FOREIGN TRADE UNIVERSITY FACULTY OF INTERNATIONAL ECONOMICS *** ECONOMETRICS REPORT THE IMPACT OF GDP PER CAPITA AND OTHER FACTORS ON LIFE EXPECTANCY IN SOME COUNTRIES IN 2015 Instructor PhD Đinh Thị[.]
FOREIGN TRADE UNIVERSITY FACULTY OF INTERNATIONAL ECONOMICS -*** - ECONOMETRICS REPORT THE IMPACT OF GDP PER CAPITA AND OTHER FACTORS ON LIFE EXPECTANCY IN SOME COUNTRIES IN 2015 Instructor: PhD Đinh Thị Thanh Bình Group: Course: KTEE309.1 Hanoi, October 2021 GROUP MEMBERS Name STT Student ID Contribution (%) Đỗ Quỳnh Chi 2012150014 20% Nguyễn Thị Hải Hà 17 2012150026 20% Nguyễn Phương Hoa 24 2012150034 20% Đỗ Thị Minh Huyền 28 2012150041 20% Trần Hải Diệu Linh 42 2012150054 20% TABLE OF CONTENTS INTRODUCTION CHAPTER I LITERATURE REVIEW CHAPTER II RESEARCH METHODOLOGY AND MODEL BUILDING Research methodology 1.1 Collecting data method 1.2 Processing data method Model building 2.1 Identifying type of model 2.2 Investigated variables and measurement of investigated variables Data description 3.1 Data source 3.2 Data statistics description 10 3.3 Correlation among variables in model 10 CHAPTER III CHECK FOR THE PROBLEMS OF THE MODEL AND STATISTICAL INFERENCE 13 Model estimation 13 Testing and fix the defects of the model 14 2.1 Test the model's omitted variables (the correct form of the model) 14 2.2 Multicollinearity testing 15 2.3 Heteroskedasticity testing 15 2.4 Normality of u testing 17 Final regression model and estimation result 18 Testing for the overall significance of the model and testing for the significance of the independent variables 19 4.1 Testing for the overall significance of the model 19 4.2 Testing for the significance of the independent variables 19 Analyzing the estimated results and policy implication 21 5.1 The effect of GDP per capita on national life expectancy 21 5.2 The effect of the proportion of deaths caused by non-communicable diseases (% of total deaths) on national life expectancy 21 CONCLUSION 22 REFERENCES 23 APPENDIX 24 INTRODUCTION Econometrics is the application of statistical methods to economic data in order to give empirical content to economic relationships While econometric theory uses statistical theory and mathematical statistics to evaluate and develop econometric methods, applied econometrics uses theoretical econometrics and realworld data for assessing economic theories, developing econometric models, analysing economic history, and forecasting Gross Domestic Product (GDP) is a monetary measure of the market value of all the final goods and services produced in a specific time period As a measurement, it is often described as being a calculation of the total size of an economy Therefore, citizens' quality of life is closely linked with GDP per capita as it affects our living conditions, our access to education and healthcare, and even our opportunities It is important to understand the impact of GDP per capita on standard of living and how that in turn affects our longevity As economics-based students, realizing the importance of applying econometric methods in research and problem analysis, our group decided to write a report using the Ordinary Least Square Regression method (OLS) This paper explores the impact of GDP per capita and some other factors on life expectancy GDP per capita is the primary independent variable, while life expectancy is the dependent variable We would like to express our deepest appreciation towards PhD Đinh Thị Thanh Bình – lecturer of Econometrics class for providing us with knowledge, advice as well as detailed instruction throughout our researching and conducting process for this report Given the fact that this is our first time conducting research using an econometric method, due to the lack of practical knowledge, research methods and execution time, the inevitable shortcomings are for sure Our group is looking forward to getting the opinions of your precious Master with a view to fulfilling our knowledge in this field CHAPTER I LITERATURE REVIEW Several papers have acknowledged the relationship between GDP per capita and life expectancy, however, the correlation is still not completely understood (Dayanikli, et al., 2016) examined an association between a country’s GDP and mortality rates across different nations in 2013 The study found a correlation between income and life expectancy, with higher income being associated with a longer life expectancy The most significant conclusion of this paper is that GDP per capita only affects life expectancy up until a certain threshold After some GDP, the correlation between the variables weakens, which can be represented by their analysis: below-median GDP to life expectancy regression is stronger than the above-median GDP to life expectancy regression Additionally, this paper also points out that while education may very well lead to increases in life expectancy, there is a strong collinearity between increases in education and increases in per capita GDP, and thus the explanation cannot be labeled as causal (Taylor, 2021) also found that national income was positively correlated to life expectancy More specifically, for every tenfold increase in GDP per capita a country can expect the life expectancy of its citizens to increase by approximately 11 years Additionally, there is a statistically significant relationship between population growth and life expectancy, although it’s effect is much smaller than that of national income For every unit increase of 1% in population growth one can expect to see the average life expectancy of a country to fall by 0.81 years Furthermore, this paper believed that the relationship between national income and life expectancy is causal, since countries that are able to produce more with a smaller population (the result being a higher GDP per capita) are able to improve their healthcare infrastructure and its citizens are likely to increase their standard of living overtime in comparison to citizens of poorer countries In (Deshpande, et al., 2014), apart from recognizing the positive correlation between national income and life expectancy, they also examined the relationship between national health expenditure and life expectancy Based on the data of health expenditure, national income, government spending, literacy rate and physician density from 81 countries, including both developed, developing and underdeveloped ones, this paper shows that there is no significant correlation between healthcare spending and life expectancy in developing countries, but it does exist in developed countries Additionally, when the multiple regression for least developed countries was run, the only statistically significant variable is physician density, which was significant at the 1% level, specifically indicating that, in developing countries, access to healthcare is a large issue In these places, the fact that there is an available doctor nearby can have a significant impact, therefore the variable of physician density can have a statistically significant impact when it comes to developing countries (Jetter, et al., 2016) leveraged a gigantic dataset of 197 countries over 213 years (1800 to 2012), leading to a systematic and economically sizable relationship between income levels and life expectancy This paper’s estimations produced firm evidence of a consistently positive relationship until a value of approximately US$15,478 (using international price levels in 2005), corresponding to approximately 95 percent of the 4,325 sample observations GDP per capita alone is able to explain over 64 percent of the variation in life expectancy across countries and years Overall, this paper suggests that income levels are by far the strongest factor in raising life expectancy across the globe (Maity, et al., 2017) analyzed the relationship between different variables and life expectancy across several countries This paper has studied the topic with a new approach by including uncommon independent variables such as GNI per capita (PPP), poverty headcount ratio at $1.00, among others However, the variables chosen are probably not the best measures of average life expectancy Although average life expectancy can be influenced by gender, genetics, lifestyle, etc and though these variables might be correlated with some of the variables that were studied in this paper, since the studied variables not necessarily have direct influence on average life expectancy The study concluded that further analysis should be done in order to study good health and well-being, including using a dependent variable such as infant mortality, which may be more easily affected by the dependent variables studied in this paper This paper seeks to find the relationship between not only life expectancy and variables like GDP per capita, health expenditure and Gini Index but also with variables that have not been fully leveraged in research related to this topic such as poverty per headcount or cause of death by non-communicable diseases across different countries all around the world Based on the relationship found by previous studies, our hypothesis is that countries with higher levels of national income (as measured by GDP per capita) and spending on health expenditure will have higher life expectancies Citizens of countries that have a higher level of national income likely lead healthier lifestyles and have access to better healthcare infrastructure, both of which are contributing factors in life expectancy Additionally, countries with a booming population are more likely to have a higher life expectancy As examined in one study mentioned above, in countries with a large population, the nationals have higher life expectancy compared to other countries CHAPTER II RESEARCH METHODOLOGY AND MODEL BUILDING Research methodology 1.1 Collecting data method Cross-sectional data was collected across 200+ countries The data table contains 80 observations The data is synthesized from World Bank (2015) 1.2 Processing data method Using Excel and Stata to process data and correlation matrix among variables Model building 2.1 Identifying type of model Model containing variables: - Dependent variable: Life expectancy at birth · - Unit: years Independent variable: variables Ø GDP per capita · Unit: US$ Ø Population · Unit: people Ø Gini index (World Bank estimate) · Unit: none Ø Poverty headcount ratio at $1.90 a day (2011 PPP) · % of population Ø Current health expenditure · % of GDP Ø Cause of death, by non-communicable diseases · % of total 2.2 Investigated variables and measurement of investigated variables 𝑙𝑖𝑓𝑒 = 𝛽0 + 𝛽1 𝑙𝑜𝑔𝑔𝑑𝑝𝑐𝑎𝑝 + 𝛽2 𝑙𝑜𝑔𝑝𝑜𝑝 + 𝛽3 𝑔𝑖𝑛𝑖 + 𝛽4 𝑝𝑜𝑣 + 𝛽5 ℎ𝑒𝑎𝑙𝑡ℎ𝑒𝑥𝑝 + 𝛽6 𝑛𝑐𝑑 + 𝑢̂ · loggdpcap: GDP per capita · logpop: Population · gini: Gini index (World Bank estimate) · pov: Poverty headcount ratio at $1.90 a day (2011 PPP) · healthexp: Current health expenditure · ncd: Cause of death, by non-communicable diseases Estimated means of variables: ● 𝛽1 >0: GDP per capita increases, life expectancy at birth increases ● 𝛽2 chi2 = 0.0088 Source: Run hettest command in STATA With significance level α=5% From this result, with p-value = 0.0088 < α=5% => Reject H0 Conclusion: Model has heteroskedasticity With heteroskedasticity, the OLS estimators are still linear and unbiased, but not the best However, the variance of the coefficient will be biased, which will lead to an inexact hypothesis test To solve this problem, we use the Robust Standard Errors method Table 5: Regression result from Robust standard error method -| life | Robust Coef Std Err t P>|t| [95% Conf Interval] -+ -loggdpcap | 2.553445 2493826 10.24 16 0.000 2.056426 3.050464 logpop | 0691501 1702427 0.41 0.686 -.2701431 4084433 gini | 0321522 049574 0.65 0.519 -.0666487 1309532 pov | -.0537637 029349 -1.83 0.071 -.1122562 0047289 healthexp | 1595712 1718498 0.93 0.356 -.1829249 5020673 ncd | 1410777 0397491 3.55 0.001 0618578 2202975 _cons | 37.48252 4.310312 8.70 0.000 28.89208 46.07296 Source: Run reg life loggdpcap logpop gini pov healthexp ncd, robust command in STATA The estimated coefficients of the regression model are unchanged; however, the standard error has been fixed into their robust ones Therefore, heteroskedasticity has been solved 2.4 Normality of u testing If ui does not have normal distribution, T statistics may have no T-distribution Therefore, the hypothesis test is inexact To test the normality of u, we use the Jaque – Bera test, with significance level α=5% Examine the hypothesis: H0: ui has normality distribution H1: ui does not have normality distribution Table Jaque - Bera test result Variable | Obs Pr(Skewness) Pr(Kurtosis) adj chi2(2) Prob>chi2 -+ u | 80 0.0096 0.0780 8.58 p-value = 0.0137 < 0.05 => Reject H0 ⇨ ui does not have normal distribution However, with quite a large number of observations (80), the estimations and hypothesis testings are still credible 17 0.0137 Final regression model and estimation result The final regression model is: 𝑙𝑖𝑓𝑒 = 𝛽0 + 𝛽1 𝑙𝑜𝑔𝑔𝑑𝑝𝑐𝑎𝑝 + 𝛽2 𝑙𝑜𝑔𝑝𝑜𝑝 + 𝛽3 𝑔𝑖𝑛𝑖 + 𝛽4 𝑝𝑜𝑣 + 𝛽5 ℎ𝑒𝑎𝑙𝑡ℎ𝑒𝑥𝑝 + 𝛽6 𝑛𝑐𝑑 + 𝑢̂ The final estimation result is: reg life loggdpcap logpop gini pov healthexp ncd, robust Linear regression Number of obs = 80 F(6, 73) = 101.46 Prob > F = 0.0000 R-squared = 0.8684 Root MSE = 2.3507 -| Robust life | Coef Std Err t P>|t| [95% Conf Interval] -+ -loggdpcap | 2.553445 2493826 10.24 0.000 2.056426 3.050464 logpop | 0691501 1702427 0.41 0.686 -.2701431 4084433 gini | 0321522 049574 0.65 0.519 -.0666487 1309532 pov | -.0537637 029349 -1.83 0.071 -.1122562 0047289 healthexp | 1595712 1718498 0.93 0.356 -.1829249 5020673 ncd | 1410777 0397491 3.55 0.001 0618578 2202975 _cons | 37.48252 4.310312 8.70 0.000 28.89208 46.07296 Number of observations 80 Coefficient of determination R2 (R-squared) 0.8684 P-value 0.0000 18 F(6, 73) 101.46 Root Mean Square Error (Root MSE) 2.3507 R-squared is equal to 0.8684, which means that the independent variables in the model can explain 86.84% of the sample variation of life expectancy This indicates a strong association The sample regression function is: 𝑙𝑖𝑓𝑒 = 37.48252 + 2.553445𝑙𝑜𝑔𝑔𝑑𝑝𝑐𝑎𝑝 + 0.0691501𝑙𝑜𝑔𝑝𝑜𝑝 + 0321522𝑔𝑖𝑛𝑖 − 0.0537637𝑝𝑜𝑣 + 0.1595712ℎ𝑒𝑎𝑙𝑡ℎ𝑒𝑥𝑝 + 0.1410777𝑛𝑐𝑑 + 𝑢̂ Testing for the overall significance of the model and testing for the significance of the independent variables 4.1 Testing for the overall significance of the model Testing for the overall significance of the model at the 95% confidence interval (α=5%): {𝐻0 : 𝛽1 = 𝛽2 = 𝛽3 = 𝛽4 = 𝛽5 = 𝛽6 = 𝐻1 : 𝛽12 + 𝛽22 + 𝛽32 + 𝛽42 + 𝛽52 + 𝛽62 ≠ The P-value associated with the F-value is 0.0000, smaller than α=0.05; therefore, we reject the H0 hypothesis In conclusion, the model is statistically significant 4.2 Testing for the significance of the independent variables Testing for the significance of the independent variables at the 95% confidence level (α=5%): {𝐻0 : 𝛽𝑚 = 𝐻1 : 𝛽𝑚 ≠ , 𝑚 ∊ {1, 2, 3, 4, 5, 6} This is tested by comparing each of the P-values to α: 19 Because 𝑃 − 𝑣𝑎𝑙𝑢𝑒1 = 0.000 < 0.05, we reject the hypothesis H0 for the independent variable loggdpcap Because 𝑃 − 𝑣𝑎𝑙𝑢𝑒2 = 0.686 > 0.05, we accept the hypothesis H0 for the independent variable logpop Because 𝑃 − 𝑣𝑎𝑙𝑢𝑒3 = 0.519 > 0.05, we accept the hypothesis H0 for the independent variable gini Because 𝑃 − 𝑣𝑎𝑙𝑢𝑒4 = 0.071 > 0.05, we accept the hypothesis H0 for the independent variable pov Because 𝑃 − 𝑣𝑎𝑙𝑢𝑒5 = 0.356 > 0.05, we accept the hypothesis H0 for the independent variable healthexp Because 𝑃 − 𝑣𝑎𝑙𝑢𝑒6 = 0.001 < 0.05, we reject the hypothesis H0 for the independent variable ncd In conclusion: The independent variables that have a statistically significant effect on life expectancy at the 95% confidence level (α=5%) are: loggdpcap, ncd The independent variables that not have any statistically significant effect on life expectancy at the 95% confidence level (α=5%) are: logpop, gini, pov, healthexp 20 ... write a report using the Ordinary Least Square Regression method (OLS) This paper explores the impact of GDP per capita and some other factors on life expectancy GDP per capita is the primary independent... 5.1 The effect of GDP per capita on national life expectancy 21 5.2 The effect of the proportion of deaths caused by non-communicable diseases (% of total deaths) on national life expectancy. .. impact of GDP per capita on standard of living and how that in turn affects our longevity As economics-based students, realizing the importance of applying econometric methods in research and problem