Figure 1. Based on this figure, the volume of
remittances in 2020 reached 651.05 billion USD where this figure was 17.10% greater than
the volume in 2015, which amounted to 556.02 billion USD, so that for some developing
countries, foreign remittances are a vital source of foreign exchange, exceeding income from
main exports, and covering most imports (Adams and Page 2005).
Remittances that contribute more and are more stable than other capital inflows such as
portfolio investment, FDI, and ODA are common in many countries, especially in many
developing countries. This makes migrant workers dubbed as foreign exchange heroes for the
country, because the remittances they send into the country can increase the country's
income. Figure 2 shows a comparison of the movement of several foreign capital inflows
3
including remittances, FDI, ODA, and portfolio investment. Based on Figure 2, it can be seen
that the amount of remittance flows in the world has increased rapidly and has been stable for
a long time. However, FDI and portfolio investment have fluctuated quite a bit over the past
two decades, so it can be said that the development of FDI and portfolio investment is not
quite as stable as remittances. In addition, ODA, which is foreign official development
assistance, has been quite stable from year to year but the volume provided has not been very
high and is still below the value of remittances, portfolio investment, and ODA.
The increasing and stable volume of world remittances indicates that there is currently a
lot of migration in the world. This can be due to the limited number of job vacancies
compared to the proportion of the labor force in the country of origin, so it can be said that
migration is carried out as a form of effort to reduce the level of unemployment that occurs in
the country. In addition to reducing the unemployment rate, the number of migrations that
occur can also be due to the difference in wages between the country of origin and the
country that is the destination of migration. The difference in currency value (exchange rate)
between the country of origin and the destination country can make the wages of migrant
workers much higher than in the country, where this factor can be one of the reasons job
seekers work abroad as migrant workers.
The previous explanation allows migrant workers who work abroad to provide benefits
or assistance to their families or relatives in the country in order to improve the standard of
living of their families and relatives in the country. In other words, migrant workers are often
referred to as foreign exchange heroes for a country. If many migrants make remittances to a
country, this can have a positive impact on the country's economy. According to Azam et al.
(2016) international aid (remittances) is one of the efforts that can be made to stimulate
economic development and reduce unemployment. Even remittances sent by migrants can
reduce the level and severity of poverty in developing countries (Simionescu and Dumitrescu
2017).
Figure 3 shows the development of unemployment rate movements in the world, as
well as in high-income countries and low- and middle-income countries. The unemployment
rate can proxy for the poverty rate. If the unemployment rate is high, it is likely that the
poverty rate is also high. Based on the table, the percentage of unemployment in the world
has fluctuated over the past decade. High-income countries as well as low- and middle-
income countries also experience movements in unemployment rates that tend to fluctuate. In
2020 there was an increase in the percentage of unemployment rates in the world, both in
high-income countries and low and middle-income countries which is the effect of COVID-
4
Todaro and Smith (2010) reveal that prosperity and welfare will not be achieved if most
of the population is still unemployed and in poverty and misery. The World Bank report on
poverty states that the right policies and measures can reduce the unemployment rate, so that
the poor can have an income and can later participate and contribute to economic growth. If
this can be achieved, rapid unemployment and poverty reduction will be consistent with
sustainable economic growth. Poverty alleviation is a key goal of the 17 Sustainable
Development Goals (SDGs), a world without poverty and an end to poverty in all its forms
everywhere. In addition to poverty alleviation, unemployment, in this case getting decent
work, is also included in the 17 pillars of the SDGs, which is in the eighth position in the
goal. This statement can be a strong reason why poverty alleviation and unemployment
reduction must be considered in order to realize the country's development.
1.1 Problem Formulation
A relatively high level of unemployment does not allow a society to achieve robust
economic growth. This can be clearly seen from the various negative economic impacts
caused by unemployment. The most important of these negative impacts is the increase in
poverty. Poverty is an unavoidable issue when discussing the economy. Although there has
been a significant improvement in poverty in the world over the last half century, poverty
cannot be completely eliminated and is still common, especially in developing countries
(Todaro and Smith 2010).
When the unemployment rate increases, the economy will not produce maximum
output, so efforts are needed to reduce unemployment. One of the efforts that can be made is
by creating jobs. However, the lack of jobs in the country sometimes requires a country to
export domestic labor services abroad, where the wages that will be received by these migrant
workers are known as remittances.
Remittances are currently receiving special attention, because remittances are
considered to be quite significant in dealing with the issue of unemployment. Adams and
Page (2005) show that remittances can significantly reduce unemployment rates in various
developing countries. This is also corroborated by the data summarized in Figure 2, where it
can be seen that remittances show a fairly large volume and a fairly stable trend.
A comparison of the development of remittance receipts and unemployment rates in the
world during the period 2000-2021 can be seen in Figure 4. Based on this figure, the graph of
remittances and unemployment moves in the opposite direction. This means that it is true that
the receipt of remittances obtained by a country can reduce the unemployment rate in a
5
country at a certain time. However, there are also some points that are inconsistent and
inappropriate, which means that the remittances received can increase the
u n e m p l o y m e n t rate. However, this is possible because there are many other aspects
that can affect the unemployment rate. In the long run, remittances received by a country can
not only reduce the unemployment rate but also have an impact on poverty alleviation and
welfare improvement. The movement of remittances and the unemployment rate in more
detail can be seen in Figure.
Remittances are transfers of funds made by migrants to their relatives in their home
country. Migrants generally make a lot of remittances. This can be seen in Figure 1 where the
trend of remittance development continues to increase from year to year with a promising
volume. The large and promising volume of remittances can be seen in Figure 2 where it can
be seen that among the four foreign capital inflows, the development of remittances is stable
and also has a large volume, and when viewed from the trend, it is estimated that the volume
will continue to increase in the future.
The growing value of remittances around the world indicates that the world's
population is increasingly migrating. International migration is one of the solutions to reduce
the gap between the growth of the labor force and the availability of employment
opportunities, especially in developing countries. Another goal of migrating in addition to
getting a job is of course expected to increase income for workers who migrate, so as to
increase individual and state income in the long run and reduce unemployment rates and
increase welfare.
The data shows that remittance receipts that continue to increase from year to year are
also in line with the decline in unemployment rates in the world, despite the decline in
remittances and the increase in unemployment in 2020 which is the effect of COVID-19.
Adams and Page (2005) provide significant evidence of the role of international remittances
in reducing unemployment in developing countries. Several other studies, namely, Vacaflores
(2017) in 17 Latin American countries, Wagle and Devkota (2018) in Nepal, and Akobeng
(2015) in Sub-Saharan Africa found that international remittances have a significant negative
effect or are proven to reduce unemployment in developing countries the amount of
unemployment. Meanwhile, Imai et al. (2014) found that remittances have a significant
positive influence on unemployment in 24 countries in Asia. This indicates that the higher the
remittances, the higher the unemployment rate. Abdi (2016) found that remittances did not
significantly affect the decrease of unemployment in 15 provinces in Indonesia.
6
The relationship between remittances and unemployment is still uncertain. This
research is conducted to obtain a clear conclusion on the analysis of remittances and other
macroeconomic variables on their impact on reducing the unemployment rate with a wider
cross section of data and a longer and more recent time series. In contrast to previous studies,
it appears that there is no research related to remittances and unemployment that takes the
scope of the country and compares them. In this study, in addition to using the scope of
countries, namely 65 countries in the world; these countries will also be grouped based on the
level of remittance receipts into three categories, namely low remittance countries (0%-3%),
medium (>3%-10%), and high (>10%). Based on the description above, the problems that
will be discussed are as follows.
1. What are the dynamics of remittance receipts, unemployment rates, and
macroeconomic variables in the world?
2. What is the short-run and long-run effect of remittance receipts on the unemployment
rate in the world and are there differences in the results between countries with low,
medium and high remittance levels?
3. How do remittances and other economic variables affect the convergence of
unemployment rates in the world?
Unemployment:
Unemployment is a term used to refer to people who are not working at all, or are
looking for work. Unemployment is generally caused because the number of labor force or
job seekers is not proportional to the number of jobs available. According to the Central
Bureau of Statistics (BPS) in labor indicators, unemployment is a population that is not
working but is looking for work or is preparing a new business or a population that is not
looking for work because it has been accepted for work but has not yet started working.
Unemployment is a macroeconomic problem that directly affects human survival. For
most people the loss of a job represents a decline in living standards. The topic of
unemployment is still a hot topic that is often discussed in political debates by politicians
who often assess that the policies they offer will help create jobs (Mankiw 2000).
Unemployment is a term that refers to individuals who are employable and actively
seeking work but are unable to find a job. Also included in this group are people in the labor
force who are working but do not have a suitable job. Examples include not meeting the
minimum number of hours spent working in a week, or wages earned not meeting the
minimum standard to be considered employed. Usually measured by the unemployment rate,
7
which divides the number of unemployed people by the total labor force, unemployment
serves as an indicator of a country's economic status.
a. Types of Unemployment
According to Sukirno (1994), unemployment based on the causes of unemployment can be
divided into four groups, namely:
1. Normal or Frictional Unemployment
If there is unemployment in an economy of only 2% or 3% of the total workforce, the
economy is considered to have reached full employment. The 2% or 3% unemployment is
called normal unemployment or frictional unemployment. This means that these unemployed
people are unemployed not because they cannot find a job, but because they are looking for a
better job. In a country with a thriving economy, the unemployment rate is low and jobs are easy
to obtain, so workers in this unemployment context have minimal worries. On the other hand,
employers find it difficult to obtain labor, so they offer higher salaries. This will encourage
workers to leave their old job and look for a new job that pays a higher salary or is more in
line with their skills. In the process of looking for a new job, these workers are temporarily
referred to as normal unemployment or frictional unemployment.
2. Cyclical Unemployment
The economy does not always develop steadily. There are times when aggregate
demand is higher, and this encourages employers to increase production. More workers are
hired and unemployment is reduced. At other times, however, aggregate demand may decline
drastically. For example, in countries that produce agricultural raw materials, this decline
may be due to falling commodity prices. This downturn has an effect on other related firms,
which will also experience a decline in the demand for their products. This reduction in
aggregate demand results in firms reducing their workforce or closing down, resulting in an
increase in unemployment. Unemployment with this description is called cyclical
unemployment.
3. Structural Unemployment
Not all industries and companies in the economy will continue to thrive, but there will
be some sectors that experience setbacks. These setbacks can be caused by the following
factors: The emergence of new and better goods, technological advances that reduce the
demand for a good, expenses that have become so high that they cannot compete, and exports
of industrial production that have greatly decreased due to more serious competition from
other countries. These things will cause production activities in an industry to decline, and
8
some workers will be forced to be dismissed so that they become unemployed. Such
unemployment is called structural unemployment because it is caused by changes in the
structure of economic activity as described earlier.
4. Technology Unemployment:
Unemployment can also be caused by the replacement of human labor by technological
advances, such as the loss of jobs due to new machines. An example is workers on the
expressway (toll road) where it is known that currently at each toll gate there are no workers
but only automatic machines, so the workers who originally worked there lost their jobs.
Another recent example in Indonesia is self-service gas stations at several state-owned gas
stations in the capital city. Where this gas station has no workers and requires gasoline buyers
to serve themselves in purchasing gasoline which is done on monitors or sophisticated
machines in the gas station. An example in agriculture, namely the poison of weeds and
grass, where this has reduced the use of labor to clean plantations, rice fields and other
agricultural land. In some companies or factories there are also robots that have replaced
work done by humans. Unemployment caused by the use of machines and technological
advances is called technological unemployment.
b. Measurement of Unemployment:
Not everyone who is not working is called unemployed. To be considered unemployed
according to the statistical agency, a person must not only be out of work but also actively
looking for work, for example by sending resumes of job applications to various companies
or agencies. The labor force includes those who have jobs and those who are looking for
work. The unemployment rate is the percentage of the labor force that is looking for work.
The labor force is only a fraction of the population. The ratio of the labor force to the
working-age population is called the labor force participation rate.
People who are of working age (15-64 years) but are not working and not looking for
work, such as students and housewives are not included in the labor force. The labor force also
excludes people who are unemployed and have been unable to find work for so long that they
have stopped looking. Such discouraged workers are one reason why unemployment statistics
can understate the true demand for jobs in an economy, this is known as disguised
unemployment. Another form of disguised unemployment is those who are working but the
time spent working is less than normal working hours or less than 35 hours a week.
9
Remittances:
Remittances are remittances of part of the income received by workers who work
abroad from the place where they work to their home country. Salama (2004) states that the
effect of international migration on household welfare is characterized by the sending of
remittances by migrants. The remittances received can be used to improve the quality of life
of the recipient, including to repair the house, pay for children's education, pay for family
health, for business capital, and so on. In addition, remittances are considered a major
component in preserving ties with the area of origin because they can improve the standard of
living of the area of origin (Chami et al. 2003).
The flow of remittances stems from workers choosing to work abroad or migrate.
Migration is a change in the place of residence of individuals either permanently or semi-
permanently, and there is no distance limit for the change of residence. One of the reasons for
migration is due to differences in the area of origin and destination of migration. This
difference can be caused by economic, social, and environmental factors. However, several
studies have shown that the main factor determining migration is economic, namely to obtain
employment and higher income. In other words, the purpose of migration is a form of effort
to improve the quality of life of families who migrate.
The conditions described above are in line with Todaro and Smith's (2010) migration
model, which states that migration flows take place in response to income differences
between origin and destination areas. However, the income in question in Todaro's migration
theory is not actual income but expected income. Based on Todaro's migration model,
migrants consider and compare the labor markets available to job seekers in the origin and
destination regions. After that they can choose the one that is considered to maximize their
expected gains.
Migration is a means for the inflow of remittance receipts. Remittance flows are a very
important source of external finance for countries, especially developing countries. The
Ministry of Finance states that remittances are an alternative source of foreign exchange used
as external financing, apart from government loans and private investment. This remittance
inflow, as recorded in the current account of each country's balance of payments, will
increase foreign exchange reserves. The increase in foreign exchange reserves indicates that
the supply of foreign exchange is also increasing and in turn will affect a country's exchange
rate against foreign exchange. This in turn will have an impact on reducing unemployment,
economic growth and poverty alleviation.
10
Economic Growth:
Every country in the world is very concerned about the progress of its country's
economic growth. Economists in the world focus on the rules for increasing income with the
aim of increasing economic growth. This is because economic growth has been believed to be
a measure of the value of national economic development (Todaro and Smith 2010).
Economic growth shows the growth of production of goods and services in an economic
region in a certain time interval. Economic growth is measured based on the growth rate of
national income or Gross Domestic Product (GDP). The calculation of the GDP rate is
determined by the following formula:
Real Interest Rate
There are various types of interest rates including lending rates and savings rates. In
addition, interest rates can also be categorized based on inflation. Based on inflation, interest
rates are divided into two, namely nominal interest rates and real interest rates. Nominal
interest rates are interest rates that do not include inflation as a determining factor in the
decline in purchasing power. While the real interest rate is a pure interest rate that has
included inflation as a determining factor in the decline in purchasing power or can be said to
be a corrected interest rate. This interest rate displays the net return that will be obtained by
customers or economic actors after deducting inflation.
Levinson (2006) states that for example, if an investor can ensure an interest rate of
5% for the coming year and anticipates a price increase of 2%, then they would expect to earn
a real interest rate of 3%. This can be explained more formally by the Fisher equation, which
states that the real interest rate is close to the nominal interest rate minus the inflation rate.
Broadly speaking, the relationship between nominal interest rates, inflation and real interest
rates can be seen in the following Fisher Equation:
Trade Openness:
International trade theory states that with an abundance of resources, each region that
has a comparative advantage will specialize in producing commodities that are relatively
cheap so that they can compete in domestic and international markets. The opposite condition,
namely when there are limited resources, the need for the desired commodity at a relatively
cheap price can be met through import activities, where this is due to the comparative
advantage of other regions or countries. Import-export activities are considered to be able to
11
improve the welfare of the population of a region or country because consumers will get the
desired commodities at a reasonable price relatively cheap, due to specialization between
regions or countries (Salvatore 1997).
Regional economic integration through the formation of regional free trade blocs has
implications for the welfare of member countries, namely positive effects in the form of trade
creation and negative effects due to trade diversion. The change in welfare level is
determined by how much trade creation and trade diversion occur. If creation is greater than
trade diversion, then welfare increases and vice versa. Trade creation is a situation where a
free trade agreement (FTA) can create unprecedented trade between members. With trade
creation, an FTA member country will obtain goods that are produced more efficiently from
other FTA member countries (Krugman and Obstfeld 2005).
Based on the theory of international trade, the greater the capacity of international
trade, the lower the price of goods or services needed by consumers. Indirectly, wages will be
lower. In other words, companies or job providers can attract more workers and the
unemployment rate will decrease, thus indirectly the relationship between the level of trade
openness and the unemployment rate is predicted to be inversely proportional or have a
negative relationship.
Inflation:
Inflation is a symptom of rising price levels at the aggregate level in the economy on a
continuous basis, so it can be said that inflation is defined as a change that occurs in the
aggregate price level and is continuous (Mankiw 2000). Based on Bank Indonesia's
definition, inflation is a tendency of increasing prices of goods and services in general and
continuously. Meanwhile, the inflation rate is the development of price increases of a number
of goods and services in general over a period of time. Based on the general definition of
inflation there are three important aspects, namely:
1. There is a tendency for prices to increase, meaning that over a period of time, prices
show an increase.
2. Price increases are sustained, meaning that they increase over time.
3. The definition of price is the general level of price, meaning that it covers the whole
commodity and not just one or a few commodities, but in general or in aggregate.
12
Relationship between Remittances and Unemployment
The relationship between unemployment and remittances is explained by the vicious
circle of poverty theory by Nurkse (1961). Where this theory illustrates that a poor country
will remain poor and will always be entangled in poverty which proxies the amount of
unemployment. This situation can be
Poverty will continue to occur if the poverty chain is not broken. Low productivity will result
in insufficient income to fulfill the necessities of life, so poverty cannot be avoided. Therefore,
to get out of the poverty problem, the chains of poverty must be broken through an increase
in national income reflected by GDP growth, so that people's welfare can be achieved.
Relationship between Economic Growth and Unemployment (Okun's Law)
Okun's law was proposed by Arthur Okun (1962), he explained the relationship
between unemployment and economic growth. Okun's Law states that there is a negative
relationship between unemployment and economic growth. According to him, a 1% increase
in the unemployment rate will lead to a decrease in economic growth by 2% or more.
Conversely, a 1% increase in output in economic growth will cause a decrease in the
unemployment rate by 1% or less.
Okun's Law plays a role in economic theory in understanding that the unemployment
rate can describe the state of a country's economy, whether it is healthy or sick. In this case, a
low unemployment rate indicates the health of a country's economy. Conversely, a high
unemployment rate illustrates the sickness of the country's economy.
Therefore, various efforts to reduce the unemployment rate actually have a major
impact on economic growth where a decrease in the unemployment rate also equates to a
better level of public welfare (Mankiw 2000). In more detail, this ridge can be seen in Figure.
Inflation and Unemployment Relationship (Phillips Curve Theory)
In 1958, A.W. Phillips, a New Zealand-born economist, published an article that
examined the relationship between unemployment and the rate of change of wages in the UK
from 1861-1957. In the article, Phillips showed a negative relationship between the
unemployment rate and the inflation rate (wage change rate) in a case study for the UK.
Phillips showed that years with low unemployment tend to be accompanied by high inflation,
and years with high unemployment tend to be accompanied by low inflation (Samuelson and
Nordhaus 1990).
13
A.W. Phillips (1958) in Mankiw (2000) describes how the relationship between
inflation and unemployment rate is based on the assumption that inflation is a reflection of an
increase in aggregate demand. With the increase in aggregate demand, it is in accordance
with the theory of demand that if demand rises, prices will rise. The Phillips curve proves that
price stability and high employment are not possible simultaneously, which means that if we
want to achieve high employment in the sense of low unemployment, then as a consequence
we must be willing to bear the burden of high inflation.
The Phillips curve is a negative relationship between the unemployment rate (U) and
the inflation rate (P), as explained in Figure 8. A high growth rate of aggregate demand
stimulates an increase in output, an increase in output makes firms need more labor to
produce output, hence lowering the unemployment rate. A high growth rate of demand also
leads to a steady increase in prices (rising inflation), so the Phillips curve postulates a trade-
off between the inflation rate and the unemployment rate.
1.2 Framework of Thought
Figure 9 shows the flow of thought about the research conducted by the author. A
high unemployment rate can be caused by limited job vacancies. The limited job vacancies
are sometimes one of the reasons why people in the world migrate to look for jobs abroad. In
addition to the possibility of higher job vacancies, wage differences can also be one of the
reasons for migration. Wages earned by migrants that are transferred to family or relatives in
their home country are called remittances. Remittances are an important source of financial
support for migrant families, which can directly increase their household income, thereby not
only reducing unemployment but also reducing poverty and inequality. In the long run,
remittances can improve the welfare and standard of living of the country of origin because
the flow of remittances can strengthen the economy, education, health, and so on. The flow of
thought in the author's research can be seen conceptually in Figure 9.
Data Type and Source
This study uses secondary data from the World Bank (World Development Indicators,
WDI). The type of data used is panel data, which is a combination of time series and cross-
section data. The time series data used in the study covers the period 2000-2021, while the
cross-section data used covers 65 countries. In detail, the data used in the study can be seen
in Table.
14
1.1 Operational Definition
The operational definitions in this research related to the decline in the unemployment
rate are as follows:
1. Unemployment
Unemployment in this study is the unemployment rate, which is the percentage of
unemployment out of the total labor force.
2. Remittances
Remittances in the study are the ratio of remittances inflow in cash to GDP or the
share of remittances inflow to GDP where the unit is in percentage.
3. Exchange rate
The exchange rate or exchange rate is the agreed exchange rate between two
countries. In this study, it is the domestic currency (local currency of a country under
study (cross section in this study) per international dollar or in this study, namely
against the United States dollar (USD). Exchange rates are used in making
remittances both inflow and outflow, international trade between countries and others.
In this study, the exchange rate is made in logarithm (log) form to equalize units.
4. Real interest rate
The real interest rate is an inflation-adjusted lending rate measured by the deflator, so
this study uses real interest rates that have been corrected by inflation.
5. Trade openness
In this study, trade openness used is the ratio of trade (exports and imports) to GDP.
6. Inflation
Inflation in this study is measured from the consumer's point of view, so the inflation
used is measured by the Consumer Price Index (CPI).
7. GDP per capita
GDP per capita used in this study is the GDP of a country divided by the mid-year
population where the data is already available in the World Development Indicators.
In this study, GDP per capita is made in logarithm (log) form to equalize units.
Data Analysis and Processing Methods
The analysis methods used in the research are qualitative and quantitative methods.
Qualitative methods are used to descriptively analyze the development of the variables used
in the study, namely by interpreting the results of quantitative analysis methods. The
quantitative method used is the dynamic panel data method, using the Generalized Method of
15
Moment (GMM) method. The data processing carried out in the study used R Studio 4.2.2
software and Microsoft Excel 2021.
Exploratory Descriptive Analysis
Exploratory descriptive analysis is an analysis in examining a group of objects, a
condition, a system of thought, or a class of events in the present (Nazir 1999). The purpose
of exploratory descriptive analysis is to make a descriptive, systematic description or painting
that is factual about the facts and characteristics and relationships between phenomena that
occur. The explorative descriptive analysis conducted in this study is to see the relationship
between remittance receipt data, and other macroeconomic variables to unemployment rate
data. The analysis is carried out in 65 countries which are grouped into three groups based on
their remittance receipts, namely countries with low, medium, and high remittance receipts
during the period 2000-2021.
Panel Data Model
In econometrics, there are three types of data, namely time series data, cross section
data, and panel data. Panel data (panel pooled data) is a combination of time series data and
cross section data. Panel data itself consists of two types, namely static panel and dynamic
panel. The difference is that the dynamic panel data method applies dynamism, which means
that it is continuously updated.
The dynamic panel is characterized by the presence of lag variables (Arellano and Bond
1991), which change continuously in the present or current data and have a relationship to
previous or past data.
Thus, in dynamic panels, usually the value in the previous period (lag) of the dependent
variable will be used as an independent variable because it is considered that the lag of the
dependent variable (
𝑥
𝑡
−1
) has
relationship with the current dependent variable (𝑦𝑡 ). It can be said that the data
panel is an observation of individual units over a period of time.
certain.
Regression using panel data is called a panel data regression model. Baltagi (2008)
states that the advantages and benefits of using panel data analysis methods in ecometric
analysis are:
1. Panel data provides or presents data that is more informative, more varied, has little
collinearity between variables, has greater degrees of freedom, and makes estimation of
16
parameters more efficient.
2. Panel data can control the problem of individual data heterogeneity, meaning that it can
control the problem of heterogeneity in variables that are not included in the model
(unobserved individual heterogeneity).
3. Panel data is better at identifying and estimating influences that cannot be detected in
pure time series and cross section data.
4. Panel data can be used to test studies with dynamic behavioral problems that are more
complex than pure time series or cross section data analysis.
5. Panel data at the micro level can minimize or eliminate bias that occurs due to data
aggregation to the macro level.
6. The problem of panel data at the macro level is unlike the problem of the nonstandard
distribution type of the unit root test in time series data analysis when it has a longer
time series, so the problem can be more easily overcome.
7. Panel data is a more appropriate method for studying adjustment dynamics.
Panel data analysis generally uses data in level form to facilitate model interpretation.
However, if the research uses data containing time series elements that have a trend, it is
necessary to test the stationarity assumption first, namely by conducting a unit root test. If the
unit root test results show a trend in the level data so that the data is not stationary at the
level, then the problem must be overcome, one of which is by converting the level data into
first difference data. This is done to ensure that the relationship between the dependent
variable and the independent variable does not show spurious regression and avoid
misleading results. This unit root test is suggested by Baltagi (2008) for panel data with N
and T that are relatively not large. In the case of this study, it uses 22 years of time series data
and cross section data 65 countries where both time series and cross sections include a large
amount of data, so this unit root test can be skipped.
Dynamic Panel Data Regression Estimation
Panel data consists of a collection of observations of an object in a period of time
(Gujarati 2006). This study uses dynamic panel data regression estimation, where its use is
driven by the widespread fact that the development of dynamic relationships between
economic variables in relation to the analysis of adjustment dynamics. The dynamic
relationship is characterized by the lag of the dependent variable
(
𝑦
𝑡
−1
) between the
independent variables (Jacobs et al. 2019).
The existence of the lag of the dependent variable as an independent variable results in
17
the emergence of endogeneity problems, so that if the model is estimated with static panel
data analysis, it will produce biased and inconsistent estimators (Verbeek 2004). Arellano and
Bond (1991) proposed the use of the Generalized Method of Moments (GMM) approach to
overcome the endogeneity problem in order to produce unbiased and consistent estimators.
There are two estimation procedures that are usually used in the GMM method, the first is
First- Differences GMM (FD-GMM) and the second is System GMM (Sys-GMM). Parameter
estimation on dynamic panel data uses Spatially Corrected Arellano-Bond (SCAB) parameter
estimation. Parameter significance testing aims to determine the significance of the
relationship between the independent variable and the dependent variable in the model
(Baltagi 2005).
The advantages of the GMM estimator include being a common estimator that can
provide a more useful framework for comparison and assessment and is a simple alternative
to other estimators, especially the Maximum Likelihood Estimator (MLE). However, the
disadvantage is that the GMM estimator is efficient only in sample size large, but less
efficient if the number of samples is limited. This GMM estimator in its calculations requires
some programming implementation so that software is needed that supports the use of the
GMM approach; for example, such as Stata, R Studio, E- Views and others. The estimation
method for dynamic panel data analysis with the GMM approach can use several estimation
procedures as follows:
1. First Difference GMM (FD-GMM)
Arellano and Bond (1991) state that additional instruments can be obtained in a
dynamic panel data analysis model when using the orthogonality condition that occurs
between the lag value
𝑦
𝑖𝑡
and the error component
𝜀
𝑖𝑡
. The
weakness of the FD GMM
approach is that the estimator at a limited sample size or number of time periods will be
biased, even more biased toward
lower than the estimator with the FEM approach. This occurs because the lag level of the
data series is weakly correlated with the next first difference, causing the instruments
available for the next first difference equation to be weak (Blundell and Bond 1998).
2. System GMM (Sys-GMM)
Blundell and Bond (1998) stated the importance of utilizing initial conditions in
producing efficient estimators in dynamic panel data analysis when t is small. The basic idea
of using the Sys-GMM method is to estimate a system of equations both at first difference
and at level where the instrument used at level is the lag first difference of the series.
18
The Wald test is used as a simultaneous significance test of model parameters, while
the Z test is used as a partial significance test of model parameters. Model specification
testing uses panel unit root test, Arellano-Bond test (consistency test), Sargan test (instrument
validity test) and unbiasedness (Arellano and Bond 1991). Interpretation of dynamic panel
model with GMM method can see the short-term effect, long-term effect, speed of
convergence and half time of convergence.
a. GMM Model Selection Testing Stage
Based on Firdaus (2011), the steps or testing stages in determining the selection of the
best and appropriate GMM dynamic panel model use several criteria as follows:
1. Unbiased
The unbiasedness test of the estimator with the GMM approach is carried out by
checking for contemporaneity. This means that the coefficient value of the estimator with
GMM must be between the coefficients of the PLS and FEM estimators. The lag coefficient
of the dependent variable generated by PLS will be biased upwards, while the lag coefficient
of the dependent variable generated from FEM will be biased downwards. To avoid bias, the
value must be between PLS and FEM.
2. Valid Instrument
The test to determine whether the instrument is valid or not is to use the Sargan test.
The instrument variable validity test is used to see the possibility of bias in the estimated
parameters due to the inappropriate use of instrument variables in the equation. To test the
validity of the instrument variable, the Sargan Specification Test is used (Gujarati 2006). The
Sargan test is used to determine the validity of the use of instrument variables that exceed the
number of parameters being estimated. The hypothesis in the Sargan test is as follows
H0 = the instrument used in the model estimation is valid
H1 = the instrument used in the model estimation is invalid
The test criterion
H
0
is rejected if
𝑝
-
𝑣𝑎𝑙𝑢𝑒
< α
.
If the Sargan test shows rejection of the null hypothesis, it is said to be valid. In the first
stage using the FD-GMM method, if the results of the FD-GMM method show that the
instrument used is invalid, the Sys-GMM method is used. The sargan test is used for over
identifying restriction to test the validity problem in the instrument used. If the instrument is
valid, there is no correlation between the instrument and the error component.
3. Consistent
19
The consistency test of the estimator with the GMM approach is carried out by testing
the statistical significance of m1 and m2 which are usually calculated automatically through
software. The m1 and m2 test is also often referred to as the Arrellano-Bond test because in data
processing, to check this consistency is by looking at m1 and m2 which are seen from the AR(1)
and AR(2) values.
Remittances in the World and Country Groupings Based on Remittance Receipt Rates
Remittance receipts in the world show a positive trend, which basically increases every
year. The development of this remittance receipt in detail can be seen in Figure 1 earlier.
Based on this figure, it can be seen that the volume of remittances in 2020 reached 651.05
billion USD, which is 17.10% greater than the volume in 2015, which amounted to 556.02
billion USD.
Remittance receipts can be categorized into several groups. In this study, the division of
groups is divided into three group categories based on the amount of remittance share of
GDP received by a country, namely groups of countries with low, medium, and high
remittance receipts. There is no definite benchmark in determining whether a country belongs
to the low, medium, or high category. However, the World Bank report in 2022 stated in its
article that remittances in Sub-Saharan Africa rose to 10.3%, which is a high remittance
receipt. The determination of the country group division category in this study will be based
on the following benchmarks; low remittances 0-3%, medium remittances >3-10%, and high
remittances >10%.
Figure 10 shows the number of countries per group based on their remittance receipts.
Based on the figure, it can be seen that the category most occupied by countries in the world
is the low remittance group, which is occupied by 39 countries. This means that the majority
of countries in the world receive low remittances, which range from 0-3% of their country's
GDP share. This shows that although currently remittances have developed quite
significantly from year to year, but for most countries the receipt of remittances still does not
contribute much to their country's income.
Countries that fall into the category of countries with low remittance receipts include
the United States, Australia, the United Kingdom, China, and South Korea. The examples of
these countries show that in the category of low remittance countries, the majority come from
countries with developed country specifications. In this case, Indonesia, which is a
developing but developed country, is also included in the category of high remittance
countries. In the medium remittance receiving category, there are 15 countries that fall into
20
this category. Examples of countries are Nigeria, Mongolia, Uganda, and others, which
shows that the majority of those in the medium remittance category come from developing
countries. In countries with high remittance categories, there are 11 countries included in it.
These countries include Bangladesh, India, Jamaica, Guatemala and the Philippines. These
countries tend to belong to group of countries that are still low-income or poor and
developing countries.
Determination of the Best Model GMM Method
In this study, panel data testing was carried out using the dynamic panel method,
precisely by using the Generalized Method of Moment (GMM). Before concluding the results
of the GMM estimation, first test whether the GMM model used is correct, and whether the
estimation results are good or not.
In determining the best model and the accuracy of the estimation, the GMM method has
its own criteria that are used by conducting three stages of testing as follows:
1. Instrument validity test (Sargan test)
In order for the instrument used to be valid, it must not reject H0 or p-value
must be instrumentally insignificant, meaning
𝑝
-
𝑣𝑎𝑙𝑢𝑒
> α
2. Consistency test (Arellano-Bond correlation test)
For the instrument to be consistent, the Auto-Correlation AR(2) must not reject H0 or must be
instrumentally insignificant, meaning AR(2) > α.
3. Unbiasedness test
In order for the instrument used to be unbiased, the value of the lag coefficient of the
dependent variable in the GMM method must be between the value of the lag coefficient of
the dependent variable in the Pooled Least Squares (PLS) and Fix Effect Model (FEM)
methods. That is, the dependent lag coefficient in FEM < GMM < PLS.
Based on these three tests, the best model in the study will be obtained, so that further
estimation can be continued and valid. Furthermore, after processing all 65 countries directly,
the results of this study will also be discussed based on group categories for each stage of the
test conducted.
21
Determination of the Best Model in 65 Countries
a. Instrument validity (Sargan Test)
Table 3 GMM instrument validity test results (Sargan test) in 65 countries
Sargan Test Statistic Value
P-Value
FD-GMM
43,86022
0,01908735**
Sys-GMM
45,76528
0,13047
Table 3 shows the results of the instrument validity test on the entire sample of
countries, namely 65 countries in the world. The test conducted is using the Sargan test. The
Sargan estimation results show a p-value on FD-GMM of 0.01908735 and on Sys-GMM of
0.13047. The criterion in determining whether the instrument is valid is to compare the p-
value of the Sargan test estimation results with the real level. In this study using the real level
α = 5%. The rejection criterion is if 𝑝 -
𝑣𝑎𝑙𝑢𝑒 < α then reject H0, so in FD-GMM the value is under
real level, where rejecting H0 means that there is a correlation between residuals or the
instrument used in model estimation is invalid. As for Sys-GMM, the decision is not to reject
𝐻0 , which means that there is no correlation between residuals and overidentifying
restrictions or the instrument variables used are more than the number of parameters
estimated so that the instrument is invalid.
The instruments used are valid, so it can be concluded from the Sargan test results that only
in Sys-GMM where the instruments used are valid.
FD-GMM Sargan test result = 0.01908735 < 0.05; so it is not valid. Sargan Sys-GMM test
result = 0.13047 > 0.05; so it is valid.
b. Consistency (Arrellano-Bond Test)
Table 4 GMM consistency test results on 65 countries
Test
Value
Probability
FD-GMM
AR (1)
-5,209869
1.8897e ***-05
AR (2)
-0,06389889
0,94905
Sys-GMM
AR (1)
-5,090988
3.562e ***-05
AR (2)
-3,4218394
0,67314
22
c. Unbiased
Table 5 GMM unbiased test results on 65 countries
Variables
Coefficient
FD-GMM
Sys-GMM
FEM
PLS
𝑢𝑛𝑒𝑚𝑝
it−1
0,8605021
0,8907018
0,8702047
0,9843626
Table 5 shows the estimation results of the GMM unfamiliarity test on 65 countries,
where there are FD-GMM, Sys-GMM, FEM, and PLS. From the results, it can be seen that
the value of the lag coefficient of the dependent variable in the Sys-GMM method is between
the value of the lag coefficient of the dependent variable in the PLS and FEM methods, while
for FD-GMM the value is below FEM and PLS. It is known that the estimator from PLS will
be biased upwards
and the estimator from FEM will be biased downwards.
Based on
the estimation results that meet the criteria is Sys-GMM whose dependent lag coefficient
value is between the FEM and PLS values. This indicates that the model used is not biased
for the Sys- GMM model. So, the lag coefficient of unemployment (𝑢𝑛𝑒𝑚𝑝𝑖𝑡−1 ) in Sys-
GMM is the value of
is in between FEM and PLS.
FD-GMM is biased downward, the coefficient of unemp(-1) in FD-GMM < FEM and PLS.
Sys-GMM is unbiased, unemp(-1) coefficient on FEM < Sys-GMM < PLS.
Based on the Sargan test, consistency test, and unbiased test, the model that meets the
criteria of the three is Sys-GMM, so for the estimation of all countries as a whole, the Sys-
GMM method will be used.
Determining the Best Model in Low Remittance Countries
a. Instrument validity (Sargan Test)
Table 6 Results of instrument validity test (Sargan test) GMM of low remittance countries
Sargan Test Statistic Value
P-Value
FD - GMM
28,48362
0,074553
Sys - GMM
39
0,72301
23
Table 6 shows the results of the instrument validity test in countries with low
remittance receipts. The test conducted is using the Sargan test. The Sargan estimation results
show the p-value on FD-GMM of 0.074553 and on Sys-GMM of 0.72301. The criterion in
determining whether an instrument is valid is to compare the p-value of the Sargan estimation
results with the real level. In this study using the real level α = 5%, the criterion for rejection
is if 𝑝𝑝 - 0.074553.
𝑣𝑎𝑙𝑢𝑒 < α then reject H0, so that in both FD-GMM and Sys-GMM the decision is not to reject
𝐻0 . This means that there is no correlation between
residuals and overidentifying restrictions or instrument variables used
more than the number of estimated parameters. It can be concluded from the Sargan test
results that the instrument used is valid.
FD-GMM Sargan test result = 0.074553 > 0.05; so it is valid. Sargan Sys-GMM test result =
0.72301 > 0.05; so it is valid.
b. Consistency (Arrellano-Bond Test)
Table 7 GMM consistency test results on low remittance countries
Test
Value
P-value
FD-GMM
AR (1)
-3,981213
0,00068564***
AR (2)
1,154814
0,24817
Sys-GMM
AR (1)
-3,859724
0,00011352***
AR (2)
0,7143179
0,47503
The model consistency test for the category of countries with low remittances can be
seen from the results of the Arellano-Bond AR (1) and AR (2) statistical tests. The model is
declared consistent if H0 AR (1) is rejected and the significance test of AR (2) is not
significant. Both consistency tests were tested using the Arellano-Bond test. Based on Table
7, it can be seen that the dynamic panel method with the GMM Arellano-Bond approach has
met the criteria for the best model statistically consistent. The test results using FD-GMM
show that the AR (1) test has a value of -3.981213 with a p-value of 0.00068564. The AR (1)
test results show reject 𝐻0 . Test
AR (2) has a value of 1.154814 with a p-value of 0.24817 which means that it is not
enough evidence to reject 𝐻0 .
24
While for the estimation results using Sys-GMM, AR (1) is valued at -3.859724 with a p-
value of 0.00011352. The AR (1) test results show reject 𝐻0 . The AR (2) test is worth
0.7143179 with a p-value of 0.7143179.
value 0.47503 which means there is not enough evidence to reject 𝐻0 . From
conclusion of AR (1) test and AR (2) test both by using FD-GMM
The Arellano-Bond test results prove that the estimation is consistent and there is no auto-
correlation in the second-order first difference error, so it can be concluded that the absolute
growth convergence model in the study period is consistent.
Arellano-Bond AR(2) with FD-GMM = 0.24817 > 0.05; thus consistent. Arellano-Bond
AR(2) with Sys-GMM = 0.47503 > 0.05; thus consistent.
c. Unbiased
Table 8 GMM unbiased test results in low remittance countries
Variables
Coefficient
FD-GMM
Sys-GMM
FEM
PLS
𝑢𝑛𝑒𝑚𝑝
it−1
0,8979351
0,9050505
0,8757652
0,9868812
Table 8 shows the estimation results of the GMM unfamiliarity test.
Where
there are FD-GMM, Sys-GMM, FEM, and PLS. From these results it can be seen that the
value of the lag coefficient of the dependent variable in the FD-GMM and Sys-GMM
methods is between the value of the lag coefficient of the dependent variable in the
PLS and
FEM methods. Based on the estimation results, both
using FD-GMM and Sys-GMM,
the dependent lag coefficient value remains between the FEM and PLS values. This indicates
that the model used is not biased.
These results indicate that the estimation models using FD-GMM and Sys-GMM have
both met the criteria so that the estimation results can be concluded. However, because in the
category of countries with low remittance income, FD-GMM and Sys-GMM both meet the
requirements, in determining the model to be used, it can be seen from the number of
variables that are consistent in the estimation results of each method.
25
Determining the Best Model in Medium Remittance Countries
a. Instrument validity (Sargan Test)
Table 9 shows the results of the instrument validity test in countries with medium
remittance receipts. The test conducted is using the Sargan test. The Sargan estimation results
show the p-value on FD-GMM of 0.7226 and on Sys-GMM of 0.69731. The criterion in
determining whether the instrument is valid is to compare the p-value the result of the
estimation of the sargan with the real level, which is α = 5%. The rejection criterion is if 𝑝 -
𝑣𝑎𝑙𝑢𝑒 < α then reject H0, so both in FD-GMM and Sys-GMM the decision is not to reject 𝐻0 .
This means that the instrument variables used are more than the number of estimated
parameters.
It can be concluded from the results of the Sargan test that the instruments used are valid.
FD-GMM Sargan test results = 0.7226> 0.05; so it is valid. Sargan Sys-GMM test result =
0.69731> 0.05; so it is valid.
Table 9 GMM instrument validity test results (Sargan test) on medium remittance countries
b. Consistency (Arrellano-Bond Test)
Table 10 GMM consistency test results (Arellano-Bond test) on medium remittance countries
The model consistency test is seen from the results of the Arellano-Bond AR statistical test.
Sargan Test Statistic Value
P-Value
FD-GMM
15
0,7226
Sys-GMM
15
0,69731
Test
Value
Probability
FD-GMM
AR (1)
-2,853283
0,004327***
AR (2)
-1,629591
0,10319
Sys-GMM
AR (1)
-2,848432
0,0043935***
AR (2)
-1,607797
0,10788
26
(1) and AR (2). The model is declared consistent if the AR (2) significance test is not
significant. This means that AR (2) p-value > real level. Based on Table 10, it can be seen
that the dynamic panel method with the Arellano-Bond GMM approach has met the criteria
for the best model statistically consistent. The test results using FD-GMM show that the AR
(2) test is -1.629591 with a p-value of 0.10319 which means there is not enough evidence to
reject 𝐻0 . While for the estimation results using Sys-
GMM, the AR (2) test result is -1.607797 with a p-value of 0.10788 which is
also means there is not enough evidence to reject 𝐻0 . From the conclusion of the AR (2) test,
both using FD-GMM and Sys-GMM, it proves that the estimation carried out is consistent
and there is no auto correlation.
on the second-order first difference error, so from the results of the Arellano-Bond test it can
be concluded that the absolute growth convergence model in the study period has been
consistent.
p-value of Arellano-Bond AR(2) with FD-GMM = 0.10319> 0.05; thus consistent.
p-value of Arellano-Bond AR(2) with Sys-GMM = 0.10788 > 0.05; thus consistent.
c. Unbiased
Table 11 GMM unbiasedness test results on medium remittance countries
Variables
Coefficient
FD-GMM
Sys-GMM
FEM
PLS
𝑢𝑛𝑒𝑚𝑝
it−1
0,9518214
0,8977412
0,8795941
0,9469632
Table 11 shows the estimation results of the GMM unbiasedness test for countries
with medium remittance receipts. It is known that the estimator of PLS will be biased
upwards and the estimator of FEM will be biased downwards. Based on these results, it can
be seen that the value of the lag coefficient of the dependent variable in the Sys-GMM
method is between the value of the lag coefficient of the dependent variable in the PLS and
FEM methods. While in FD-GMM the value is above the value of FEM and PLS. The
criterion in determining whether the model used is unbiased is that the value using GMM
must be between the values when estimated with FEM and PLS. Based on the estimation
results, the one that meets the criteria is Sys-GMM. This indicates that the Sys-GMM model
used is unbiased.
27
FD-GMM: unemp(-1) coefficient in FEM < PLS < FD-GMM; hence biased. Sys-GMM:
unemp(-1) coefficient on FEM < Sys-GMM < PLS; thus unbiased.
Determining the Best Model in High Remittance Countries
a. Instrument validity (Sargan Test)
Table 12 GMM instrument validity test results (Sargan test) on high remittance countries
Notes: ***, ** significant at 1%, 5% level
Table 12 shows the results of the instrument validity test in countries with high
remittance receipts. The Sargan test estimation results show the p-value on FD-GMM of
0.69238 and on Sys-GMM of 0.71561. The criterion in determining whether the instrument is
valid is by comparing the p-value of the Sargan estimation results with the real level, which is
α = 5%. The rejection criteria are if 𝑝 - 𝑣𝑎𝑙𝑢𝑒 < α then
reject H0, so the value of both p-values in both FD-GMM and Sys-GMM
is above 0.05 so the decision is not to reject 𝐻0 . This means that the instrument variables
used are more than the number of parameters assumed, so it can be concluded from the
Sargan test results that the instruments used are more than the number of parameters
assumed.
used is valid.
FD-GMM Sargan test result = 0.69238> 0.05; so it is valid. Sargan Sys-GMM test result =
0.71561> 0.05; so it is valid.
b. Consistency (Arrellano-Bond Test)
GMM consistency test results (Arellano-Bond test) on high remittance countries
Test
Value
Probability
FD-GMM
AR (1)
-0,6848566
0,0049343***
AR (2)
-1,650468
0,098847
Sargan Test Statistic Value
P-Value
FD-GMM
11
0,69238
Sys-GMM
11
0,71561
28
Sys-GMM
AR (1)
-2,727053
0,0063903***
AR (2)
-2,202232
0,276409
The model consistency test for the category of countries with high remittances can be
seen from the results of the Arellano-Bond AR (1) and AR (2) statistical tests. The model is
declared consistent if H0 AR (1) is rejected or significant and AR (2) significance test is not
significant. Both consistency tests were tested using the Arellano-Bond test. Based on Table
13, it can be seen that the dynamic panel method with the GMM Arellano-Bond approach has
met the criteria for the best model statistically consistent. The test results using FD-GMM
show that the AR (1) test has a value of - 0.6848566 with a p-value of 0.0049343. The AR (1)
test results show reject 𝐻0 . The AR (2) test is worth -1.650468 with a p-value of
0.098847 which means there is not enough evidence to reject 𝐻0 .
While for the estimation results using Sys-GMM, AR (1)
value of -2.727053 with a p-value of 0.0063903. AR test results
(1) shows reject 𝐻0 . The AR (2) test is -2.202232 with a p-value of 0.276409 which means
there is not enough evidence to reject 𝐻0 . From the conclusions of the AR (1) test and AR (2)
test both using FD-GMM and Sys-GMM
This proves that the estimation is consistent and there is no auto-correlation in the second-
order first difference error, so from the Arellano-Bond test results it can be concluded that the
model of the unemployment rate in the study period is consistent.
Arellano-Bond AR(2) with FD-GMM = 0.098847 > 0.05; thus consistent. Arellano-Bond
AR(2) with Sys-GMM = 0.276409 > 0.05; thus consistent.
c. Unbiased
Table 14 GMM unbiased test results on high remittance countries
Variables
Coefficient
FD-GMM
Sys-GMM
FEM
PLS
𝑢𝑛𝑒𝑚𝑝
it−1
0,778895
0,8892936
0,8685024
1,0727706
The estimation results of the GMM unfamiliarity test on countries with high remittance
29
receipts are shown in Table 14. In determining whether the model is biased or not is to
compare the value of the dependent lag coefficient with the results by estimating using
GMM, FEM, and PLS. The value of the dependent lag coefficient in GMM must be between
the FEM value and the PLS value. Based on the estimation results in Table 14, it can be seen
that the value of the lag coefficient of the dependent variable in the FD-GMM method for 65
countries in the world with a high remittance category does not meet the requirements or is
biased. This is because the value of the dependent lag coefficient using FD-GMM is lower
than FEM or PLS. While the estimation results with Sys-GMM the lag coefficient value of
the dependent variable is between the coefficient values when estimated using the PLS and
FEM methods. Thus, based on the estimation results, the one that meets the criteria is Sys-
GMM. This indicates that the Sys-GMM model used is unbiased.
FD-GMM: unemp(-1) coefficient on FD-GMM < FEM < PLS; thus biased. Sys-GMM:
unemp(-1) coefficient on FEM < Sys-GMM < PLS; ⸫so unbiased.
Short-Term Estimation Results with GMM Method
Table 15 shows the overall estimation results of 65 countries using Sys-GMM. Based
on the table, it can be seen that in the Sys-GMM estimation results there are four significant
variables. Two variables are significant at the 1% real level, namely the previous year's
unemployment variable and GDP, one variable is significant at the 5% real level, namely the
exchange rate variable, and 1 variable at the 10% real level, namely the real interest rate.
Based on Table 15 in the Sys-GMM results, it can be seen that there are four variables
that have a significant effect at various real levels. The most significant variables or with the
smallest real level are the previous year's unemployment variable and GDP. If the previous
year's unemployment increases by 1%, the current unemployment will increase by
0.8907018%, cateris paribus. Meanwhile, if this year's GDP increases by 1%, current
unemployment will decrease by 1.4250182%, cateris paribus. As for the variable that is the
focal point of this research, namely remittances, when estimated in 65 countries as a whole, it
gives results that do not significantly affect the unemployment rate in a country.
30
Table 1 5 Short-run estimation results with Sys-GMM for 65 countries
Variables
Coefficient
Sys-GMM
𝑢𝑛𝑒𝑚𝑝
it−1
0,8907018
(2,2e-16 )***
Log_GDP
-1,4250182
(7,332e-05 )***
Remit
-0,3059113
(0,4991232)
Log_EXR
-0,8202765
(0,0026518)**
RIR
0,5849983
(0,0406857)*
Trade
-0,8065722
(0,1451189)
INF
-0,9732014
(0,1154935)
Although the results given are in accordance with the hypothesis, namely a negative
effect and the amount given is also relatively large, the results for the remittance variable are
not valid because the results are not statistically significant. Similar to the remittance
variable, the trade openness and inflation variables also have statistically insignificant results.
Other statistically significant variables are exchange rate, real interest rate and GDP. In the
exchange rate variable, if the exchange rate of a country increases by 1%, the unemployment
rate will decrease by 0.8202765%, cateris paribus. In the real interest rate variable, if the
interest rate increases by 1%, the unemployment rate will increase by 0.5849983%, cateris
paribus. As for GDP, if the GDP variable increases by 1%, it will reduce unemployment by
1.4250182%, cateris paribus.
Short-Term Estimation Results in Low Remittance Countries:
Table 16 shows the comparison of estimation results using FD-GMM and Sys-GMM in
countries with low remittance categories. Based on the table, it can be seen that in the FD-
GMM estimation results there are six significant variables with two variables at the 1% real
level, three variables at the 5% real level, and 1 variable at the 10% real level. The Sys-GMM
estimation results have four statistically significant variables, where two variables are
significant at the 1% real level, one variable at the 5% real level, and 1 at the 10% real level.
When compared between the FD- GMM and Sys-GMM models, the best model for low
remittance category countries is FD-GMM.
Table 16 Comparison of estimation results with FD-GMM and Sys-GMM in low remittance countries in
the short run
Based on Table 16 in the FD-GMM results, it can be seen that all variables have a
significant effect at various real levels. The most significant variables or with the smallest
real level are the previous year's unemployment variable and GDP. If the previous year's
unemployment increases by 1%, the current unemployment will increase by 0.8979351%,
cateris paribus. Meanwhile, if this year's GDP increases by 1%, current unemployment will
decrease by 1.542783%, cateris paribus. As for the variable that is the focal point of this
research, namely remittances, it is also significant to the unemployment rate, but the
significance is at a real level that is still quite high, namely 10%. The amount of contribution
Variables
Coefficient
FD-GMM
Sys-GMM
𝑢𝑛𝑒𝑚𝑝
it−1
0,8979351
0,9050505
(2,2e-16 )***
(2,2e-16 )***
Log_GDP
-1,542783
-1,0954015
(8,288e-06 )***
(3,772e-06 )***
Remit
-0,0561187
-0,0042315
(0,05815)*
(0,0721)*
Log_EXR
-0,7163167
-0,3051080
(0,04298)**
(0,1533)
RIR
0,0177723
0,0059971
(0,01204)**
(0,0127)**
Trade
-0,6140375
0,0507451
(0,03788)**
(0,3584)
INF
-0,0030571
-0,0049673
(0,39175)
(0,5527)
32
given is also still very low, i.e. if remittances increase by 1%, the unemployment rate will
decrease by 0.0561187%, cateris paribus. This shows that the contribution of remittances to
the reduction of unemployment in countries with low remittance receipts is still relatively
small in influence or contribution. This is in accordance with the fact that in countries with
low category, remittance receipt does not contribute too much in influencing the decrease of
unemployment, this is because countries with low remittance receipt are more influenced by
other variables, such as their country's GDP in eradicating unemployment. Furthermore, the
other independent variables that are significant to the decrease in unemployment all fulfill the
hypothesis in their influence on the unemployment rate.
Short-Term Estimation Results on Medium Remittance Countries:
Table 17 Short-run estimation results with Sys-GMM in medium remittance countries
Variables
Coefficient
Sys-GMM
𝑢𝑛𝑒𝑚𝑝
it−1
0,8977412
(2,2e-16 )***
Log_GDP
-1,2911585
(1,814e-05 )***
Remit
-0,0618531
(0,05967)*
Log_EXR
-0,3721507
(0,03878)**
RIR
0,0187841
(0,13564)
Trade
-0,2461660
(0,02556)**
INF
-0,0359289
(0,02245)**
Based on the estimation results in Table 17, it can be seen that in countries with
medium remittance categories, the previous year's unemployment rate and GDP have a
significant effect at the 1% real level on reducing the unemployment rate. Meanwhile, the
exchange rate, trade openness, and inflation have a significant effect at the 5% real level.
Meanwhile, remittances, which are the key variable in this study, are also significant but at
the 10% real level on the decline in the unemployment rate. The only variable that has no
significant effect is interest rate. The variable that positively influences unemployment is
unemployment in the previous year, where every 1% increase in unemployment in the
previous year will increase unemployment in this year by 0.8977412%, cateris paribus.
The variables of remittances, GDP, exchange rate, trade openness, and inflation
negatively affect the decrease of unemployment. This means that if there is an increase in
these variables, it will decrease the unemployment rate. The biggest influence is on the
increase of GDP variable, where if there is a 1% increase of GDP, it will decrease
unemployment by 1.2911585%. As for remittances themselves, although they have a
significant negative effect, their impact on reducing unemployment in countries with medium
remittances is still less influential than in countries in the high remittance category, where
every 1% increase in the remittance variable will reduce unemployment by 0.0618531%.
Another variable that has a positive effect is inflation, where if inflation increases by 1% then
unemployment will decrease by 0.0359289%, cateris paribus. This is in accordance with the
Phillip Curve theory, which states that inflation and unemployment have a negative
relationship. This means that if you want to reduce the unemployment rate, you must be
willing to bear higher inflation.
Short-Term Estimation Results in High Remittance Countries:
Table 18 Short-run estimation results with Sys-GMM in high remittance countries
Variables
Coefficient
Sys-GMM
𝑢𝑛𝑒𝑚𝑝
it−1
0,8892936
(2,2e-16 )***
Log_GDP
-1,1848122
(5,311e-06 )***
Remit
-0,1210119
(0,06135)*
Log_EXR
-0,6766757
(0,03932)**
RIR
0,0438938
(0,03593)**
34
Trade
-0,4648102
(0,12053)
INF
-0,4278400
(0,02385)**
Notes: ***, **,* significant at 1%, 5%, 10% level.
Table 18 shows the Sys-GMM estimation results for countries with high remittance
receipts. Based on these results, all variables are significant except the trade openness
variable. Based on the estimation results in Table 18, it can be seen that the previous year's
unemployment and GDP have a significant effect at the 1% real level on the decline in the
unemployment rate in countries with high remittance income. Meanwhile, exchange rates,
interest rates, and inflation have a significant effect at the 5% real level. Meanwhile,
remittances, which are the key variable in this study, have the same result as countries with
low and medium categories, i.e. in countries with high remittance category, the significance
remains at 10% real level on the decrease of unemployment rate. There is one variable that
does not have a significant effect, namely trade openness.
The variables that positively affect unemployment are the previous year's
unemployment and interest rate. Every 1% increase in unemployment in the previous year
will increase unemployment in this year by 0.8892936%, cateris paribus. Meanwhile, for
interest rate variable, that is, if there is a 1% increase in the interest rate variable, the
unemployment rate will increase by 0.0438938%, cateris paribus.
The variables of GDP, remittances, exchange rate, and inflation negatively affect the
decrease of unemployment. This means that if there is an increase in these variables, it will
reduce the unemployment rate. The biggest influence is on the increase of GDP variable,
where if there is an increase of GDP by 1%, it will decrease unemployment by 1.1848122%,
cateris paribus. As for remittances themselves, although they have a significant negative
effect, their impact on reducing unemployment in countries with high remittances is still less
influential than countries in the low and medium remittance categories, but the coefficient
value itself is the largest among the three categories of countries. Every 1% increase in the
remittance variable will reduce unemployment by 0.1210119%, cateris paribus. For the
exchange rate variable, every 1% increase in the exchange rate variable will decrease the
unemployment rate by 0.6766757%, cateris paribus. For the inflation variable, if inflation
increases by 1%, unemployment will decrease by 0.42784%, cateris paribus.
Results of Long-Term Estimation with GMM Method:
In the analysis using the GMM method, the estimation results obtained after processing
the data for the first time are the estimation results for the short term. But in addition to
getting the results of short-term effects, the GMM method can also do further estimation to
determine the long-term effect on the dependent variable in the study. The advantage of using
the GMM method in knowing the response or influence of the variables used in the long term
is that researchers do not need to worry about endogeneity problems.
Long-run Estimation Results on Medium Remittance Countries Table 21 Long-run response
on medium remittance countries
In Table 21, which displays the results of the long-term estimation of the decline in the
unemployment rate in the category of countries with medium remittance receipts, the response results
are more or less the same as the short-term response. The difference is that in the long-term
estimation, the coefficient value tends to be smaller than the short-term estimation. The long-term
result shows that if there is an increase in remittances by 1%, it will inhibit the unemployment rate in
the long run by 0.0604868%, ceteris paribus. A 1% increase in the exchange rate will also reduce the
unemployment rate in countries with medium remittances by 0.3639302%, cateris paribus in the long
run. Trade openness and GDP growth by 1% in the long run will reduce the unemployment rate by
0.2407284% and 1.2626380%, cateris paribus. In the inflation variable, if there is an increase in
inflation by 1%, the unemployment rate will be suppressed by 0.351353%, cateris paribus. The
values (coefficients) of the magnitude of the effect in the long run are lower than the coefficients in
the short run. This is true because in the long run the variables involved can already adjust in the long
run.
Long-Run Estimation Results in High Remittance Countries:
Based on the estimation results in the table, the results show that in the long run if there
is an increase in remittances by 1% then in the long run the unemployment rate decreases by
0.10930965%, ceteris paribus. On the exchange rate variable in the long run there is an
Variables
Probability
Response
Long Term
Log_GDP
(1,814e-05 )***
-1,2626380
Remit
(0,059679)*
-0,0604868
Log_EXR
(0,038781)**
-0,3639302
RIR
(0,13564)
0,0183692
Trade
(0,02556)**
-0,2407284
INF
(0,02245)**
-0,0351353
36
increase in the response coefficient which is lower than in the short run. If the exchange rate
increases by 1%, the unemployment rate will decrease by 0.61123494%, cateris paribus. This
can be due to the fact that in the long run, countries that are categorized as countries with
Most of the high remittance receipts come from poor and developing countries. This makes
the development of trade openness, remittances, investment and so on that require
transactions with foreign currencies in this case related to the exchange rate will be faster
than other categories. The trade openness variable also has the same effect in the long term as
in the short term, but with a response coefficient value that is slightly lower than the short-
term effect. This is also the same as the exchange rate variable where in high remittance
income countries the development to be more advanced will adjust more to shocks and so on
in the long run.
Table 22 Long-term response in high remittance countries
Variables
Probability
Response
Long Term
Log_GDP
(5,311e-06 )***
-1,07022944
Remit
(0,06135)*
-0,10930965
Log_EXR
(0,03932)**
-0,61123494
RIR
(0,03593)**
0,03964812
Trade
(0,12053)
-0,41985845
INF
(0,02385)**
-0,38646323
Based on Table 22, in detail, every 1% increase in trade openness variable will decrease
the unemployment rate by 0.41985845% in the long run, cateris paribus. The GDP variable
also has the same effect in the short term, which negatively affects the unemployment rate in
the long term with a coefficient of -1.07022944%. Meanwhile, the real interest rate will have
a positive effect, similar to its effect in the short term. More clearly, if there is an increase in
the value of real interest rate by 1%, then in the long run the unemployment rate will grow by
0.3964812%, cateris paribus. The inflation variable in the long run is the same as the
inflation variable in the short run, which negatively affects the unemployment rate in the long
run. In detail, if there is an increase in inflation by 1%, then the unemployment rate will grow
by 0.38646323%, cateris paribus.
4.1 Convergence Result
Convergence is the state of heading towards a meeting point. Converging means
leading to a meeting point. If it is related to the unemployment rate, then what is meant by
unemployment convergence is that whether the unemployment rate in countries in the world,
both in 65 countries as a whole and in the category of countries with low, medium, and high
remittances, will lead to one point or one percentage level (convergent). If the values of the
unemployment rate do not converge to a single percentage point then is called divergence.
Convergence theory states that the level of prosperity experienced by developed and
developing countries will one day converge or meet at the same point (Kiha and Rindayati
2013).
Convergence theory states that there will be a catching up effect, which is a condition
when developing countries manage to "catch up" or in the economic context is to equal the
position or position of developed countries. This theory is based on the assumption that
developed countries will experience a steady state condition, which is a condition where a
country has a constant income growth rate or cannot increase anymore. This can happen
because all production costs have been covered by existing investments, so that additional
savings in countries that have reached the steady state point cannot be used as additional
investment. No additional investment means no additional income. Meanwhile, developing
countries still have a level of investment that can still grow. This means that additional
savings in the country will be used as additional investment and will increase the country's
income. So when the condition of developed countries is relatively no longer moving or at
rest, then developing countries will continue to catch up until at some point the developed
countries will be caught and eventually be at the same level or converge (Quah 1995).
Barro and Martin (1991) state that price convergence is present when there are
significant price differences between regions and time. In the presence of price differences,
the government will carry out policies so that prices between time and between regions can
be suppressed or price convergence can occur. This can also be applied in the context of
unemployment. If the convergence condition is achieved, it can be said that the policy made
by the government has been successful, because unemployment will be at the same point
which follows the point of developed countries with low unemployment rates.
