OVERVIEW OF THE TOPIC
Unemployment and related terms
Unemployment refers to the situation where individuals actively seeking work are unable to secure employment It serves as a key indicator of economic health, with the unemployment rate being the primary measure This rate is calculated by dividing the number of unemployed individuals by the total number of people in the labor force.
Unemployment serves as a critical economic indicator, reflecting the capacity of workers to secure gainful employment and contribute to overall economic productivity A higher number of unemployed individuals results in diminished economic output, while their ongoing need for basic consumption persists, leading to a disconnect between consumption needs and production capabilities Prolonged high unemployment rates can indicate significant economic distress and may trigger social and political instability.
Unemployment is classified into two fundamental types, which are Cyclical unemployment and Natural unemployment
Cyclical unemployment refers to joblessness linked to the fluctuations in the business cycle, occurring when aggregate demand in the economy falls short of providing employment for all job seekers This decrease in demand leads to reduced production and a subsequent need for fewer workers Consequently, cyclical unemployment is at its lowest during economic prosperity and peaks when economic growth begins to decline.
Natural unemployment persists even when the labor market is in equilibrium and the economy achieves its potential output due to objective factors This type of unemployment is classified into four distinct categories.
+ Frictional unemployment occurs when it takes time for workers to search for the jobs that best suit their tastes and skills
+ Structural unemployment focuses on structural problems in the economy and inefficiencies such as a mismatch between the skills of the unemployed workers and the skills needed for the available jobs
+ Seasonal unemployment occurs at seasonal jobs which require working in certain moments of a year
+ Classical unemployment occurs when real wages for a job are set above the market-clearing level, causing the number of job-seekers to exceed the number of vacancies.
Economic theories
Our research aims to identify the key factors influencing the fluctuations in the unemployment rate, with a primary focus on their long-term relationships Previous studies have highlighted several significant long-term determinants of unemployment, including population (POP), inflation rate (INF), Gross Domestic Product (GDP), and Foreign Direct Investment (FDI).
1.2.1 The effect of Population on Unemployment
The size, structure, and quality of a population significantly influence the quantity and quality of human resources Countries with larger populations typically possess more extensive human resources, while the age distribution within a population plays a crucial role in shaping the labor force Although the population serves as the foundation for the labor market, the relationship between population size and labor resources is not always direct; it is influenced by fluctuations in population growth over time Rapid population growth often correlates with rising unemployment rates, highlighting the interconnectedness of these factors.
ASEAN countries are experiencing a significant demographic shift characterized by an aging population As life expectancy rises, the median age across these nations increases, leading to a growing number of elderly individuals This demographic trend results in a stable workforce, which creates a negligible or even negative correlation between population growth and unemployment rates.
1.2.2 The effect of Inflation rate on Unemployment
In 1958, Sir A W Phillips introduced the Phillips Curve, illustrating the relationship between wage inflation and unemployment in the UK from 1861 to 1957 This concept indicated a short-run trade-off, where decreasing unemployment could lead to rising inflation, while reducing inflation might necessitate higher unemployment levels To lower unemployment, the government could boost aggregate demand, which, although it might temporarily increase employment, could also trigger inflationary pressures in labor and product markets Phillips posited that lower unemployment rates resulted in a tighter labor market, compelling firms to raise wages to attract limited labor resources, while higher unemployment diminished this pressure The Phillips Curve encapsulated the average relationship between unemployment and wage behavior throughout the business cycle, indicating the wage inflation rate expected if a specific unemployment level persisted.
1.2.3 The effect of Foreign Direct Investment on Unemployment
Foreign Direct Investment (FDI) refers to investments made by individuals or firms in one country into business interests in another, often involving the establishment of foreign operations or acquisition of assets A key objective of FDI is to lower production costs by employing local workers, which not only boosts job creation but also enhances workforce training opportunities This relationship highlights that a skilled workforce is crucial for attracting FDI As a result, local residents benefit from improved job access, contributing to a decrease in unemployment rates, thereby reinforcing the connection between FDI and reduced unemployment.
1.2.4 The effect of Gross Domestic Product on Unemployment
Gross Domestic Product (GDP) is a crucial indicator of a country's economic performance This article explores the relationship between unemployment and GDP, referencing Okun’s Law, which suggests an inverse correlation between the two Research indicates that this relationship generally holds true across various regions and the nation as a whole However, the magnitude of Okun’s coefficients varies significantly, largely due to regional differences in productivity growth Consequently, effective policy solutions should integrate both conventional aggregate demand or supply management strategies and region-specific approaches.
MODEL SPECIFICATION
Methodology in the study
2.1.1 Method to collect and analyze the data
This study utilizes high-precision secondary panel data sourced from the World Bank, covering a 19-year period from 2000 to 2018 The dataset includes key macroeconomic variables such as unemployment rate, inflation rate, population, GDP, and FDI across ten ASEAN countries: Brunei Darussalam, Cambodia, Indonesia, Lao PDR, Malaysia, Myanmar, Philippines, Singapore, Thailand, and Vietnam.
Our group has used STATA to analyze the dataset and interpret the correlation matrix between the dependent variable and other explanatory variables
2.1.2 Method to derive the model
This research employs OLS regression for panel data to analyze the statistical relationship between a dependent variable and multiple explanatory variables Specifically, it examines how the unemployment rate is influenced by factors such as the inflation rate, population, GDP, and FDI.
There are three types of panel regression models that are commonly used, which are Pooled OLS (POLS) model, Fixed Effects (FE) model and Random Effects (RE) model
In this section, we will analyze our panel data using the regression method, employing three distinct models: Pooled Ordinary Least Squares (POLS), Fixed Effects (FE), and Random Effects (RE) We will determine the most appropriate model for interpretation based on our analysis.
Theoretical model specification
Our research team has developed a function to examine the connections between the unemployment rate and various macroeconomic variables, as well as the impact of these variables on the dependent variable.
GDP: Gross Domestic Product (Current USD)
FDI: Foreign Direct Investment net inflows (Current USD)
Expected sign of regression coefficient UEM Unemployment rate
INF Inflation rate (measured by the CPI) −
This report examines the relationship between percentage changes in a country's GDP and FDI and their impact on the unemployment rate To analyze this data effectively, we will apply the natural logarithm transformation to the two independent variables: GDP and FDI.
To analyze the impact of macroeconomic factors on the unemployment rate, our group has chosen to utilize regression analysis as our primary model, in line with established economic theories.
𝛃 𝟏 : the intercept term of the model
𝛃 𝟐 : the regression coefficient of “population” 𝐏𝐎𝐏 𝐢𝐭
𝛃 𝟑 : the regression coefficient of “inflation rate” 𝐈𝐍𝐅 𝐢𝐭
𝛃 𝟒 : the regression coefficient of “natural logarithm of GDP” 𝐥𝐨𝐠𝐆𝐃𝐏 𝐢𝐭
𝛃 𝟓 : the regression coefficient of “natural logarithm of FDI” 𝐥𝐨𝐠𝐅𝐃𝐈 𝐢𝐭
𝐚 𝐢 : all unobserved and time-constant factors that affect 𝐔𝐄𝐌 𝐢𝐭
2.2.3 Description of the data a Data sources
The panel dataset, sourced from the official World Bank website, comprises 190 observations spanning ten ASEAN countries—Brunei Darussalam, Cambodia, Indonesia, Lao PDR, Malaysia, Myanmar, the Philippines, Singapore, Thailand, and Vietnam—over a 19-year period from 2000 to 2018 This dataset provides a comprehensive statistical description of the variables involved.
To prepare for the analysis of a panel dataset in STATA, we begin by executing the command "xtset country year," which specifies the cross-sectional variable (country, n = 10) and the time-series variable (year, T = 19) This step is essential for declaring the dataset as a balanced panel, ensuring that all 10 ASEAN countries have complete measurements across all time periods.
We execute the command xtsum UEM POP INF logGDP logFDI to analyze the dataset, yielding results that encompass the number of observations (Obs), average value (Mean), standard deviation (Std Dev.), and the minimum (Min) and maximum (Max) values, as shown in the table below.
Variable Mean Std Dev Min Max Obs
INF overall 4.850346 6.917394 -2.314972 57.07451 N = 190 between 3.935656 0.3280127 13.9202 n = 10 within 5.816883 -9.179019 48.00466 T = 19 logGDP overall 24.90271 1.642864 21.27208 27.67233 N = 190 between 1.599855 22.46288 26.93039 n = 10 within 0.6190561 23.45635 26.05089 T = 19 logFDI overall 21.51932 1.904104 15.30871 25.30585 N = 190 between 1.642618 19.02659 24.28045 n = 10 within 1.088296 17.69944 23.74255 T = 19
There are three different types of statistics: overall, between, and within, in which “overall” statistics are ordinary statistics that are based on 190 observations;
The "between" statistics are derived from summary data across 10 countries without considering the time period, whereas the "within" statistics are calculated from summary data over 19 time periods irrespective of the country Additionally, a correlation matrix is utilized to analyze the relationships between the variables.
Run the command corr UEM POP INF logGDP logFDI to analyze the correlation between the variables, we have the result is the table of correlation matrix between variables:
UEM POP INF logGDP logFDI
According to the Correlation matrix between variables:
The correlation coefficient between population (POP) and unemployment rate (UEM) is 0.2549, indicating a positive but moderate relationship This suggests that fluctuations in population levels can influence changes in the unemployment rate.
The correlation coefficient between inflation (INF) and unemployment (UEM) is -0.2241, indicating a weak negative relationship This suggests that fluctuations in the inflation rate will result in only minor inverse changes in the unemployment rate.
The correlation coefficient of 0.2151 between logGDP and UEM indicates a positive yet weak relationship This suggests that fluctuations in GDP will result in minor changes in the unemployment rate.
The correlation coefficient between logFDI and UEM is -0.0902, indicating a negative and very weak relationship This suggests that foreign direct investment (FDI) has a minimal impact on unemployment, with any changes in FDI resulting in only slight fluctuations in the unemployment rate.
PANEL DATA ANALYSIS
Choosing the most suitable model
3.1.1 Breusch-Pagan Lagrange Multiplier Test (LM)
To determine whether to use the Pooled OLS model or the Random Effects/Fixed Effects model, we will conduct the Breusch-Pagan test to assess the significance of differences across units (panel effect) or the presence of heterogeneity.
State the Hypotheses: {H 0 : a i does not exist or Var(a i ) = 0
To perform the Breusch – Pagan test on STATA, we will run these two following commands: xtreg UEM POP INF logGDP logFDI, re xttest0
We have the result: chibar2(01) = 1060.10
The p-value obtained from the test is (Prob > chibar2) = 0.0000 < 0.05, we can reject H0, accept H1 Then a i does exist, we should choose RE/FE model rather than using POLS model
We will check if the Random Effects model or the Fixed Effects model is more efficient for the panel data collected by using the Hausman test The RE model will be chosen if a i are correlated with x it : cov(a i , x it ) = 0, otherwise, if a i are not correlated with x it , the FE model is more efficient to choose
State the Hypotheses: {H 0 : cov(a i , x it ) = 0
To conduct the Hausman test in STATA, execute the following commands: first, run "xtreg UEM POP INF logGDP logFDI, fe" and store the fixed effects results with "est sto fe." Next, run "xtreg UEM POP INF logGDP logFDI, re" to obtain the random effects results Finally, apply the Hausman test with the command "hausman fe" to compare the two models.
We have the result: chi2(3) = 53.43
The p-value from the test is 0.0000, which is less than the 0.05 significance level, allowing us to reject the null hypothesis (H0) and accept the alternative hypothesis (H1) This indicates a correlation between the variables a and x Consequently, the Fixed Effects model is the most appropriate model to utilize in this analysis.
Diagnostic testing the problems of the model
3.2.1 Diagnosing the problem of Multicollinearity
Multicollinearity is a statistical issue in regression analysis where one independent variable can be accurately predicted from others, indicating a strong correlation among explanatory variables This phenomenon can complicate the interpretation of regression coefficients and affect the reliability of the model's results.
To identify multicollinearity in our model, we utilize the command "corr UEM POP INF logGDP logFDI" to generate a correlation matrix among the variables, yielding the following results.
UEM POP INF logGDP logFDI
The correlation matrix reveals a strong relationship between logFDI and logGDP, with a correlation coefficient of 0.8485, indicating potential multicollinearity in the model.
To make sure if the multicollinearity is existing in our model, we will use variance inflation factors (VIF) to help detect multicollinearity by using these following commands: reg UEM POP INF logGDP logFDI vif
The result obtained as following:
The variance inflation factors (VIF) for each explanatory variable, as well as the mean VIF, are all below 10, indicating that our model is not experiencing multicollinearity issues at this time.
3.2.2 Diagnosing the problem of Heteroskedasticity
The most likely deviation from homoskedastic errors in the context of a panel data is likely to be error variances specific to the cross-sectional unit We will use the Wald test to calculates a modified Wald statistic for groupwise heteroskedasticity in the residuals of the Fixed Effects regression model
To perform the Wald test on STATA, we will run these following commands: xtreg UEM POP INF logGDP logFDI, fe xttest3
We have the result: chi2(10) = 1285.94
The p-value obtained from the test is (Prob > chi2) = 0.0000 < 0.05, we can reject H0 and temporarily accept H1 We can conclude that our model is heteroskedastic at a significance level of 5%
3.2.3 Diagnosing the problem of Autocorrelation d Testing for serial correlation with the Wooldridge’s test
Serial correlation, also known as autocorrelation, refers to the phenomenon where error terms in a time series are carried over from one period to the next The most prevalent type of autocorrelation is first-order serial correlation, which can manifest as either positive or negative.
State the Hypotheses: {H 0 : no first − order autocorrelation
To implement the Wooldridge test for serial correlation in the panel data, we use the following command: xtserial UEM POP INF logGDP logFDI
The p-value obtained from the test is (Prob > F) = 0.0012 < 0.05, we can reject
H0 and temporarily accept H1 at a significance level of 5% Therefore, our panel dataset is having the problem of serial correlation or autocorrelation e Testing for cross-sectional correlation with Breusch – Pagan LM test
In panel data models, such as xtreg, it is typically assumed that error terms are independent across cross-sections, which serves identification purposes rather than ensuring descriptive accuracy Given our dataset's characteristics of large time periods (T = 19) and a small number of cross-sections (N = 10), we utilize the LM test statistic developed by Breusch and Pagan to evaluate this assumption.
(1980) can be used to test for cross-sectional dependence
State the Hypotheses: { H 0 : cross sectional dependence
To conduct the Breusch-Pagan LM test for contemporaneous correlation in STATA, we utilize the command xttest2, which tests the null hypothesis of dependence among residuals If the null hypothesis is rejected, it indicates that the test found no cross-sectional dependence in the residuals The command to run this test following a fixed effects regression is xtreg UEM POP INF logGDP logFDI, fe, followed by xttest2.
We have the result: chi2(45) = 157.506
The p-value from the test is Pr = 0.0000, which is less than the significance level of 0.05 This allows us to reject the null hypothesis (H0) and accept the alternative hypothesis (H1) Consequently, our panel dataset shows no dependence among the residuals, indicating that there is no issue of cross-sectional correlation.
In conclusion, our model is having two problems of heteroskedasticity and serial
SECTION 4 ESTIMATED FIXED EFFECTS MODEL
Correcting the model
Our model encounters issues of heteroskedasticity and serial correlation, making clustered standard errors a more reliable option While the use of clustered standard errors does not alter the coefficient estimates, it does affect the t-statistics due to the changes in standard errors This approach provides more accurate p-values for the model.
We will use the following command for the cluster option in STATA: xtreg UEM POP INF logGDP logFDI, fe cluster(country)
The estimated results of fixed effects model
We have the estimated result of Fixed Effects regression and Fixed Effects regression using clustered standard errors as the table belows:
Explanatory variables FE model FE cluster(country) model
* siginificant at 𝛼 = 10%, ** siginificant at 𝛼 = 5%, *** siginificant at 𝛼 = 1%
The values of t are in parentheses
According to the estimated result from Fixed Effects regression using clustered standard errors, we obtained the SRF of the FE model as below:
Meanings of estimated results
The regression coefficient of POP is estimated to be 𝛃̂
𝟐 = −4.72e − 08: Holding other explanatory variables unchanged, if the population (POP) increases by 1000 people, the expected value of unemployment rate (UEM) will decrease by 0.0000472%
The regression coefficient of INF is estimated to be 𝛃̂
𝟑 = −0.0055513: Holding other explanatory variables unchanged, if the inflation rate (INF) increases by 1%, the expected value of unemployment rate (UEM) will decrease by −0.0055513%
The regression coefficient of logGDP is estimated to be 𝛃̂
𝟒 = −0.0608111: Holding other explanatory variables unchanged, if GDP increases by 1%, the expected value of unemployment rate (UEM) will decrease by 0.000608111%
The regression coefficient of logFDI is estimated to be 𝛃̂
𝟓 = −0.0640177: Holding other explanatory variables unchanged, if FDI increases by 1%, the expected value of unemployment rate (UEM) will decrease by 0.000640177%
The R² value of 0.2615 indicates that 26.15% of the variation in the unemployment rate (UEM) within a country can be attributed to the explanatory variables, including population (POP), inflation rate (INF), GDP (logGDP), and foreign direct investment (FDI) (logFDI).
R 2 (between) = 0.0844 means 8.44% the total variation in the dependent variable between household units is explained by the explanatory variables
R 2 (overall) = 0.0645 means 6.45% of the total variation in the dependent variable is explained by the explanatory variables
In the Fixed Effects model, Rho represents the proportion of variance attributed to the individual-specific term, indicating how much variation is explained by the constant term that remains unchanged over time Specifically, 1.87% of the variance is accounted for by the error term, while a significant 98.13% is explained by the constant term.
Hypothesis Testing
4.4.1 Testing the significance of an individual regression coefficient 𝜷̂
According to the results from STATA using the Fixed Effect regression, we obtained the confidence interval for the regression coefficients of each variable at a significance level of 5% as below:
The analysis reveals that the variable POP has a regression coefficient that is statistically significant at a 5% significance level, as the value of 0 falls outside the confidence interval of [−7.37e−08; −2.07e−08] In contrast, the remaining variables, INF, logGDP, and logFDI, do not demonstrate similar significance.
The presence of 0 within the confidence interval for each variable indicates insufficient evidence to reject the null hypothesis (H0) Consequently, the regression coefficients for these variables are not statistically significant at a 5% significance level This analysis utilizes the T-distribution approach to assess the significance of the results.
n: the number of observations or sample size, n = 190
α: the significance level, α = 0.05, for the two-tailed test, α
⁄ = 0.025 2 According to the test statistic t s = β ̂ j −0 SE(β ̂ j ) of each variable at the significance level of 5% obtained from the results, we have:
For the variable POP, its absolute value is |t s | = 4.03 > 1.972, we can reject
H 0 Therefore, the regression coefficient of POP is statistically significant at a significance level of 5%
For the variable INF, its absolute value is |t s | = 0.86 < 1.972, we don’t have enough evidence to reject H 0 Therefore, the regression coefficient of INF isn’t statistically significant at a significance level of 5%
For the variable logGDP, its absolute value is |t s | = 0.28 < 1.972, we don’t have enough evidence to reject H 0 Therefore, the regression coefficient of logGDP isn’t statistically significant at a significance level of 5%
The analysis of the variable logFDI reveals that the absolute value of its t-statistic is |t s| = 0.86, which is less than the critical value of 1.972 As a result, there is insufficient evidence to reject the null hypothesis (H0), indicating that the regression coefficient for logFDI is not statistically significant at the 5% significance level.
The p-value is the lowest significance level at which the Null Hypothesis H 0 can be reject
For the variable POP, its p-value is approximately equal to 0.003, which is less than 0.01 Therefore, the regression coefficient of POP is statistically significant at a significance level of 1%
For the variable INF, its p-value is approximately equal to 0.411, which is more than 0.10 Therefore, the regression coefficient of INF isn’t statistically significant at a significance level of 10%
For the variable logGDP, its p-value is approximately equal to 0.782, which is more than 0.10 Therefore, the regression coefficient of logGDP isn’t statistically significant at a significance level of 10%
For the variable logFDI, its p-value is approximately equal to 0.413, which is more than 0.10 Therefore, the regression coefficient of logFDI isn’t statistically
In summary, our analysis using three methods to evaluate the significance of individual regression coefficients reveals that only the POP coefficient is statistically significant at the 1% level, whereas the coefficients of all other variables lack statistical significance, even at the 10% level.
4.4.2 Testing the significance of the overall model
H 1 : R 2 ≠ 0 a The F-test of significance approach
The F-test of overall significance indicates whether your linear regression model provides a better fit to the data than a model that contains no explanatory variables
Specify the critical F-value F c = F n−k k−1 = F 185 4 ≈ 2.42, where:
n: the number of observations or sample size, n= 190
(5−1) ( 1−0.0645 )= 3.19 > 2.42, we can reject H 0 and accept H 1 Therefore, the overall model is statistically significant at a significance level of 5% b The p-value approach
The analysis conducted using the Fixed Effect model in STATA reveals a P-value of P(F s > F c ) = 0.0000, which is less than the 0.05 threshold Consequently, we reject the null hypothesis (H 0) and accept the alternative hypothesis (H 1), confirming that the overall model is statistically significant at the 5% significance level.
In conclusion, after testing the significance of the whole model, we can conclude that the overall model is statistically fitted at a significance level of 5% Theoretically, it is shown that:
When the population increases, the unemployment rate is expected to remain unchanged or even decrease, and vice versa, holding other variables remain unchanged
When the inflation rate increases, the unemployment rate is expected to decrease, and vice versa, holding other variables remain unchanged
When the GDP increases, the unemployment rate is expected to decrease, and vice versa, holding other variables remain unchanged
When the FDI increases, the unemployment rate is expected to decrease, and vice versa, holding other variables remain unchanged
Meanwhile, the estimated coefficients of each explanatory variable are:
The estimated coefficient of population 𝛃̂
The estimated coefficient of inflation rate 𝛃̂
The estimated coefficient of natural logarithm of GDP 𝛃̂
The estimated coefficient of natural logarithm of FDI 𝛃̂
The regression model results align with economic theories, indicating consistency in findings Notably, the regression coefficient for population (POP) is statistically significant at the 1% level, while the coefficients for other variables lack statistical significance even at the 10% level Despite this, all explanatory variables will be retained in the model.
An increase in population can influence the unemployment rate, particularly in ASEAN countries where it often remains stable or decreases As life expectancy rises, the median age of the population also increases, leading to a higher number of elderly individuals Consequently, the workforce remains relatively unchanged, resulting in a minimal or even negative correlation between population growth and unemployment rates.
In addition, there is always a trade-off between inflation and economic growth leads to economic growth It is conventionally measured as the increase in real GDP
As the scale of the production rises, firms will employ more workers leading to a fall in unemployment
Foreign Direct Investment (FDI) can negatively impact the unemployment rate by creating job opportunities for local residents Increased foreign investment often leads to easier access to employment, contributing to a reduction in unemployment For instance, in Vietnam, the period from 2017 to 2018 saw a decline in the unemployment rate from 2.05% to 1.99%, driven by simultaneous growth in population, inflation, GDP, and FDI.
Recommendations
Unemployment is inevitable, we can not delete it but only reduce it
Our group has discussed and recommended some solutions to solve the problems of unemployment in some ASEAN countries
To boost Foreign Direct Investment (FDI), it is essential to develop innovative technologies and enhance the skills of local workers, making the region more appealing to foreign investors An increase in FDI not only attracts investment but also creates more job opportunities for local residents, ultimately reducing the unemployment rate.
To reduce the unemployment rate, we may need to accept a certain level of inflation, as an increase in output typically leads to lower unemployment With more individuals employed, consumer spending rises, which can trigger demand-pull inflation and elevate price levels This trade-off requires careful consideration.
Our research analyzed the statistically dependent relationship between the unemployment rate and key economic factors such as population, inflation rate, Gross Domestic Product (GDP), and Foreign Direct Investment (FDI) The findings align with established economic theories and previous studies, indicating a negative correlation between these factors and the unemployment rate Specifically, as population, inflation rate, GDP, and FDI increase, the unemployment rate tends to decrease.
The report reflects the collective effort of our group and the knowledge acquired in class Despite facing challenges in data collection and understanding, we committed ourselves to learning about econometric models, analyzing their components, and exploring the relationships between variables While we acknowledge that further analysis of additional variables is necessary for a comprehensive understanding, our experience has significantly enhanced our knowledge of the process.
We extend our heartfelt gratitude to Ph.D Dinh Thi Thanh Binh for your invaluable guidance and suggestions that have steered us in the right direction for completing our report Acknowledging that there are still several omissions and errors, we eagerly welcome your feedback to enhance our report to its fullest potential.
1 Cashell, W B., 2004, “Inflation and unemployment: What is the connection?”, Federal Publications
2 Damodar N Gujarati, and Dawn C Porter, “Basic Econometrics”, 5th Edition
3 El-Agrody, N M., Othman, A Z., & Hassan, M B.-D., 2010, “Economic Study of Unemployment in Egypt and Impacts on GDP”
4 Lui, L Q., 2009, “Inflation and Unemployment: The roles of goos and labor market institution”
5 Shu-Chen Chang, 2006, “The dynamic interactions among foreign direct investment, economic growth, exports and unemployment: Evidence from Taiwan”
6 Marjetka Troha, 2015, “Impact of Population ageing on Unemployment and
Entrepreneurial activity: The case of Slovenia”
7 M Palát, 2011, “The impact of Foreign direct investment on Unemployment in Japan”
8 N Gregory Mankiw, “Principles of Macroeconomics”, 6th Edition
1 Elvis Picardo, May 2019, “How Inflation and Unemployment are related” https://www.investopedia.com/articles/markets/081515/how-inflation-and- unemployment-are-related.asp
2 Jim Chappelow, May 2019, “Unemployment” https://www.investopedia.com/terms/u/unemployment.asp
3 Ryan Furhmann, May 2019, “Okun's Law: Economic Growth and Unemployment” https://www.investopedia.com/articles/economics/12/okuns-law.asp
4 Tejvan Pettinger, May 2017, “Trade off between unemployment and inflation” https://www.economicshelp.org/blog/571/unemployment/trade-off-between- unemployment-and-inflation/
The dataset of ten ASEAN countries during 2000 – 2018
Obs Country Year UEM INF POP GDP FDI
Do-file
To analyze the dataset located at "C:\Users\admin\Downloads\datafinal.dta," first import the data and encode the country variable Set the data as a panel dataset using the command `xtset country year`, followed by summarizing the variables with `xtsum uem pop inf loggdp logfdi` Assess correlations among the variables using `corr uem pop inf loggdp logfdi` Choose between a Pooled OLS or Random Effects (RE) model with `xttest0`, then run a Fixed Effects (FE) regression with `xtreg uem pop inf loggdp logfdi, fe` and store the results Next, compare the FE and RE models using the Hausman test with `hausman fe` To check for multicollinearity, use the `vif` command after running a regression Install the necessary packages for testing heteroskedasticity and autocorrelation with `ssc install xttest3` and `xtserial uem pop inf loggdp logfdi`, respectively Finally, detect cross-sectional correlation using `ssc install xttest2`, and run the FE regression with clustering by country using `xtreg uem pop inf loggdp logfdi, fe cluster(country)`.
The STATA command’s outputs
The ouput of the command xtset country year
Figure 1 Panel dataset declared result delta: 1 unit time variable: year, 2000 to 2018 panel variable: country (strongly balanced) xtset country year
The ouput of the command xtsum UEM POP INF logGDP logFDI
The ouput of the command corr UEM POP INF logGDP logFDI:
The correlation matrix reveals significant relationships among various economic variables, including log foreign direct investment (logfdi), log gross domestic product (loggdp), inflation (inf), population (pop), and unemployment (uem) With a total of 190 observations, the data indicates that logfdi has an overall mean of 21.51932 and a standard deviation of 1.904104, while loggdp shows a mean of 24.90271 with a standard deviation of 1.642864 Inflation exhibits a mean of 4.850346 and a standard deviation of 6.917394, highlighting its variability The population variable has a mean of 5.89e+07 and a standard deviation of 6.88e+07, indicating a substantial range in population figures Finally, the unemployment rate (uem) shows a mean of 2.959126 with a standard deviation of 2.148455, suggesting diverse unemployment levels across the dataset These insights are crucial for understanding the interplay between these economic indicators.
corr uem pop inf loggdp logfdi
The ouput of the command xtreg UEM POP INF logGDP logFDI, re:
Figure 4 Random Effects regression result
The ouput of the command xttest0:
Figure 5 Breusch-Pagan Lagrange Multiplier Test result rho 92044805 (fraction of variance due to u_i) sigma_e 6358061 sigma_u 2.1627131
_cons 9.875268 2.602501 3.79 0.000 4.774459 14.97608 logfdi -.0826315 0777981 -1.06 0.288 -.235113 06985 loggdp -.1507737 15068 -1.00 0.317 -.4461011 1445538 inf -.005936 0087629 -0.68 0.498 -.023111 0112389 pop -2.30e-08 7.32e-09 -3.14 0.002 -3.73e-08 -8.64e-09 uem Coef Std Err z P>|z| [95% Conf Interval] corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.0000 Wald chi2(4) = 39.32 overall = 0.0648 max = 19 between = 0.0900 avg = 19.0 within = 0.2391 min = 19 R-sq: Obs per group:
Group variable: country Number of groups = 10 Random-effects GLS regression Number of obs = 190 xtreg uem pop inf loggdp logfdi, re
Estimated results: uem[country,t] = Xb + u[country] + e[country,t]
Breusch and Pagan Lagrangian multiplier test for random effects
The ouput of the command xtreg UEM POP INF logGDP logFDI, fe:
Figure 6 Fixed Effects regression result
The ouput of the command hausman fe:
The ouput of the command vif (after using reg):
F test that all u_i=0: F(9, 176) = 185.41 Prob > F = 0.0000 rho 981256 (fraction of variance due to u_i) sigma_e 6358061 sigma_u 4.6002844
_cons 8.65802 2.437831 3.55 0.000 3.846877 13.46916 logfdi -.0640177 0744515 -0.86 0.391 -.2109502 0829149 loggdp -.0608111 147047 -0.41 0.680 -.3510135 2293913 inf -.0055513 0083598 -0.66 0.508 -.0220496 0109471 pop -4.72e-08 8.98e-09 -5.26 0.000 -6.49e-08 -2.95e-08 uem Coef Std Err t P>|t| [95% Conf Interval] corr(u_i, Xb) = -0.8911 Prob > F = 0.0000 F(4,176) = 15.58 overall = 0.0645 max = 19 between = 0.0844 avg = 19.0 within = 0.2615 min = 19 R-sq: Obs per group:
Group variable: country Number of groups = 10 Fixed-effects (within) regression Number of obs = 190 xtreg uem pop inf loggdp logfdi, fe
Test: Ho: difference in coefficients not systematic
B = inconsistent under Ha, efficient under Ho; obtained from xtreg b = consistent under Ho and Ha; obtained from xtreg logfdi -.0640177 -.0826315 0186138 loggdp -.0608111 -.1507737 0899625 inf -.0055513 -.005936 0003848 pop -4.72e-08 -2.30e-08 -2.42e-08 5.20e-09 fe Difference S.E.
When analyzing coefficients, ensure that they are on a similar scale to avoid unexpected estimations Consider scaling your variables accordingly It's crucial to verify that the output aligns with your expectations, as discrepancies may lead to complications in test computations Additionally, note that the rank of the differenced variance matrix does not match the number of coefficients being tested The Variance Inflation Factor (VIF) values for the variables indicate potential multicollinearity issues, with specific values highlighting the relationships between them.
The ouput of the command xttest3:
Figure 9 Modified Wald Test result
The ouput of the command xtserial UEM POP INF logGDP logFDI:
Figure 10 Wooldridge Test result for autocorrelation
The ouput of the command xttest2:
Figure 11 Breusch-Pagan LM Test for cross-sectional correlation
Prob>chi2 = 0.0000 chi2 (10) = 1285.94 H0: sigma(i)^2 = sigma^2 for all i in fixed effect regression model Modified Wald test for groupwise heteroskedasticity xttest3
Prob > F = 0.0012 F( 1, 9) = 21.404 H0: no first-order autocorrelation Wooldridge test for autocorrelation in panel data xtserial uem pop inf loggdp logfdi
Based on 19 complete observations over panel units
Breusch-Pagan LM test of independence: chi2(45) = 157.506, Pr = 0.0000