ABSTRACT

There have been little or no empirical studies on the impact of foreign investment, financial openness, trade openness, unemployment and tax revenue on income between the Top 20% (T20) and Bottom 20% (B20) level of income. This paper investigates this impact with the use of panel data during the years of 2013-2018. The panel data consist of 60 countries that were randomly selected among the developed and developing countries. The Hausman test was conducted to determine the appropriate model and conclusion was arrived at from the three models that fixed effect model is more preferable as the probability value from the different income group were greater than the 5% chosen level of significance, thus the null hypotheses that random effect is preferable was rejected. The result obtained from pool regression, fixed effect set the basis of findings for this study and also serves as the basis for policy recommendations. It was revealed that foreign direct investment and financial openness have effect on Top 20% (T20) and Bottom 20% (B20) level of income across the selected countries. Tax revenue is the only variable that has a positive and significant effect on the Top 20% (T20) level of income but a negative although, insignificant impact on the Bottom 20% (B20).

1         CHAPTER 1: RESEARCH OVERVIEW

1.1        Introduction

This study aims at finding out the effect of foreign investment, financial openness, unemployment rate, trade openness and government tax revenue on different levels of income among the 60 selected countries. Chapter 1 introduces the background, problem, significance and objective of the research, which focuses on specific or different levels of income instead of aggregate income.

1.1        Research Background

Different levels of income resulting from the same economic input and policies in the developing and underdeveloped world have been seen for a very long time as a major source of concern to researchers, economists and policymakers. Either in a country or across the countries in the world, a lot of factors that are likely to contribute to different income level has been identified. It is worthy to note that there are merits and demerits that await the population of a country that increases her trade openness.

Majority of the aspects of globalization have been revealed as causes of different levels of income, particularly when it is being examined from both the functional and primary distribution of income. Nonetheless, research on this same issue focuses on aggregate indicators. These indirectly assume the factors that can affect the Top 20% (T20) and Bottom 20% (B20) level of income. Over the time, the focus has been shifted to examining the effect of trade openness on specific levels of income but despite this, it has been global finance and technical change that has attracted much attention.

Over the years, foreign investment, financial and trade openness have grown and extended across the world. Although, they are regarded as development facilitators, their effect on income different across all levels. This may be less apparent particularly at the end of the bottom income earner. In deciding the economic development of a country- reflected by income, productivity and production levels, foreign trade, openness to trade- capital mobility has been said to play important roles. As openness to trade is considered to be contributing immensely to industrialization and jobs generation, foreign investment has also been found to be offering the same opportunities in terms of income growth and local production. The explanation is that trade raises competitiveness at the national and international level, which in turn affects the efficiency of an economy and is also expected to have reflected on the earnings of all workers across various levels (Lim & McNelis, 2014).

Global exchange flows rose from 8.3 trillion dollars in 1990 (constant 2000 US$) to 24.7 billion dollars in 2007 (constant 2000 US$). With the rise of foreign venture capital and trade flows, the impact of these factors on the economy of the country has become important to examine. In addition, low-income countries have increased their GDP growth, enabling them to come closer to their constant state.

Previous studies have shown that free trade improves the economy of a region and that exports have a beneficial influence on economic development by improved production and efficiency. on the other hand, scholars have shown that the exports depend on the condition of the host region. Imports boost rivalry, calculated by the level of security.

Recent research by Jaumotte, Lall, & Papageorgiou (2008) has revealed that the impact of globalization was comparatively minor on various income levels, even though the technological change was a major contributor to the widening gap among income distribution in developed and developing countries. This is because liberalization of exchange usually contributes to a decline in income distribution, while financial liberalization generally increases income inequality. Thus, foreign investment, foreign trade and openness seek to combat the discrepancy of income at various levels.

Financial openness, trade openness and foreign investment have been widely regarded as a catalyst to development. Despite this, its benefits have not been evenly felt across all levels of income. The extent of Foreign Investment’s effect on income has also been debated by scholars and researchers across the globe. Prior study indicates that foreign investment has a positive relationship on income, provided there is the availability of skilled labor and functional financial sector. However, some studies also revealed that foreign investment, financial openness and trade to be insignificant on the income level among the developing countries.  (Tebaldi, 2011).

Although some of the available literature has revealed a positive relationship between foreign investment and income distribution, there exist some other literature that revealed a contrary report of a negative relationship. Perhaps, the controversy may partially be as a result of the aggregation of panel analysis framework, data periods, different countries and methodological approaches.

On the impact of foreign investment on different income level, it was reported that wage spillovers may cause financial investment to a negative impact on income distribution as a result of the wage discrimination policy of the multinational corporations to pay wages to skilled workers handsomely than the unskilled workers because of their surplus capital benefits. Therefore, the possibility is that the presence of multinational companies will limit the proportion of local companies and affect their profitability. Local companies are forced to reduce their costs by reducing the level of wages and the number of employees (skilled and unqualified). The implication of this is that higher foreign investments, financial and trade openness will lead to greater poverty (Babatunde, 2018).

Economists and policymakers agreed that foreign investment, financial and trade openness will lead to developing country’s growth and development by transferring new technologies and management expertise, improving human resources and entering the export markets. While foreign investment’s potential position is once again the center of focus in the least developed and developing countries’ phase, several key problems remain unresolved. Among such problems is probably the most complex and contentious of the effects of foreign investment on different income level. Although less developed countries have experienced a rise in inequality in recent decades, economic activities are also rapidly globalized through foreign trade and investment (Ravinthirakumaran and Ravinthirakumaran, 2020).

Evidence has revealed that there are numerous amounts of research that has been carried out by scholars on the impact of financial development on income distribution across all levels. However, the relationship between foreign investment, financial and trade openness and income distribution particularly on the Top 20% (T20) and Bottom 20% (B20) happens to be an area that has gained little or no attention. To set out the uniqueness of this study, the identified gap will be filled with the sole aim of making this study a source of policy recommendations and to contribute to the current empirical literature. This study will try to investigate the effect of foreign investment, financial and trade openness on income across various levels of income through a 60-country panel from 2013 to 2018. This is to establish the forms of relationship that exists between variables.

1.2        Research Problem

Past research on different levels of income focuses on different combinations of effect of foreign investment, financial openness and trade openness. Their results are inconclusive. Nonetheless, past literature only used aggregate income level while current policy trends around the world have changed to targeting specific levels of income. For example, the Top 20% (T20) and Bottom 20% (B20) income levels are policy targets for subsidies provision. In addition, it is doubtful that increase in foreign investment, higher financial openness and increase in trade will benefit equally the higher income group of capitalists and lower income group of unskilled labor.

Therefore, research on income should be more oriented to specific or different levels of income and not just aggregate income. In doing so, it will give better insights for policymakers in their new trend of policymaking which targets a specific income level group.

 

1.3        Research Objectives

Generally, this research aims to analyze the effect of foreign investment, financial openness, trade openness, government tax and unemployment to various levels of income.

Specifically, this research has five objectives:

(i)     To determine the relationship between foreign investment and Top 20% and Bottom 20% levels of income;

(ii)    To determine the relationship between financial openness and Top 20% and Bottom 20% levels of income;

(iii)      To determine the relationship between trade openness and Top 20% and Bottom 20% levels of income;

(iv)       To determine the relationship between government tax revenue and Top 20% and Bottom 20% levels of income and

(v)        To determine the relationship between unemployment and Top 20% and Bottom 20% levels of income;

1.4        Hypothesis of the Study

A hypothesis is a statement of claim which illustrates the relationship between two or more variables of interest. The null and alternative hypothesis for this paper is determined as follows:

1.4.1        Foreign investment

𝐻1₀: Foreign investment has no significant relationship with Bottom 20% and Top 20% levels of income.

𝐻1₁: Foreign investment has a significant relationship with Bottom 20% and Top 20% levels of income.

 

1.4.2        Financial openness

𝐻2₀: Financial openness has no significant relationship with Bottom 20% and Top 20% levels of income.

𝐻2₁: Financial openness has a significant relationship with Bottom 20% and Top 20% levels of income.

 

1.4.3        Trade openness

𝐻3₀: Trade openness has no significant relationship with Bottom 20% and Top 20% levels of income.

𝐻3₁: Trade openness has a significant relationship with Bottom 20% and Top 20% levels of income.

 

1.4.4        Government Tax

𝐻4₀: Government tax has no significant relationship with Bottom 20% and Top 20% levels of income.

𝐻4₁: Government tax has a significant relationship with Bottom 20% and Top 20% levels of income.

 

1.4.5         Unemployment

𝐻5₀: Unemployment has no significant relationship with Bottom 20% and Top 20% levels of income.

𝐻5₁: Unemployment has a significant relationship with Bottom 20% and Top 20% levels of income.

1.5        Research Significance

By using panel data techniques, with observations across 60 countries and over the period of 2013-2018, in this research, the main significance is to focus on the impact of the macroeconomic variables — FDI, financial openness, trade openness, unemployment rate and government tax that may affect various levels of income.

Researchers in the past were only studied the impact of various variables to the aggregate income level; they were ignoring the specific income levels like Top 20%, Bottom 20% and others (T20, B20 and others). Thus, this study will reinforce previous findings by targeting different income levels. The gap identified from the review of literatures and research problems, shall set basis for recommendations and implications that would be suggested at the end of this research.

This study is considered to contribute immensely in shaping the perspective of economists, researchers and policymakers as regards income across all levels. Factor affecting overall (aggregate) income, top level income (the richest group) and bottom level income (the poorest group) may be different or at different magnitude. For example, trade openness may benefit the Top 20% (T20) through export earning but harm the Bottom 20% (B20) through suppress wages in export sector and higher import price due to policy to maintain competitive (lower) domestic currency value. High unemployment may reflect negatively on Bottom 20% (B20) but may not to Top 20% (T20).

 

In short, by studying the relationship between foreign investment, financial openness, trade openness, unemployment rate, government tax and various levels of income in this research, this paper will provide a good framework and suggestions for future policymakers. Thus, they care able to make their policy implementation decisions better and respond to any economic changes.

2         CHAPTER 2: LITERATURE REVIEWS

2.1        Introduction

Chapter 2 will link the variables to the underlying theories and explain the relationship between income with FDI, financial openness, trade openness, government tax and also unemployment rate. Other than that, a proposed theoretical framework and hypothesis development will be present in section 2.3 and 2.4. The last section will provide a direction to proceed with next chapter by summarizing the literature review from the journals and past research.

2.2        Reviews of Literature

By studying the past research, the relationship between various income levels with FDI, financial openness, trade openness, government tax and also unemployment rate are being discussed in this part.

2.2.1        Dependent Variable – Income

National income has been conceptualized as the total monetary value of goods and services produced by the residents of a particular country in a given at a given period of time. For the purpose and uniqueness of this study, income has been extensively conceptualized in all possible recognizable ways. In relationship with foreign investment and financial openness, income is looked into from at all levels (T20 & B20) and Gross National Income (GNI). This research mainly studies T20 and B20. T20 refers to the top tier income earners account for 20% of the country’s total income, while B20 are the bottom tier income earners account for 20% of the country’s total income. Gross National Income, an accurate parameter for measuring National income and an accurate parameter for measuring National income is the total amount of income earned in a nation is by her citizen at a particular period of time. The effect of the three classifications of income in this context shall be investigated in relation to foreign direct investment, financial openness, trade openness, unemployment and government tax.

2.2.2        Independent Variable – The FDI

In general, foreign direct investment is deemed to be one of the fewer forces that drives a nation’s economic growth. Previous studies have shown that there exist positive spillover effects of foreign direct investment on national income. In addition, findings from existing literature of different scholars have revealed that foreign direct investment has a positive effect on any economy blessed with skilled labor and sophisticated financial sector. Furthermore, findings of other researchers also validate the fact that foreign direct investment may positively impact the economic growth of developing countries.

Previous studies from the works of different scholars have revealed that positive relationship exists between foreign investment and economic growth of any country. Although, there are numerous studies that find otherwise. In the research work done by (Khawar, 2005) to examine the impact of Foreign Direct Investment on the growth of National income proxied by GDP per capita from 1970 to 1992. The findings of his investigation showed that a strong positive correlation exists between foreign direct investment and per capita GDP growth. More importantly, it was revealed from his findings that an increase in foreign direct investment, resulting in a GDP growth.

In contrast to other studies that find a positive relationship between income and foreign investment, other studies have found out the possibility of a negative relationship between income across all levels and foreign direct investment. Among these findings is the study of  (Wu & Hsu, 2012). This study aims to examine the effect of foreign direct investment between the periods of 1980-2005 on income distribution. The cross-sectional dataset that consists of 54 countries (21 developed countries and 33 developing countries) was employed for the analysis. It was found that foreign direct investment could affect various levels of income distribution of countries whose levels of absorptive capacity is low.

On the same vein, (Sylwester, 2005) carry out research on the relationship that exists between foreign direct investment and income distribution on 29 least developed countries from 1970–1989. The findings of the study revealed no substantial evidence to opine that there exists a relationship between foreign direct investment and income at various levels within this group of selected less developed countries.

Baiashvili and Gattini (2019) focused their research on the growth impact of FDI inflows. They focused on the interaction between country income levels and foreign direct investment as they perceived that previous studies have not extensively analyzed it.

Using panel GMM techniques that are robust to sample size, endogenous issues and instrument proliferation, a total of 111 countries ranging from developed economies to developing and emerging markets beginning in 1980 were examined. The results of their study revealed that benefits from foreign direct investment do not spread evenly across the countries selected. In addition, an inverted-U-shaped relationship between the country’s income level and the growth impact of foreign direct investment was detected.

Wijeweera, Villano, and Dollery, (2010) also added to the current literature on the influence of FDI on National income. The Stochastic Frontier model was employed on 45 countries between the period of 1997 and 2004. The results of the study indicate that when the labor force uses foreign technology, the highly skilled labor of the host country contributes to a positive effect of foreign direct investment inflows on economic growth. when the labor force uses foreign technology. They further suggested that Poor countries can also increase their income rate by encouraging direct foreign investment, strengthening education and reducing corruption.

The impact of foreign direct investment inflows on income inequality in the Asia-Pacific Economic Cooperation was investigated from 1990-2015 by Ravinthirakumaran and Ravinthirakumaran (2016). Trade openness, Gini Coefficients, Foreign Direct Investment inflows, Gross Domestic Product per capita and Human capital constituted the variables used for this research. They conducted the following econometric tests panel heterogeneous non-causality and panel Autoregressive Distributed Lag (ARDL). The findings revealed by the panel ARDL pointed out that foreign Direct Inflows in the long run help in reducing income inequality. Furthermore, the result also revealed that trade openness GDP per capita helps in reducing income inequality to the barest minimum level while differences in human capital have been seen to widen the disparity income inequality.

Dimelis and Papaioannou (2010) show econometric evidence that in the community of developed countries, impact of FDI is significantly positive, although, it appears to be positive, but insignificant in the developing countries.

An OLS regression and a panel regression were performed by (Carkovic & Levine, 2002) from 1960 to 1995 to reassess the relationship that exists between economic growth and FDI. It was found that FDI is insignificant when using OLS regression but important in panel estimation to control financial development or international openness. It can be assumed that FDI has no independent impact on the development of the economy.

The extent of the effect of foreign direct investment on income and economic growth is still being discussed and can never be overstressed. While previous research has shown that the spillover effects of FDI have been positive for economic development, some studies on the contrary also indicate that FDI is insignificant for developing countries’ economic growth. It is on the premise that this study seeks to include the variable as part of the independent variables for this study.

2.2.3        Independent Variable – The financial openness

It is possible to conceptualize financial openness as a means of incorporation into international financial markets. It is one of the variables that change a country’s production structures as well as the methods of doing business.

Financial openness is synonymous with but distinct from that of financial growth. It also opens up to international capital and becomes more closely aligned with global financial structures as a financial system evolves and becomes more complex (Gemma Estrada & Ramayandi, 2015).

Scholars across the globe have done great work as regards the relationship that exists between financial openness and National income. Among these scholars is (Serdaroğlu, 2015) whose studies aimed at analyzing the effects of financial openness on Total Factor Production in Turkey. Empirical results obtained from the studies revealed that financial openness exerts a positive and significant effect on Total Factor Production together with the other determinants of TFP. This study, however, noted that the relationship between financial openness and TFP may present varying results, depending on the sub-periods considered (Serdaroğlu, 2015).

Financial openness appears intuitively to have a positive effect on the growth of any economy. Inflows of foreign direct investment will enhance economic growth by attracting advanced foreign technologies and managerial skills. This will also increase the competitiveness of domestic markets through the entry of foreign companies. By allowing domestic firms to access foreign savings, non-FDI inflows are also likely to contribute to economic growth. Nevertheless, in the absence of a sound and effective financial system, foreign capital inflows may be improperly distributed, resulting in a financial crisis that hinders growth. (Gemma Estrada & Ramayandi, 2015).

Thang Cong Nguyen and Ha (2019), also employed numerous techniques of estimation in their study which was focused on financial openness.  For the period 1961–2017, a total of 21 emerging economies were sampled. Robust findings from this study revealed that an inverted U-curve relationship exists between financial development and income inequality. What this outcome meant was that gap among different levels of income can be widened or contract at the early stage of financial growth after a certain degree has been reached.

In their own research, Belloumi and Alshehry (2020) investigate the effect of openness to trade on sustainable development in Saudi Arabia. To arrive at a robust finding, Autoregressive distributed lag cointegration framework was employed on the annual data that spanned from 1971 to 2016. Importantly, the findings of the study show that a long-term relationship exists between trade openness and growth of an economy. The findings obtained also suggest that openness to trade does not impact short-term economic growth. While trade openness has for a very long term had a considerable negative impact on economic growth. The study concluded by revealing that trade openness could be linked to the depletion of sustainable development in Saudi Arabia over the past fourteen years.

The relation between financial structure and income inequality has recently been researched by (Michael Brei & Gambacorta, 2019). They used panel data from 97 countries over the period 1989-2012 and discovered that the relationship was not monotonic. The results are consistent with well-established literature showing that a strengthened financial structure can help to reduce the gap in income distribution in developing countries. Also, it aligns with recent proof of rising disparity in different economies that are financially advanced.

Huchet-Bourdon, Mouël, Vijil & (2016), carried out research that was focused on proposing a more elaborated way of measuring trade openness. The results of their studies have shown that economies that export goods of higher quality are growing faster.

 

 

 

Significantly, an interesting non-linear trend was found between the ratio of trade dependence and quality trade, showing that trade could have a negative effect on growth for countries that have specialized in low-quality goods. Also, findings from the duo’s work shows that a non-linear relationship exists between export, trade ratios and growth which indicates that countries that exports a variety of goods will grow faster before the economic dependence on trade reaches a certain level.

The relationship between financial growth, financial and trade openness in 29 Asian developing economies from 1994-2008 was investigated by (Hanh, 2010) using GMM estimator. It was found that bidirectional causality exists between financial openness trade openness. In addition, the relationship between financial growth and financial transparency through various measures has been found to be heterogeneous.

In conclusion, this research was conducted with the intention to contribute to earlier studies. With the forgoing literature, there is no time better than now to explore the impact, if any that exists between financial openness and different income levels.

2.2.4        Independent Variable – The trade openness

Openness to trade is conceptualized as the degree to which countries participate with other countries or economies in trading activities. Also, it has been widely regarded as a main determinant of growth to the extent that the developed and developing countries rely more on international trade  (Petrakos, Kallioras, & Ageliki, 2007).

Researchers and economists use export intensity and import penetration as independent variables when evaluating the effect of trade transparency on GDP growth. The strength of exports can be determined by the export share of GDP and the rate of export growth. The penetration of imports means the willingness of foreigners to compete with domestic companies. The penetration of imports is measured by average effective protection rates, GDP shares of imports, import growth rates, and flows of imports. Researchers usually use time series and panel data modeling techniques to examine the relationship between foreign trade and economic development.

Developing countries benefit from free trade because trade with developed countries (a) enables them to use advanced technologies and skills to increase productivity in their home country and (b) increases demand for domestic goods and services produced. Therefore, international trade is a driving force of growth  (Awokuse, 2007).

Some research has confirmed that free trade is advantageous for the economy of a nation. Exports have a beneficial effect on economic growth through the increase of production and productivity. The addition of imports to the growth model improves the model’s accuracy, and previous research has shown that inward trade has a positive effect on economic growth.

The effect of trade openness on income distribution in both developing and developed countries was investigated by (Polpibulaya, 2015). The researcher divided the data used for the study into five years (10 periods) and 86 countries were selected for the survey. In order to observe improvements in trade openness and income distribution during the study period, the investigator also divided the countries into eight regions. To obtain the results that revealed that trade openness leads to a wider gap between various levels of income in overall countries, the ordinary least squared regression approach was used. By running the regressions independently for emerging and developed countries an interesting fact was arrived at: although increasing trade openness widens the gap between the incomes at various levels in developing countries, it also helps to combat income differential in developed countries, although the effect is not substantial.

An analysis was undertaken by Lim and McNelis (2014), to investigate the relationships between the Gini coefficient and trade transparency, assistance and foreign direct investment flows. For the total data collection (42 low to medium income countries), panel data projections are given. The duo’s empirical findings revealed that trade openness proves to very effective in improving income at various levels than either of foreign direct investment or foreign aid, however, its effectiveness depends on the development stage.

The studies of Fida (2016), empirically investigates the impact of trade liberalization on income inequality in China and the Regional Cooperation Countries Association of South Asia. For the period from 1973 to 2012, panel data analysis was performed. The results showed that in these nations, free trade policies have increased income inequality. The independent variables used have distinct impacts on income distribution. Per capita income has a growing influence on income inequality, although it is shown that education, financial growth, financial transparency, democracy and government size minimize differences across all income level.

Paulino (2012), asserted that even if trade liberalization broadened across all income level, Poverty reduction would still not occur, although it leads to positive economic growth.

In testing for Granger causality between variables, and use time series data from Pakistan to analyze the impact of exports and foreign economic performance on overall productivity factors and output growth, Fatima (2002) adopted vector error correction model and a multivariate approach. The results of this study show that, although they are linked in the long term, there are no causal relationship between exports and economic growth in the short run. It is however revealed that foreign economic success affects overall production, but not short-term domestic growth.

Fatima also points out that productivity is affected by the increase in exports and by the rise in external economic performance due to the spillover between countries which can occur without the aid of trade. In addition, the findings from their empirical and simulation results indicate that trade and financial openness can be effective policies in low-income countries to reduce inequality, provided that capital gains are redistributed.

2.2.5        Independent Variable – The unemployment rate

Unemployment and Economic growth are the core factors that all economies select and revolve all economic policies around. Unemployment refers to the existence of a workforce in and out of the working force who is willing to work at the current wage and cannot find a job.

Over time, Economic growth and unemployment are the main parameters that both policy makers and the public observed meticulously in providing a strong view of a country’s economic progress. Economic and financial challenges are the major concerns of policy makers both in the developed or developing countries. Among these challenges are the increasing rate of unemployment and economic growth fluctuation. (Srinivas,2018).

Furthermore, the connection between unemployment and economic growth as a macroeconomic problem covers a broad range of theoretical and empirical study. In economics, it is generally understood that an economy’s higher GDP growth rate raises jobs and decreases unemployment  (Srinivas, 2018)

According to  (Soylu, Çakmak, & Okur, 2018), high rate of economic growth and massive reduction of unemployment are two main priorities of developed and developing country economies. They also pointed out that in terms of the country’s economic performance, economic growth and employment are two extremely important macroeconomic variables, and they are the core elements of many countries’ economic policies.

Tregenna (2011), in their studies examined the relation between unemployment and income inequality in South Africa).  The results of the study indicate that the distribution of employment  in the formal and informal sectors is negligible to explain income inequality, as is the division of incomes within each of these groups. The results indicated that there is greater need to reduce the level of unemployment in South Africa in order to extremely address high level of inequality.

The objective study of the study (Sadikua, Ibraimi, & Luljeta, 2015) was to investigate empirically the relation between economic growth and the unemployment rate by following the law of Okun’s in the FYR of Macedonia. For the study of the coefficient of Okun’s, four types of models are used for quarterly data from 2000 to 2012.

Empirical findings from all models do not show robust evidence and do not support the inverse relationship that exists between the unemployment and economic development, as stated in Okun’s law. On the basis of the VAR methodology and the Engel-Granger cointegration test, there no evidence of causal link between the two variables and it also appears that changes in real GDP growth rate do not affect the unemployment rate and vice versa  (Sadikua, Ibraimi, & Luljeta, 2015).

Soylu, Çakmak, and Okur (2018), also attempt on investigating the relationship that exists between economic growth and unemployment in Eastern European Countries from 1992-2014. They used panel data framework for the study. The results of their study suggest that economic growth and the unemployment rate have remained stable from the beginning, and the unemployment rate has been positively affected by economic growth. In essence, under all conditions remain unchanged, if GDP grows by 1%, the unemployment rate will fall by 0.08%.

2.2.6        Independent Variable – The government tax

Generally speaking, the tax income maximization is incompatible with the Gross Domestic Product maximization. Many economists agree with the assertion that” ‘High taxes are not good for economic growth’ and they often employed the tax multiplier when evaluating this unfavorable correlation. Over the time, Development of growth regressions have been the most widely used by researchers and scholars to investigate the impact of tax revenue on growth.

The impact of a number of external factors on economic development, including taxes and public investment, is measured using sophisticated data sets. It can be inferred, that taxation is directly related to inflation – i.e the higher tax levels and government spending, the lower the amount of growth.

The emergence of the theory of endogenous growth has opened the door to investigate the impact of taxes on economic growth. Explicit simulation of human choices that results to growth facilitates the examination of the tax incidence and the prediction of its effect on growth.

Myles (2000), in his studies reviews the empirical and theoretical evidence to investigate how taxation affects the rate of economic growth. It is shown from the findings of this study that some mechanisms by which taxation can influence economic growth were isolated by theoretical models used in the analysis; however, these effects are very important.

The duo of Đurović-Todorović and Milenković, (2019) conducted a research that was aimed at identifying a potential linear correlation between direct taxes and economic growth between the period of 1996-2016. The conclusions from the matrix correlation indicate that there is a statistically significant link between growth in tax receipts, personal income tax, income tax and gross national product in OECD countries. The pair nevertheless observed that property tax is substantially correlated at the OECD level with gross domestic product. This is rational in view of their low share of tax.

Feng and Suyono (2014) examined the connection between tax revenue and Hebei Province’s economic growth in 1978-2011, based on the tax multiplier effect. The pair employed a basic and updated theory of the effect of tax multipliers and the distributed polynomial lag model. The findings suggest that the negative influence of tax hikes on economic development cannot be as severe as one might imagine and that reduction in tax might produce more favorable consequences for the province of Hebei.

 

2.3        Proposed Theoretical Framework

Figure 2.1 Proposed Theoretical Framework

Figure 2.1 above demonstrates the theoretical framework that will be used to set basis for this research. The research is performed to examine the extent at which the income at various levels be affected by foreign investment, financial openness, trade and government tax revenue across a pool of selected countries. The dependent variable, as shown in the theoretical framework is identified as “Income which for the sake of this study is grouped into the Top 20% (T20) Income, Bottom 20% (B20) Income and Gross National Income”, whereas the independent variables selected for the study are foreign investment, financial openness, trade and government tax revenue. The hypothesis H₁, H₂, H₃, H₄ and H₅ have been made according to the theoretical framework to assume the relationship between every independent variable and dependent variable.

2.4        Conclusion

From the literature reviews, it has been discovered that the foreign investment, trade openness, financial openness, government tax and unemployment rate are factors that may severely affect various levels of income across all economies. The following chapters will discuss about the research methodology.

 

 

3         CHAPTER 3: RESEARCH METHODOLOGY

3.1        Introduction

This chapter introduces the methods of data collection and research. All the models used to analyze the research data and the source from which the data is obtained will be listed in section 3.2 and 3.3.

3.2        Data collection methods

3.2.1        Secondary Data

Data collected for the sake of the study are secondary data. They are obtained from two relevant sources; World Development Indicators and the Chinn-Ito Index. The data for Foreign Investment, Trade openness, Tax and Unemployment rate are collected from World Development Indicators, while the data Financial openness are collected from Chinn-Ito Index. The collection of data periods is covered from 2013 to 2018 in collaboration with the observation of 60 sample sizes in total. Further information from previous researchers is also cited to ensure the accuracy of the unit measurement of each variable.

3.3        Research Instruments and Empirical Methodology

Static Panel through Pool Ordinary Least Square (POLS), Fixed Effect Model (FEM) and Random Effect Model (REM). All data can be available in the data base of Chinn-Ito Index and that of the World Bank’s World Development Indicator.

 

 

3.3.1        Data and Models

The data used for this study is standardized in order to avoid the adverse impact of the variation in the units used to express the independent and dependent variables (Gujarati, 2004). Thus, we identify the factors that can influence that influence income of the Top 20% (T20) and Bottom 20% (B20) across 60 selected countries.

Specifically, the data is analyzed to establish the influence of foreign investment, financial openness, trade openness and government tax with the interactions on income across all levels by means of panel data random effects. The description of this technique has been provided later in this chapter.

The sole aim of this empirical study is to investigate the impact of the five identified independent variables on income across various levels. To have a result worthy of policy recommendations, income that constituted the dependent variable has been consciously split into three categories. They are Top 20% (T20), Bottom 20% (B20) and Gross National Income. What this simply means is that the impact of all the independent variables (Foreign Direct Investment, Financial openness, Trade openness, Unemployment rate and Government tax) shall be separately examined on the income of the Top 20% (T20), Bottom 20% (B20) and GNI.

Before this study proceeds to analyzing this impact, it is necessary to perform some regression assumptions testing. These tests are important so that none of the assumptions of regressions will be violated since it is generally known that any violation of these assumptions may lead to spurious or inaccurate results.

It was stated by  (Stock & Watson, 2015) that testing for linear regression assumptions when dealing with panel data analysis is not necessarily due to the fact that one of the panel data analysis has the advantage of coping better with the issues resulting from the violations of the assumptions of regression. For this reason and some other reasons that is not cited in this study, it will be safe to state that testing for regression assumptions may be omitted. However, performing this test is still of essence in order to have an accurate and reliable result for this study.

Saunders, Lewis, & Thornbill, (2000) suggests the relevant tests necessary to be conducted when testing for violations of regression assumptions. They are itemized as (a) normality test. This test is conducted to test whether data is normally distributed, (b) heteroscedasticity test. This is conducted to know whether the variances of the residuals used in a model are not equal, (c) autocorrelation test. This test is done to check if the residuals of a model depends on each other, and (d) multicollinearity test. This is test is also carried out to know if two or more explanatory variables are strongly correlated with each other in the model. There are some relevant tests that are not itemized by these scholars but this study adopted.

3.3.1.1       Fixed Effects Model (FEM)

Fixed Effects model assumes that with different intercept, differences between individual observations can be accommodated. This estimation technique is often in most times referred to as Least Squares Dummy Variable (LSDV).

Despite the fact that Fixed effect model is different from common effect model, the two model still uses the same ordinary least square assumption. However, any other model that resulted to a constant intercept for each time-series and cross-section data is considered to be less realistic, in this regard, more models are needed so as to capture the difference.

In panel data, the regression equation for fixed effects model is:

Fixed Effect Equation

Description:

for i = 1,2,3,…., n and t = 1,2,3…., T.

Where n = number of observations

t = the number of time.

3.3.1.2       Random Effects Model (REM)

This model is used to estimate panel data in which variables of intervention can be related between periods as well as individuals. In the Random Effect model, error terms of each observations accommodates the difference between intercepts. Random Effect model has the advantage of being used to eliminate the problem of heteroscedasticity. In essence, random effect model is said to be different from the fixed effect model and common effect model.

Residuals are likely to be interconnected between time and individual observation or cross sections. Hence, random effects model assumes that there is a difference of intercept for each individual and the intercept is a random variable. Therefore, there are two residual components of random effect model. The first residual is an individual residual which is a random characteristic of the i-th unit observation and remains at all times.

The second residual as a whole where the residual is a combination of cross section and time series.

 

                                                Random Effect Model

for i = 1,2, …., n and t = 1,2, ….,

Where:

n = number of individuals or cross section

= the number of time periods.

it = is the residual as a whole where the residual is a combination of time series and cross section.

i = is the individual residual which is the random characteristic of unit observation the i-t hand remains at all times.

3.3.2        Best Model Selection Methods

3.3.2.1       Poolability F-Test

One of the main rationales behind pooling a time series of cross-sections is to widen the database in order to get robust and more reliable estimates of the parameters of the model. The simplest poolability test has its null hypothesis in the OLS model as:  and as its alternative hypothesis in the Fixed Effect model as:

In Stata, if xtreg is commanded with the Fixed Effect option, it will be obtained at the bottom of the output of the F-test that all α=0. Rejection of the null hypothesis means that the estimates suffer from an omission variables problem which makes it biased and inconsistent.

3.3.2.2       Breusch-Pagan Lagrange Multiplier (BP-LM) Test

Breusch-Pagan Lagrange Multiplier Test is an analysis performed in order to determine the best method to be adopted in panel data regression and to know whether common effect or random effect model is to be used. The main function of the Lagrange Multiplier test performs a function of determining the best estimate and whether to use a random effect or not.

When performing panel data regression, Lagrange Multiplier Test should be conducted if we are in two conditions, they are:

(a). if the Chow Test reveals that Common Effect of the fixed effect is the best method. So the

next step is to determine whether the Common Effect is better than the Random Effect, then the Lagrange Multiplier Test is required.

(b). if the Hausman Test revealed that the Random effect of the Fixed is the best method. So, the next step is to determine whether the Random Effect is better than the Common Effect, then the Lagrange Multiplier Test is required.

Lagrange multiplier (LM) test conducted to determine whether Random Effect model is better than Common Effect.

If Result:

H₀: Select Common Effect (p> 0.05)

H₁: Select Random Effect (p <0.05)

3.3.2.3       Hausman Test

To establish whether random effects model or fixed effects model is to be used for analysis, a formal test known as the Hausman test developed by Hausman in 1978 can be utilized. The null hypothesis in the Hausman test is that there is no significant difference between fixed effects model and the random effects model estimators. The test statistic in the Hausman test has an asymptotic χ2 distribution. If the null hypothesis is rejected, the conclusion is that random effects model is inappropriate and fixed effects model is a better alternative, while if the null hypothesis is not rejected, the random effects model is appropriate (Gujarati, 2004).

Hausman test is a statistical test that is used to determine the most appropriate Fixed Effect or

Random Effect model is used.

If Result:

H0: Select Random Effect (p> 0.05)

H1: Select Fixed Effect (p <0.05)

3.3.3        Diagnostic Checking

3.3.3.1       Heteroscedasticity

The reasons for the presence of heteroscedasticity are: poor data collecting techniques, presence of outliers, incorrect specification of the model to be tested, skewness in the distribution of one or more regressors included in the model, incorrect data transformations and through incorrect functional forms. The problem of heteroscedasticity is more likely to be present in case of cross-sectional data than time series data (Gujarati, 2004). There are a number of tests available for identifying the problem of heteroscedasticity, namely, Glejser’s (1969) Test, The Goldfeld and Quandt (1965) Test Breusch and Pagan (1979) Test, White’s (1980) Test, etc. (Baltagi, 2008).

3.3.3.2       Autocorrelation

Autocorrelation is said to be present in the data when there exists correlation between the members of series of observations ordered in time (as in time series data) or in space (as in cross-sectional data). When running regression analysis, it is assumed that autocorrelation is not present in the error terms. In other words, the model assumes that the error term related to any observation is not correlated to the error term of any other observation. The reasons for the presence of autocorrelation are: inertia, excluded variable specification bias, Cobweb phenomenon, lags, manipulation of data, data transformation, non-stationarity (Gujarati, 2004).

The Durbin-Watson Statistic is calculated to check if the problem of autocorrelation is present in the data. The value of Durbin-Watson ranges from 0-4. If the statistic is close to 4 the data has negative correlation, if it is close to 0 the data has positive correlation, while a score near 2 reveals no auto correlation. For objectives 1 and 2, The Durbin-Watson statistic is around 2 pertaining to the data corresponding to various objectives, hence, no problem of autocorrelation seems to be affecting the data corresponding to the present study. Whereas for objective 3 the score is not clearly close to 2, it raises some doubt regarding the presence of auto-correlation. The Wooldridge test of auto-correlation (Drukker, 2003) is used to address our doubt. The null hypothesis “no first order autocorrelation” was accepted, thus alleviating the doubt.

3.4        Conclusion

To conclude with, this chapter has comprehensively discussed on the methodology that will be  used to  test  for the  result in order to achieve the objectives of this study.  This chapter sets out the data, its nature and sources and the empirical model used in adopted for this study. All of the data were sourced from Chinn-Ito index and World Bank. Besides, the variables specification of measurements and the proxy of each variable were described clearly. Of all sophisticated analytical software that can be used for studies like this, STATA software is considered to be one of the most suitable for this research. All the result obtained from this software shall be analysed and interpreted in the following chapter.

 

 

 

 

 

 

4         CHAPTER 4: DATA ANALYSIS AND FINDINGS

4.1        Introduction

This chapter presents the panel data of sixty countries of three income groups ranging from top income to bottom income and gross national income of five years. The descriptive statistics was conducted and drawing conclusion from correlation results of the specified models. It follows by presenting the results of pool regression, fixed effect, and random effect. The main model for this study is fixed effect model while other three methods were included as the robustness check of results. the fixed effect model was more preferred as chosen by hausman test.

4.2         Descriptive Statistics

The descriptive statistic of three groups of income countries showing the strength of relationship the dependent and independent variables are presented in table 4.1

Table 4.1 Correlation coefficients

VARIABLES                                      T20                                B20                            GNI                      

LFDI                                               -.0.0574                            –0.0754                        0.1955  

UNEMPOLYMENT                          0.1329                                -0.0526                        -0.3685 

FO                                                   -0.4786                             0.0819                         0.1498  

TO                                                  -0.1523                               0.0191                          0.2476  

GT                                                   0.1869                                0.0447                   -0.0584 

Source : Authors computation from STATA 14.1

The simple correlation coefficients of the various income group distribution with other variables of interest are presented in table 4.1. from the results, it can be deduced that Top 20% (T20) is highly and negatively correlated with financial openness. The correlation between Top 20% (T20) and other variables were weak, while unemployment and government tax show positive correlation with the growth of income in this group. Also, foreign investment and trade openness have a negative correlation coefficient with income level.

The second column from the table reveals the correlation coefficients between Bottom 20% (B20) with other variables for the analysis. The results indicated that all the series are weakly correlated with dependent variables showing positive relationship with trade openness, financial openness and government tax. It is also negatively correlated with foreign investment and unemployment; this indicates that as the level of unemployment increases, the income level of the Bottom 20% (B20) keeps reducing which is economically plausible. As regards the foreign investment showing negative relationship, it indicates that as the foreign direct investment keep flowing into the countries under investigation, it will likely crowd out fund from those countries because of overdependence on foreign inputs which in turn will affect the income of the bottom earners.

From the third column on the table is the result of third sub-group of income which is proxied by gross national income. The result obtained indicated positive correlation between income and foreign investment, trade openness and financial openness and negatively correlated with government tax and unemployment. The result of correlation from this group followed economic expectation that foreign investment increase knowledge and technological inflow that can increase the overall income. Unemployment of resources will shift the production possibilities curve inward and thus reduces income level as a result of ineffectiveness.  The pool regression from the three income groups are presented in table 4.2.

 

4.3        Present of Results, Tables and Figures

4.3.1        Pool Regression Results

Table 4.2 Pool Regression Coefficients

	POOL REGRESSION
VARIABLES	T20	B20	GNI
LFDI	0.257 (0.198)	-1.3898 (0.431)	2.0452 (0.767)
UNEMPLOYMENT	0.1938 (0.009)***	-0.1189 (0.852)	-8.643 (0.001)***
FO	–8.3441 (0.000)***	-4.0141 (0.778)	100.267 (0.076)*
TO	-0.2507 (0.650)	2.354 (0.630)	10.533 (0.583)
LGT	0.1818 (0.466)	0.893 (0.672)	-6.9798 (0.400)
Constant	35.694 (0.000)***	60.225 (0.370)	267.507 (0.310)
R-squared	0.3247	0.0168	0.225
F-stat.(5, 77)	7.40 (0.0000)***	0.26 (0.9319)	4.48 (0.0012)***

Notes *, ** and *** are 10% , 5% and 1% level of significance respectively

Source: Authors computation from STATA 14.

 

4.3.2        Fixed Effect Model

To control for each country’s heterogeneity while obtaining unbiased estimates of coefficient, a fixed effect approach was adopted. In determining the best method to choose, a Hauman test was conducted and the test points the fixed effects estimators as the preferred technique of analysis

Table 4.3 Fixed Effect Regression Coefficients

	Robust Cluster FIXED REGRESSION
VARIABLES	T20	B20	GNI
LFDI	-0.0796 (0.241)	0.543 (0.620)	-4.225 (0.303)
UNEMPLOYMENT	0.154   (0.020)**	2.390 (0.393)	-9.385 (0.323)
FO	-0.408 (0.929)	-69.014 (0.627)	405.897 (0.239)
TO	1.0471 (0.422)	7.824 (0.720)	-15.2411 (0.858)
LGT	2.253 (0.315)	16.786 (0.571)	-135.783 (0.429)
Year 2014	-0.685 (0.061)	0.159 (0.946)	-3.215 (0.675)
Year 2015	-0.529 (0.239)	0.6048 (0.767	10.6025 (0.305)
Year 2016	-0.386 (0.486)	-1.113 (0.721)	8.740 (0.0253)**
2017	-0.561 (0.404)	-2.157 (0.531)	21.619 (0.232)
2018	-0.691 (0.743)	-48.272 (0.000)***	10.930) (0.377)
Constant	35.982 (0.000)***	-358.298 (0.645)	3522.64 (0.426)
Sigma_u	4.896	35.6454	282.9055
Sigma_e	0.636	11.899	38.906
Rho	0.983	0.899
Number of groups	25	25	25
F-stat.(10, 24)	2.76 (0.0201)***	10.90 (0.000)***	1.11 (0.396)
Hausman : ch2(5)	0.80 (0.938		7.72 (0.1723)
Heteroscedasticity: chi2(25)	2.031 (0.000)	325.53 (0.000)	45371.06 (0.000)
Serial correlation F(1, 11)	11.959 (0.0054)	0.073 (0.7915)	2.168 (0.1690)

Notes *, ** and *** are 10% , 5% and 1% level of significance respectively

Source: Authors computation from STATA 14.1

 

4.4        Interpretation on Major Findings

From the pool ordinary least square results presented in table 4.2 it can be inferred that unemployment has positive and significant impact on the Top 20% (T20) at 1%. This implies that unemployment rate is one of the key determinants of income differentials with coefficients of 0.1938 indicating that a unit increase in unemployment will widen the gap between the Top 20% (T20) level by 0.1938%. This finding is economically plausible as the top levels are not part of those that can be affected by unemployment. What this findings implies is that a unit increase in unemployment will make more people to fall and remain in the bottom level. This findings however is contrary the finding of (Anyanwu, 2014) who found that a reduction in graduate unemployment will induces the income level by increasing the productivity. This finding may be due to low rate of unemployment among high income countries which has reached full employment level thus it neither has neutral effect or positive on income across all levels. Financial openness has negative and significant impact on Top 20% (T20). This supports the findings from previous studies. Its implication is that a percent increase in financial openness will induce income level by 8.34% because the spillover effect will be felt on the Bottom 20% (B20) which will lift them from the bottom to the top thereby shrinking the gap between the Top 20% (T20) and Bottom 20% (B20).

With regard to general income group, financial openness has positive and significant effect on income level and statistically significant at one 1%.

With regards to gross national income and Bottom 20% (B20), the finding reveals that foreign investment influenced income level negatively, although insignificant. In relation with Top 20% (T20), the effect on income are opposite with negative multiplier effect on Bottom 20% (B20) while positive effect with top income although at both level the effect is insignificant.

The result presented in the table 4.3 is the finding from the fixed effect model. Having established the correlation between series for the model and there is evidence of multicollinearity and hendogeneity, we proceeded to examine the pool regression. The study further examines the fixed effect and random effect to determine if individual’s countries factors is necessary in the model. The hausman test was conducted to determine the appropriate model and conclusion was derived from the three models that fixed effect model were more preferable as the probability value from the different income group were greater than the 5% chosen level of significance, thus the null hypotheses that random effect is preferable was rejected.

4.4.1        Foreign investment

The coefficient of foreign investment from Top 20% (T20) is negative and statistically insignificant at 10%. However, the economic impact and prediction on Top 20% (T20) is quite infinitesimal. Holding all other variables fixed, a unit increase in foreign direct investment will dampen income by 0.796 (0.0796*10), ceteris paribus.

With regard to Bottom 20% (B20), the influence of foreign investment was strong and positive although insignificant. It implies that holding all other variables constant, a unit increase in foreign investment will induce income of the bottom earners by 0.543, ceteris paribus.  It effects on gross national income was negative and also insignificant which might be due to economic instability in some countries.

In conclusion, the alternative hypothesis of significant relationship between foreign investment on Top 20% (T20) and Bottom 20% (B20) was rejected and thus concluded that although foreign direct investment has positive relationship with Bottom 20% (B20) and GNI but negatively related with Top 20% (T20).

4.4.2        Unemployment

Unemployment has positive coefficient indicating that among Top 20% (T20), the unemployment has positive and significant impact on income level. This is contrary to theoretical expectation but can be related to the greediness of the policy makers among the Top 20% (T20) which may be directed at ensuring that there is unemployment in the country so that there will be surplus labour. The finding implies that a one percent increase in unemployment rate will induce income by 0.154 and is statistically significant at 5%.

As regard the Bottom 20% (B20), unemployment has positive coefficient which indicated that it has positively affected the level of income, although insignificant. The relationship between unemployment and gross national income indicated negative sign, showing that a unit increase in unemployment rate will reduces income level by 9.385. This is in line with theoretical expectation from Okun’s law that increases in the level of unemployment will shift the countries production possibility curve inward.

4.4.3        Financial openness

The coefficient of the financial openness is negative and statistically insignificant among Top 20% (T20) and Bottom 20% (B20) income level. The null hypothesis that there is no relationship was not rejected. Thus, concluded that a one percent increase in financial openness will affect the income of the Top 20% (T20) by 0.408 and reduce income from Bottom 20% (B20) income by 69.014.

With regard to the third group, financial openness will induce income positively and is statistical insignificant.

4.4.4        Trade openness

The coefficient of trade openness in Top 20% (T20) level indicated negative impact but positive with Bottom 20% (B20) as well as gross national income. However, all were statistically insignificant. The finding indicated that holding all variables constant, a one percent increase in trade openness will reduce income level among Top 20% (T20) by 1.047 and increase income among Bottom 20% (B20) by 7.842 but it influence on the third group (GNI) negatively by 15.24.The finding concluded that  trade openness has positive but insignificant impact on income in among Top 20% (T20) and Bottom 20% (B20) but negatively and statistically insignificant on the third column income.

4.4.5        Government tax

Column 1 indicated the coefficients that measure the impact of government tax on Top 20% (T20) level of income. The coefficient of government tax is positive but statistically insignificant even at 10%. However, the coefficient indicated that a one percent increase in government tax would induce income by 2.253, holding all other variables fixed.  The alternative hypothesis of significant relationship between income and government tax was rejected in favour null hypothesis.

From column 2, government tax has positive and insignificant impact on income level. The coefficient that measure the multiplier effect indicated that a one percent increase government tax will cause multiplier effect by 16.785, although is statistically insignificant. In column 3, it also indicated insignificant and negative impact on income. This implies that as government tax increases, the income decreases because of reduction in consumption expenditure that can stimulate investment.

Table 4.4 Random Regression Coefficients

	Random REGRESSION
VARIABLES	T20	B20	GNI
LFDI	-0.0075 (0.937)	-1.421 (0.475)	-4.382 (0.413)
UNEMPLOYMENT	0.0975 (0.177)	0.307 (0.693)	-6.909 (0.034)
FO	-2.0862 (0.222)	-8.0664 (0.654)	177.755 (0.023)
TO	-0.0564 (0.947)	3.654 (0.549)	21.161 (0.495)
LGT	0.208 (0.584)	1.246 (0.626)	-7.135 (0.555)
Constant	35.982 (0.000)***	49.919 (0.528)	316.956 (0.372)
Sigma_u	3.044	9.473	91.601
Sigma_e	0.6464	19.184	37.37
rho	0.9569	0.196	0.857
Number of groups	25	25	25

Notes *, ** and *** are 10% , 5% and 1% level of significance respectively

Source: Authors computation from STATA 14.1

4.5        CONCLUSION

The result from the finding indicated that fixed effect is most preferred using Hausman test meaning individual country’s heterogeneous factors are important. The hypothesis that unemployment has no significant relationship with Top 20% (T20) income group was rejected in favor of alternative hypothesis. The study concluded that unemployment has positive and significant impact on various levels of income. This findings is recommended to be subjected to further findings as its implication does not support a priori expectation The result of fixed effect indicated that none of the variable has significant impact on income in 20 Bottom income level. This would be due to institutional framework problem among low income countries.

From the comparative static, it is crystal clear that fixed effects estimators behave very poorly when dealing with time-variant as most variables appear insignificant but has significant effect from generalized least square and pool regression.  (Adeleye, 2014)

 

 

5         CHAPTER 5: SUMMARY, CONCLUSION AND RECOMMENDATIONS

5.1        Introduction

This chapter presents the summary from findings, discussion and conclusion on findings. The last part of the section encompasses the limitations of the study as well as the recommendation for further study.

5.2        Summary

The study examines the effect foreign investment, financial openness and trade openness to various levels of income. In order to obtain unbiased estimates coefficients and avoid spurious results from misspecification and problem multi-collinearity we carry out correlation as preliminary test aimed at implementing all data processing techniques. The proceeded to estimate pool panel regression across different income levels. The findings from Top 20% (T20) income group indicated unemployment has positive and significant effect on the level of income. The reason for this would be because of unemployment benefits. Financial openness indicated negative and significant effect on income level while other variables remain insignificant.

From Bottom 20% (B20) income level, all the variables indicated insignificant relationship with the level of income.

The column 3 in table 4.3 in the last chapter indicated the relationship between independent variables and income level.  The findings infer that all the variables are statistically insignificant. only the financial openness indicated positive but insignificant effect on income level. All other variables ranging from foreign investment, unemployment, trade openness and government tax indicated negative effect on income level.

5.3        Discussion and conclusion

From the preceding chapter on results presentation and interpretation the study concludes as follows;

Firstly, the study found a negative and insignificant relationship between foreign direct investment and income level among Top 20% (T20) income group. It thus concluded that foreign investment is not one of the major factors that can influences the growth of income among high income level. Likewise, the same nature of relationship was found among with gross national income while the association between foreign investment and bottom income countries’ income level are positive, although insignificant at 10%.

In addition, the finding indicated positive and significant relationship between unemployment and Top 20% (T20) income. This indicated that unemployment is of the key determinants of the level of income among the top income level. The reason for the positive relationship might be because of unemployment benefits paid by developed countries that reduce crime and social unrest in the society, thus stimulate investment for income growth.

Furthermore, financial openness has negative and insignificant effect on income level in to income countries. Thus, the study concluded that financial openness is not one of the strong determinants of income growth in developed countries. The reason for insignificant coefficient might be due to the state of the economy. Financial openness in Bottom 20% (B20) income level also has negative and insignificant effect. However, the coefficient of financial openness has positive but still remain insignificant on income in the third group of the study which is GNI. The study finally concluded that although financial openness has positive effect of gross national income and negative effect on top income and bottom countries, the conclusion from all remain insignificant.

Trade openness has positive and insignificant effect on top income and bottom income but indicated negative effect on gross national income. The conclusion was drawn that trade openness will influence top income and bottom income countries’ income positively but negatively on gross national income. The conclusion also indicated that trade openness is not one of the keys determinants of income level among different income countries.

Lastly, revenue from tax indicated positive and insignificant effect on top income countries’ income.  It also indicates positive and insignificant relationship with income level in Bottom 20% (B20) income level.  The effect on gross national income indicated negative and insignificant, thus a conclusion was that tax revenue is not one of the key determinants of income level.

The result of the year cluster indicated 2016 to have significant and positive effect on gross national income while 2018 has positive and significant effect on Bottom 20% (B20) income level. Holding all the variables constant, the income level of among Top 20% (T20) income is positive and significant relationship.

 

5.4        Limitation and recommendation for future research

As it is the case with all studies, this study is associated with some limitations. The results of this analysis must therefore be taken into account with the following limitations

1. The study failed to use more variables that can capture the full effect on the income level among different income level countries. Some existing study that use interaction effect variable obtained more efficient and unbiased results.
2. The study considered only five variables for this study which indicated error of misspecification and misspecification in the model of the study.

• The methodology used for the study failed to give the true nature of relationship between independent variables and independent variables among income different countries because of the short scope used for the study.

The following acts as the limitations can be addressed as this single study cannot cover be covered by a single research. Therefore, the following areas are hereby are recommended;

1. The study used fixed effect model that covered individual’s country heterogeneity but failed to gives the full effects on different income levels. The scope of this study was short for the fixed effect model, the study thus recommended further studies to expand scope and adopt great method of moments (GMM) for unbiased and efficient estimate.
2. Researchers in the area can carry out similar studies by increasing the numbers of independence variables that can capture all the factors that can determine income inequality among different income countries.

 

Appendices

Results

REGRESSION RESULT FOR GNI AND FDI

SUMMARY STATISTICS

 

summarize gni1 lfdi unemployment fo to lgt

 

    Variable |        Obs        Mean    Std. Dev.       Min        Max

————-+———————————————————

        gni1 |        360    179.5028    104.0625          1        359

        lfdi |         83    22.75382    1.692116   18.17406   25.81447

unemployment |        360    8.827183    6.098528      .2065      35.15

          fo |        360    .7499561    .3181932          0          1

          to |        360    1.085053    .6226261   .2242296   4.091723

————-+———————————————————

         lgt |        360     26.1042    2.810644   21.43853   34.95619

 

. correlate  gni1 lfdi unemployment fo to lgt

(obs=83)

 

             |     gni1     lfdi unempl~t       fo       to      lgt

————-+——————————————————

        gni1 |   1.0000

        lfdi |   0.1955   1.0000

unemployment |  -0.3685  -0.2538   1.0000

          fo |   0.1498   0.1986   0.2217   1.0000

          to |   0.2476   0.2661  -0.2071   0.0116   1.0000

         lgt |  -0.0584   0.0841  -0.3805  -0.2859  -0.5105   1.0000

 

 

. regress gni1 lfdi unemployment fo to lgt

 

      Source |       SS           df       MS      Number of obs   =        83

————-+———————————-   F(5, 77)        =      4.48

       Model |  191228.679         5  38245.7359   Prob > F        =    0.0012

    Residual |  657368.863        77  8537.25796   R-squared       =    0.2253

————-+———————————-   Adj R-squared   =    0.1750

       Total |  848597.542        82  10348.7505   Root MSE        =    92.397

 

——————————————————————————

        gni1 |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]

————-+—————————————————————-

        lfdi |   2.045298   6.872649     0.30   0.767    -11.63989    15.73049

unemployment |  -8.642692     2.4972    -3.46   0.001    -13.61525   -3.670131

          fo |   100.2671   55.69308     1.80   0.076    -10.63195    211.1662

          to |   10.53321   19.08731     0.55   0.583    -27.47448     48.5409

         lgt |  -6.979817    8.24231    -0.85   0.400    -23.39235     9.43272

       _cons |   267.5072   261.7761     1.02   0.310    -253.7556    788.7701

——————————————————————————

 

.

.

. xtreg gni1 lfdi unemployment fo to lgt, fe

must specify panelvar; use xtset

r(459);

 

. xtreg gni1 lfdi unemployment fo to lgt, fe

must specify panelvar; use xtset

r(459);

 

. xtset c_id year

       panel variable:  c_id (strongly balanced)

        time variable:  year, 2013 to 2018

                delta:  1 unit

 

. xtreg gni1 lfdi unemployment fo to lgt, fe

 

Fixed-effects (within) regression               Number of obs     =         83

Group variable: c_id                            Number of groups  =         25

 

R-sq:                                           Obs per group:

     within  = 0.2266                                         min =          1

     between = 0.0429                                         avg =        3.3

     overall = 0.0384                                         max =          6

 

                                                F(5,53)           =       3.10

corr(u_i, Xb)  = -0.9190                        Prob > F          =     0.0157

 

——————————————————————————

        gni1 |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]

————-+—————————————————————-

        lfdi |  -2.853515   5.825409    -0.49   0.626    -14.53781    8.830776

unemployment |  -11.82819   5.803329    -2.04   0.047     -23.4682   -.1881894

          fo |   374.4815   121.3097     3.09   0.003     131.1651    617.7979

          to |   6.600946   93.06792     0.07   0.944    -180.0697    193.2716

         lgt |  -118.9194    79.2434    -1.50   0.139    -277.8616    40.02271

       _cons |   3083.045   2098.298     1.47   0.148    -1125.609    7291.699

————-+—————————————————————-

     sigma_u |  250.76025

     sigma_e |  37.379627

         rho |  .97826258   (fraction of variance due to u_i)

——————————————————————————

F test that all u_i=0: F(24, 53) = 17.39                     Prob > F = 0.0000

 

. estimate store fe

 

.  xtreg gni1 lfdi unemployment fo to lgt, re

 

Random-effects GLS regression                   Number of obs     =         83

Group variable: c_id                            Number of groups  =         25

 

R-sq:                                           Obs per group:

     within  = 0.1657                                         min =          1

     between = 0.1235                                         avg =        3.3

     overall = 0.1833                                         max =          6

 

                                                Wald chi2(5)      =      11.99

corr(u_i, X)   = 0 (assumed)                    Prob > chi2       =     0.0349

 

——————————————————————————

        gni1 |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]

————-+—————————————————————-

        lfdi |  -4.382089   5.347273    -0.82   0.413    -14.86255    6.098374

unemployment |  -6.909117   3.265319    -2.12   0.034    -13.30903   -.5092089

          fo |   177.7554   78.04205     2.28   0.023     24.79582     330.715

          to |   21.16128    30.9966     0.68   0.495    -39.59093     81.9135

         lgt |  -7.135922   12.10114    -0.59   0.555    -30.85372    16.58188

       _cons |   316.9569   355.2014     0.89   0.372     -379.225    1013.139

————-+—————————————————————-

     sigma_u |  91.601476

     sigma_e |  37.379627

         rho |  .85725103   (fraction of variance due to u_i)

——————————————————————————

 

HAUSMAN TEST

.

. estimate store re

 

. hausman fe re

 

                 —- Coefficients —-

             |      (b)          (B)            (b-B)     sqrt(diag(V_b-V_B))

             |       fe           re         Difference          S.E.

————-+—————————————————————-

        lfdi |   -2.853515    -4.382089        1.528574        2.311289

unemployment |   -11.82819    -6.909117       -4.919077        4.797533

          fo |    374.4815     177.7554         196.726        92.87345

          to |    6.600946     21.16128       -14.56034        87.75448

         lgt |   -118.9194    -7.135922       -111.7835        78.31397

——————————————————————————

                           b = consistent under Ho and Ha; obtained from xtreg

            B = inconsistent under Ha, efficient under Ho; obtained from xtreg

 

    Test:  Ho:  difference in coefficients not systematic

 

                  chi2(5) = (b-B)'[(V_b-V_B)^(-1)](b-B)

                          =        7.72

                Prob>chi2 =      0.1723

 

Fixed effect result

 

.  xtreg gni1 lfdi unemployment fo to lgt, fe

 

Fixed-effects (within) regression               Number of obs     =         83

Group variable: c_id                            Number of groups  =         25

 

R-sq:                                           Obs per group:

     within  = 0.2266                                         min =          1

     between = 0.0429                                         avg =        3.3

     overall = 0.0384                                         max =          6

 

                                                F(5,53)           =       3.10

corr(u_i, Xb)  = -0.9190                        Prob > F          =     0.0157

 

——————————————————————————

        gni1 |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]

————-+—————————————————————-

        lfdi |  -2.853515   5.825409    -0.49   0.626    -14.53781    8.830776

unemployment |  -11.82819   5.803329    -2.04   0.047     -23.4682   -.1881894

          fo |   374.4815   121.3097     3.09   0.003     131.1651    617.7979

          to |   6.600946   93.06792     0.07   0.944    -180.0697    193.2716

         lgt |  -118.9194    79.2434    -1.50   0.139    -277.8616    40.02271

       _cons |   3083.045   2098.298     1.47   0.148    -1125.609    7291.699

————-+—————————————————————-

     sigma_u |  250.76025

     sigma_e |  37.379627

         rho |  .97826258   (fraction of variance due to u_i)

——————————————————————————

F test that all u_i=0: F(24, 53) = 17.39                     Prob > F = 0.0000

 

Diagnostic test on fixed effect model

 

Heteroscedasticity test

 

.

. xttest3

 

Modified Wald test for groupwise heteroskedasticity

in fixed effect regression model

 

H0: sigma(i)^2 = sigma^2 for all i

 

chi2 (25)  =   45371.06

Prob>chi2 =      0.0000

 

Autocorrelation In Panel Data Test

 

 

. xtserial gni1 lfdi unemployment fo to lgt

 

Wooldridge test for autocorrelation in panel data

H0: no first-order autocorrelation

    F(  1,      11) =      2.168

           Prob > F =      0.1690

 

Cluster Robust Fixed Effect

 

. xtreg gni1 lfdi unemployment fo to lgt i.year, fe robust cluster(c_id)

 

Fixed-effects (within) regression               Number of obs     =         83

Group variable: c_id                            Number of groups  =         25

 

R-sq:                                           Obs per group:

     within  = 0.2569                                         min =          1

     between = 0.0277                                         avg =        3.3

     overall = 0.0230                                         max =          6

 

                                                F(10,24)          =       1.11

corr(u_i, Xb)  = -0.9348                        Prob > F          =     0.3959

 

                                  (Std. Err. adjusted for 25 clusters in c_id)

——————————————————————————

             |               Robust

        gni1 |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]

————-+—————————————————————-

        lfdi |  -4.225835    4.01622    -1.05   0.303    -12.51491    4.063236

unemployment |  -9.385047    9.30634    -1.01   0.323    -28.59239    9.822295

          fo |   405.8975   336.0427     1.21   0.239    -287.6606    1099.456

          to |  -15.24111   84.25691    -0.18   0.858    -189.1388    158.6566

         lgt |  -135.7383   168.5605    -0.81   0.429    -483.6301    212.1535

             |

        year |

       2014  |  -3.215509     7.5668    -0.42   0.675    -18.83262     12.4016

       2015  |   10.60251   10.10869     1.05   0.305    -10.26081    31.46582

       2016  |   8.739946   7.469357     1.17   0.253    -6.676048    24.15594

       2017  |   21.61888   17.62983     1.23   0.232    -14.76731    58.00506

       2018  |   10.93005   12.15553     0.90   0.377    -14.15772    36.01782

             |

       _cons |    3522.64   4349.187     0.81   0.426    -5453.641    12498.92

————-+—————————————————————-

     sigma_u |  282.90554

     sigma_e |  38.500772

         rho |  .98181614   (fraction of variance due to u_i)

——————————————————————————

 

 

 

GENALIZED LEST SQUARE RESULT (GLS )

 

. xtregar gni1 lfdi unemployment fo to lgt, fe rhotype(dw)

 

FE (within) regression with AR(1) disturbances  Number of obs     =         58

Group variable: c_id                            Number of groups  =         20

 

R-sq:                                           Obs per group:

     within  = 0.3297                                         min =          1

     between = 0.0094                                         avg =        2.9

     overall = 0.0134                                         max =          5

 

                                                F(5,33)           =       3.25

corr(u_i, Xb)  = -0.8133                        Prob > F          =     0.0171

 

——————————————————————————

        gni1 |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]

————-+—————————————————————-

        lfdi |  -3.553112    7.54863    -0.47   0.641    -18.91092    11.80469

unemployment |  -3.564646   10.72356    -0.33   0.742    -25.38189    18.25259

          fo |   658.2902   237.4122     2.77   0.009     175.2714    1141.309

          to |  -90.93146   96.37584    -0.94   0.352    -287.0096    105.1467

         lgt |   -1.33757    9.66071    -0.14   0.891    -20.99243    18.31729

       _cons |  -194.0906   71.24349    -2.72   0.010    -339.0366   -49.14468

————-+—————————————————————-

      rho_ar |  .40326292

     sigma_u |  174.58724

     sigma_e |  46.918581

     rho_fov |  .93264337   (fraction of variance because of u_i)

——————————————————————————

F test that all u_i=0: F(19,33) = 3.57                       Prob > F = 0.0007

 

.

.

.

 

 

a   REGRESSION RESULT FOR INCOME TOP

POOL REGRESSION RESULTS

reg T20 lfdi unemployment fo to gt

 

      Source |       SS           df       MS      Number of obs   =        83

————-+———————————-   F(5, 77)        =     11.46

       Model |  345.125021         5  69.0250042   Prob > F        =    0.0000

    Residual |  463.763777        77   6.0229062   R-squared       =    0.4267

————-+———————————-   Adj R-squared   =    0.3894

       Total |  808.888798        82  9.86449754   Root MSE        =    2.4542

 

——————————————————————————

   T20  |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]

————-+—————————————————————-

        lfdi |   .0931245   .1842182     0.51   0.615    -.2737007    .4599497

unemployment |   .2023785    .058196     3.48   0.001     .0864955    .3182614

          fo |  -5.785659   1.606693    -3.60   0.001    -8.984994   -2.586325

          to |   .0554338   .4017817     0.14   0.891    -.7446157    .8554834

          gt |   5.43e-13   1.43e-13     3.79   0.000     2.58e-13    8.28e-13

       _cons |   40.84718   3.935171    10.38   0.000     33.01126    48.68311

 

FIXED EFFECT RESULTS

 

. xtreg T20 lfdi unemployment fo to gt, fe

 

Fixed-effects (within) regression               Number of obs     =         83

Group variable: c_id                            Number of groups  =         25

 

R-sq:                                           Obs per group:

     within  = 0.0332                                         min =          1

     between = 0.0054                                         avg =        3.3

     overall = 0.0174                                         max =          6

 

                                                F(5,53)           =       0.36

corr(u_i, Xb)  = -0.3484                        Prob > F          =     0.8710

 

——————————————————————————

   T20 |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]

————-+—————————————————————-

        lfdi |  -.0122108    .101097    -0.12   0.904    -.2149857    .1905641

unemployment |   .0850668    .083579     1.02   0.313    -.0825715    .2527052

          fo |  -1.258328   2.239599    -0.56   0.577    -5.750394    3.233739

          to |   .3985519   1.612363     0.25   0.806    -2.835439    3.632543

          gt |  -1.94e-13   2.86e-13    -0.68   0.501    -7.68e-13    3.80e-13

       _cons |    40.4432   3.385292    11.95   0.000     33.65317    47.23324

————-+—————————————————————-

     sigma_u |  3.2241879

     sigma_e |  .64469987

         rho |  .96155426   (fraction of variance due to u_i)

——————————————————————————

F test that all u_i=0: F(24, 53) = 44.28                     Prob > F = 0.0000

 

 

RANDOM EFFECT RESULT

 

xtreg T20 lfdi unemployment fo to gt, re

 

Random-effects GLS regression                   Number of obs     =         83

Group variable: c_id                            Number of groups  =         25

 

R-sq:                                           Obs per group:

     within  = 0.0095                                         min =          1

     between = 0.2438                                         avg =        3.3

     overall = 0.3720                                         max =          6

 

                                                Wald chi2(4)      =          .

corr(u_i, X)   = 0 (assumed)                    Prob > chi2       =          .

 

——————————————————————————

   T20 |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]

————-+—————————————————————-

        lfdi |  -.0099792   .0976824    -0.10   0.919    -.2014332    .1814748

unemployment |    .084827    .064656     1.31   0.190    -.0418963    .2115504

          fo |   -1.80252    1.82412    -0.99   0.323    -5.377729    1.772688

          to |  -.2260928   .7479795    -0.30   0.762    -1.692106     1.23992

          gt |   1.55e-13   2.16e-13     0.72   0.474    -2.69e-13    5.79e-13

       _cons |   41.33549   2.746599    15.05   0.000     35.95225    46.71872

————-+—————————————————————-

     sigma_u |  2.8850712

     sigma_e |  .64469987

         rho |  .95244022   (fraction of variance due to u_i)

——————————————————————————

 

HAUMAN TEST

 

. hausman fe re

 

Note: the rank of the differenced variance matrix (4) does not equal the number of coefficients being tested (5);

        be sure this is what you expect, or there may be problems computing the test.  Examine the output of your

        estimators for anything unexpected and possibly consider scaling your variables so that the coefficients

        are on a similar scale.

 

                 —- Coefficients —-

             |      (b)          (B)            (b-B)     sqrt(diag(V_b-V_B))

             |       fe           re         Difference          S.E.

————-+—————————————————————-

        lfdi |   -.0122108    -.0099792       -.0022316        .0260528

unemployment |    .0850668      .084827        .0002398        .0529628

          fo |   -1.258328     -1.80252        .5441926        1.299381

          to |    .3985519    -.2260928        .6246447         1.42837

          gt |   -1.94e-13     1.55e-13       -3.49e-13        1.87e-13

——————————————————————————

                           b = consistent under Ho and Ha; obtained from xtreg

            B = inconsistent under Ha, efficient under Ho; obtained from xtreg

 

    Test:  Ho:  difference in coefficients not systematic

 

                  chi2(4) = (b-B)'[(V_b-V_B)^(-1)](b-B)

                          =        0.80

                Prob>chi2 =      0.9382

 

 DIGNOSTIC TEST FOR RANDOM EFFECT

ttest0

 

Breusch and Pagan Lagrangian multiplier test for random effects

 

        720[c_id,t] = Xb + u[c_id] + e[c_id,t]

 

        Estimated results:

                         |       Var     sd = sqrt(Var)

                ———+—————————–

               T20 |   9.864498        3.14078

                       e |   .4156379       .6446999

                       u |   8.323636       2.885071

 

        Test:   Var(u) = 0

                             chibar2(01) =    55.18

                          Prob > chibar2 =   0.0000

 

 

Cluster random regression

andom-effects GLS regression                   Number of obs     =         83

Group variable: c_id                            Number of groups  =         25

 

R-sq:                                           Obs per group:

     within  = 0.0821                                         min =          1

     between = 0.2137                                         avg =        3.3

     overall = 0.3519                                         max =          6

 

                                                Wald chi2(9)      =          .

corr(u_i, X)   = 0 (assumed)                    Prob > chi2       =          .

 

                                  (Std. Err. adjusted for 25 clusters in c_id)

——————————————————————————

             |               Robust

   T20 |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]

————-+—————————————————————-

        lfdi |  -.0453236   .0652198    -0.69   0.487    -.1731521    .0825049

unemployment |   .0899402   .0933218     0.96   0.335    -.0929671    .2728474

          fo |  -1.005996   3.989142    -0.25   0.801    -8.824571     6.81258

          to |  -.0477071    .495037    -0.10   0.923    -1.017962    .9225475

          gt |   2.35e-13   1.86e-13     1.26   0.207    -1.30e-13    6.01e-13

             |

        year |

       2014  |  -.5063287   .3000678    -1.69   0.092    -1.094451    .0817934

       2015  |  -.3736349   .3600211    -1.04   0.299    -1.079263    .3319935

       2016  |  -.1733661   .3970504    -0.44   0.662    -.9515705    .6048384

       2017  |  -.3262173   .5279437    -0.62   0.537    -1.360968    .7085334

       2018  |   -.366906   .5122713    -0.72   0.474    -1.370939    .6371273

             |

       _cons |   41.39282    3.78931    10.92   0.000     33.96591    48.81973

————-+—————————————————————-

     sigma_u |   2.873273

     sigma_e |  .64498614

         rho |  .95202707   (fraction of variance due to u_i)

——————————————————————————

 

 

 

GENALISED LEAST SQUARE

xtregar T20 lfdi unemployment fo to gt, re rhotype(dw)

 

RE GLS regression with AR(1) disturbances       Number of obs     =         83

Group variable: c_id                            Number of groups  =         25

 

R-sq:                                           Obs per group:

     within  = 0.0063                                         min =          1

     between = 0.2701                                         avg =        3.3

     overall = 0.4159                                         max =          6

 

                                                Wald chi2(6)      =       7.39

corr(u_i, Xb)      = 0 (assumed)                Prob > chi2       =     0.2859

 

——————- theta ——————–

  min      5%       median        95%      max

0.6702   0.6702     0.7612     0.7743   0.7743

 

——————————————————————————

   T20 |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]

————-+—————————————————————-

        lfdi |  -.0239847   .0821006    -0.29   0.770    -.1848989    .1369295

unemployment |   .1268808   .0672783     1.89   0.059    -.0049823    .2587439

          fo |  -3.054624   1.624262    -1.88   0.060    -6.238118     .128871

          to |  -.0634719   .6659391    -0.10   0.924    -1.368689    1.241745

          gt |   3.17e-13   1.83e-13     1.74   0.082    -4.07e-14    6.75e-13

       _cons |   42.05446   2.378551    17.68   0.000     37.39259    46.71634

————-+—————————————————————-

      rho_ar |  .59412613   (estimated autocorrelation coefficient)

     sigma_u |  2.3008222

     sigma_e |  .64644957

     rho_fov |  .92683465   (fraction of variance due to u_i)

——————————————————————————

 

REGRESSION RESULTS

 

 

SUMMARY STATISTIC

 

. summarize B20 lfdi unemployment fo to lgt

 

    Variable |        Obs        Mean    Std. Dev.       Min        Max

————-+———————————————————

incomebott~1 |        360       39.65    25.11178          1         81

        lfdi |         83    22.75382    1.692116   18.17406   25.81447

unemployment |        360    8.827183    6.098528      .2065      35.15

          fo |        360    .7499561    .3181932          0          1

          to |        360    1.085053    .6226261   .2242296   4.091723

————-+———————————————————

         lgt |        360     26.1042    2.810644   21.43853   34.95619

 

 

correlation result

correlate  B20 lfdi unemployment fo to lgt

(obs=83)

 

             | income~1     lfdi unempl~t       fo       to      lgt

————-+——————————————————

incomebott~1 |   1.0000

        lfdi |  -0.0754   1.0000

unemployment |  -0.0526  -0.2538   1.0000

          fo |  -0.0819   0.1986   0.2217   1.0000

          to |   0.0191   0.2661  -0.2071   0.0116   1.0000

         lgt |   0.0447   0.0841  -0.3805  -0.2859  -0.5105   1.0000

 

 

POOL REGRESSION RESULTS

 

 

. regress B20  lfdi unemployment fo to lgt

 

      Source |       SS           df       MS      Number of obs   =        83

————-+———————————-   F(5, 77)        =      0.26

       Model |   730.85467         5  146.170934   Prob > F        =    0.9319

    Residual |  42806.3742        77  555.926938   R-squared       =    0.0168

————-+———————————-   Adj R-squared   =   -0.0471

       Total |  43537.2289        82  530.941816   Root MSE        =    23.578

 

——————————————————————————

incomebott~1 |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]

————-+—————————————————————-

        lfdi |  -1.389809   1.753775    -0.79   0.431    -4.882021    2.102403

unemployment |  -.1189376     .63724    -0.19   0.852    -1.387845    1.149969

          fo |  -4.014121   14.21186    -0.28   0.778    -32.31355    24.28531

          to |   2.354343   4.870734     0.48   0.630    -7.344526    12.05321

         lgt |   .8937551   2.103287     0.42   0.672    -3.294425    5.081935

       _cons |   60.22519   66.80049     0.90   0.370    -72.79159     193.242

 

 

Fixed Effect Model

xtreg B20  lfdi unemployment fo to lgt, fe

 

Fixed-effects (within) regression               Number of obs     =         83

Group variable: c_id                            Number of groups  =         25

 

R-sq:                                           Obs per group:

     within  = 0.3724                                         min =          1

     between = 0.0006                                         avg =        3.3

     overall = 0.0017                                         max =          6

 

                                                F(5,53)           =       6.29

corr(u_i, Xb)  = -0.9985                        Prob > F          =     0.0001

 

——————————————————————————

incomebott~1 |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]

————-+—————————————————————-

        lfdi |   2.103707   2.989738     0.70   0.485    -3.892949    8.100363

unemployment |   3.172536   2.978407     1.07   0.292    -2.801391    9.146464

          fo |  -60.93003   62.25901    -0.98   0.332    -185.8058    63.94574

          to |   35.31512   47.76467     0.74   0.463    -60.48868    131.1189

         lgt |  -156.6213    40.6696    -3.85   0.000    -238.1942    -75.0484

       _cons |   4079.127   1076.897     3.79   0.000     1919.146    6239.109

————-+—————————————————————-

     sigma_u |  339.50761

     sigma_e |  19.184114

         rho |  .99681727   (fraction of variance due to u_i)

——————————————————————————

F test that all u_i=0: F(24, 53) = 2.64                      Prob > F = 0.0017

 

 

RANDOM EFFECT

 

xtreg B20  lfdi unemployment fo to lgt, re

 

Random-effects GLS regression                   Number of obs     =         83

Group variable: c_id                            Number of groups  =         25

 

R-sq:                                           Obs per group:

     within  = 0.0031                                         min =          1

     between = 0.1737                                         avg =        3.3

     overall = 0.0119                                         max =          6

 

                                                Wald chi2(5)      =       1.31

corr(u_i, X)   = 0 (assumed)                    Prob > chi2       =     0.9337

 

——————————————————————————

incomebott~1 |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]

————-+—————————————————————-

        lfdi |  -1.421044   1.987105    -0.72   0.475    -5.315698     2.47361

unemployment |   .3072111   .7788509     0.39   0.693    -1.219309    1.833731

          fo |  -8.066428   17.97587    -0.45   0.654    -43.29849    27.16564

          to |    3.65447   6.091904     0.60   0.549    -8.285442    15.59438

         lgt |   1.246133   2.557867     0.49   0.626    -3.767195    6.259461

       _cons |   49.91889   79.01846     0.63   0.528    -104.9544    204.7922

————-+—————————————————————-

     sigma_u |  9.4735912

     sigma_e |  19.184114

         rho |   .1960529   (fraction of variance due to u_i)

——————————————————————————

 

 HAUSMAN TEST

hausman fe re

 

                 —- Coefficients —-

             |      (b)          (B)            (b-B)     sqrt(diag(V_b-V_B))

             |       fe           re         Difference          S.E.

————-+—————————————————————-

        lfdi |    2.103707    -1.421044        3.524751         2.23382

unemployment |    3.172536     .3072111        2.865325        2.874769

          fo |   -60.93003    -8.066428        -52.8636        59.60749

          to |    35.31512      3.65447        31.66065         47.3746

         lgt |   -156.6213     1.246133       -157.8674        40.58908

——————————————————————————

                           b = consistent under Ho and Ha; obtained from xtreg

            B = inconsistent under Ha, efficient under Ho; obtained from xtreg

 

    Test:  Ho:  difference in coefficients not systematic

 

                  chi2(5) = (b-B)'[(V_b-V_B)^(-1)](b-B)

                          =       33.41

                Prob>chi2 =      0.0000

 Diagnostic test on fixed effect model

 

Heteroscedasticity test

. xttest3

 

Modified Wald test for groupwise heteroskedasticity

in fixed effect regression model

 

H0: sigma(i)^2 = sigma^2 for all i

 

chi2 (25)  =     325.53

Prob>chi2 =      0.0000

 

Autocorrelation In Panel Data Test

 

xtserial B20  lfdi unemployment fo to lgt

 

Wooldridge test for autocorrelation in panel data

H0: no first-order autocorrelation

    F(  1,      11) =      0.073

           Prob > F =      0.7915

 

 

Cluster Robust Fixed Effect

xtreg B20 lfdi unemployment fo to lgt i.year, fe robust cluster(c_id)

 

Fixed-effects (within) regression               Number of obs     =         83

Group variable: c_id                            Number of groups  =         25

 

R-sq:                                           Obs per group:

     within  = 0.7813                                         min =          1

     between = 0.0310                                         avg =        3.3

     overall = 0.1831                                         max =          6

 

                                                F(10,24)          =      10.90

corr(u_i, Xb)  = -0.8233                        Prob > F          =     0.0000

 

                                  (Std. Err. adjusted for 25 clusters in c_id)

——————————————————————————

             |               Robust

incomebott~1 |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]

————-+—————————————————————-

        lfdi |   .5433736   1.082974     0.50   0.620    -1.691775    2.778522

unemployment |   2.390725   2.747424     0.87   0.393    -3.279679    8.061129

          fo |  -69.01409   60.78291    -1.14   0.267    -194.4638    56.43566

          to |   7.824321   21.61074     0.36   0.720    -36.77806     52.4267

         lgt |   16.78629   29.22729     0.57   0.571    -43.53588    77.10846

             |

        year |

       2014  |   .1598003   2.336915     0.07   0.946    -4.663355    4.982955

       2015  |   .6048507      2.014     0.30   0.767    -3.551841    4.761543

       2016  |  -1.113744   3.078369    -0.36   0.721    -7.467185    5.239697

       2017  |  -2.157194   3.392536    -0.64   0.531    -9.159044    4.844657

       2018  |  -48.27197    8.32177    -5.80   0.000    -65.44726   -31.09668

             |

       _cons |  -358.2981   767.1022    -0.47   0.645    -1941.519    1224.923

————-+—————————————————————-

     sigma_u |  35.644541

     sigma_e |  11.899989

         rho |  .89971997   (fraction of variance due to u_i)

——————————————————————————