Undergraduate Economic Thesis

Introduction

A goal underlying economic policy is to positively impact the quality of life. This paper analyzes which variables can predict quality of life, and which quality of life aspects are more easily predicted. This information is largely relevant to policymakers across the world, since knowing quality of life predictors could be very useful in crafting economic policy. It is also relevant to concerned and active voters interested in choosing the policymakers most qualified for the situation at hand. For instance, if statistics indicate a nation has a relatively low life expectancy, policymakers may wish to orient policy in a manner attempting to manipulate the variables that indicate potential influence on life expectancy.

The Theoretical Models

            I used three possible measures for Quality of Life: Life Expectancy, Gross Domestic Product per Capita (abbreviated GDP/Capita), and Gross Tertiary Enrollment Ratio (GTER). Each measure is used as a dependent variable in its own multiple regression model with three to five explanatory variables and 5% significance. Data for all models are from the year 2010 and come from DataWorldBank online. Because of the many missing data points, a few compromises were necessary in the analysis. Some variables were omitted due to insufficient data, and many countries were left out of the analysis because they lacked data for one or more explanatory variables. A list of countries may be found in the appendix, 2. The models, accompanying economic theory, and descriptive data statistics are discussed below.

Life Expectancy Model

            In the first model, the dependent variable is life expectancy at birth (denoted LE). In evaluating the relative prosperity of a nation, one would likely take into account the country’s life expectancy. Policymakers striving for prosperity may be interested in which variables, if any, can predict life expectancy with statistical significance. A likely follow-up question is which of the significant predictors, if any, have a causal relationship with life expectancy. The answers to these questions would be useful in deciding which variables to manipulate in economic policy. Chen and Ching of built a similar multiple regression model to that described above. In their analysis, Life Expectancy was predicted by Gross National Product Per Capita, Annual Population Growth, Fertility Rate, Aids Prevalence, Tuberculosis Prevalence, School Enrollment Rate, Percent of Population with Clean Water Access, and Annual Rate of Deforestation. While our model used slightly different explanatory variables, Chin and Cheng’s analysis used data from the year 2000 rather than 2010.

            The first variable used to predict life expectancy is public health expenditures as a percentage of Gross Domestic Product (denoted HEALTH$). The chosen indicator is a proportion rather than the value of gross expenditures, which adjusts for differences in scope of countries’ economies. With increased federal spending on health, there are likely to be resultant increases in: access to medical supplies, access to medical facilities, and number of physicians – all of which should positively impact life expectancy.

            The second variable used to predict life expectancy is percent of total population living in urban areas (denoted %URBAN). Chen and Cheng describe cities in relation to life expectancy as, “centers of medicine and modern advance, but also quarters of overpopulation and overcrowding.”(Chen, Ching, 5) Since there is typically greater access to medical facilities in cities, it makes intuitive sense that a larger proportion of a population living in urban areas would result in a larger proportion with access to medical care and thus a higher life expectancy.

            The third explanatory variable in the life expectancy regression model is percentage of urban population with access to clean water (denoted %CLEAN). Few will argue that environmental factors will influence life expectancy in a country. In Chen and Ching’s model, they address this by saying, “Environmental soundness measured in the forms of clean drinking water…[it] is a reflection on the salutary conditions of the country.”(Chen, Ching, 7) Because a deprivation of clean water access is likely to decrease life expectancy, a positive relationship and thus a positive coefficient is expected for this variable.

            The fourth and final explanatory variable in this model is value of international trade as a percentage of Gross Domestic Product (denoted TRADE). Unlike Net Exports, this indicator is calculated by [(X+M)/GDP]. Similar to with public health expenditures, the international trade indicator is a proportion, which adjusts for the size differences in economies. Countries that isolate themselves from international trade could very well be missing valuable medical supplies available abroad. It makes sense then that an increased value of international trade could increase access to medical supplies and thus increase life expectancy – implying a positive relationship and positive coefficient value.

Descriptive Statistics

Descriptive Statistics	Dependent Variable	Variable 1	Variable 2	Variable 3	Variable 4
	Life Expectancy	HEALTH$	URBAN%	%CLEAN	TRADE
Mean	70.278	7.0799	57.123	95.212	24.126
Standard Deviation	9.486	2.597	22.456	7.209	24.447
Observations	147	147	147	147	147

Analysis Results

 = 2.819 + .045(HEALTH$) + .175(URBAN%)* + .599(%CLEAN)* + .028(TRADE)

                           (.214)                           (.027)                     (.084)                     (.028)

F=46.20, p<.05; adj-R2=.555; dfbetween=4; observations=146

            The regression results indicate that the coefficient parameter for public health expenditures is insignificant in predicting life expectancy.  

            As expected, the coefficient value for percentage of urban population living in urban areas was positive and significant. The model suggests that the marginal effect of this is explanatory variable is as follows: each additional percentage of a population living in an urban area increases life expectancy by .175 years. Chen and Ching’s description of cities as “medical centers” is likely the primary explanation for the positive relationship.

            The coefficient value for percentage of urban population with clean water access was also positive and significant. The marginal effect of this variable is: each additional percentage point of a nation’s urban population with clean water access adds .599 years to life expectancy. That increased access to clean water will allow for a longer life is relatively straightforward and needs little explanation.

            Similar to public health expenditures, the model suggests international trade is not a significant predictor of life expectancy.

Gross Domestic Product per Capita Model

            The first variable used to predict GDP/Capita is international trade value as a percentage of GDP (TRADE).  Just as in the last model, this indicator sums exports and imports and takes the value as a proportion of GDP. A regression analysis by Bidlingermaier examines the relationship between income and international trade. He lists the following description of the gains from trade theory:

“These gains stem from specialization in production due to international trade. If countries specialize according to comparative advantage enhanced resource allocation can be achieved…Consequently, the welfare (income) of all trading nations is improved.” (Bidlingmaier, 1)

It makes intuitive sense that both exports and imports would positively impact GDP/Capita. With increased exports will come increased sales for domestic producers. Furthermore, increased imports imply cheaper production inputs for domestic producers. Because of these two factors, a positive coefficient value is expected for this variable.

         The second explanatory variable used in the GDP/Capita model is population density (denoted POPDENSE). Chen and Ching suggest that the relevance of population growth comes from the idea that a larger population implies the same resources must be spread across more people, thus decreasing prosperity. From this, an inverse relationship and a negative coefficient value is expected for population density.

            The third variable used to predict GDP/Capita is net immigration (denoted NETMIG). This indicator is calculated by subtracting the total number of emigrants from the number of immigrants for a five-year interval. Borjas’s labor market model helped build a prediction for the coefficient sign for this variable.

“When immigrants enter the country, the supply of labour expands… and the market wage falls to W₁ (all other things being equal). As a result, native workers earn a lower wage. Total employment increases. The economy’s total output also expands. Total output is represented by the area under the marginal product curve and to the left of the supply curve. This area is larger following the increase in labour supply. The expansion in output generates an increase in income for the owners of capital in local firms (and, of course, income for immigrants). Under certain conditions the loss in income for native workers is more than offset by the increase in income accruing to the owners of capital. The result is a net increase in national income.” (Moody, 11)

From this information, a positive coefficient value is expected for this independent variable.

            The fourth explanatory variable used in this model is average insolvency time (denoted INSOLVENCY). This indicator refers to the average number of years passed between initial court filing and resolution for distressed assets cases. A larger value for this indicator may imply a larger level of instability and risk, and most likely implies a lower level of liquidity. Reduced risk would likely incentivize investment spending and loans, since there is a greater chance of being repaid. Also, a reduction in liquidity would likely result in a lower level of consumption and investment spending, both of which are components of GDP. These factors suggest that this variable will be inversely related to GDP and will have a negative coefficient.

            The fifth and final explanatory variable in the GDP/Capita model is the Logistics Performance Index (denoted LPI). This indicator ranges from one to five, with a higher score suggesting more impressive logistics.  Metrics used to build this composite indicator include: quality of trade and transport-related infrastructure, frequency of shipments arriving on schedule, and ability to track and trace shipments. Because infrastructure is a key component (and somewhat of a pre-requisite) of economic growth, this predictor variable is expected to be positively related to GDP/Capita and have a positive coefficient.

Descriptive Statistics

	Dependent Variable	Variable 1	Variable 2	Variable 3	Variable 4	Variable 5
	GDP/CAP	TRADE	POPDENSE	NETMIG	IINSOLVENCY	LPI
Mean	14250.04	2304.045	249.791	-1058.72	2693.701	2950.234
Standard Dev.	18828.015	2403.578	887.332	777602.14	1277.44	554.27
Observations	129	129	129	129	129	129

Analysis Results

 = 58502.65 + 2.043(TRADE)* + (-2.978)(POPDENSE)* + .0027(NETMIG)*

                                                  (.387)                        (1.061)                         (.0011)

+ (-1.356)(INSOLVENCY) + 24.56(LPI)*

                      (.761)                         (1.792)

F=71.69, p<.05; adj-R2=.737; dfbetween=5; observations=127

            In the GDP/Capita model, four of the five explanatory variables had significant parameter estimates. The adjusted R Square statistic suggests that the model explains 73.72% of variation in GDP/Capita among nations. The F-test indicates the overall model has statistical significance.

            The coefficient value for international trade was positive and significant. The marginal effect of international trade on GPD/Capita suggested by the model is: each incremental percentage increase in international trade value as a percentage of GDP increases GDP/Capita by $2043.08. Trade theory’s explanation of improved efficiency in resource allocation resulting in cheaper domestic inputs and increased domestic sales abroad is the most probable explanation of the positive relationship between international trade and GDP/Capita.

            As expected, the coefficient value for population density was negative and significant. Calculating the marginal effect of population density on GPD/Capita gives us: each incremental increase in population density leads to a decrease in GDP/Capita equal to $2.98. The most probable explanation for the negative coefficient is Chen and Ching’s argument of resources becoming scarcer when population is increased.  

            A positive coefficient value was hypothesized correctly for net migration, which was found to be significant. The marginal effect of net migration on GDP/Capita is as follows: each incremental increase in net migration (increase by one person) leads to an increase in GDP/Capita of $.003. Borjas’s labor market model offers a possible explanation for the positive relationship in saying that increased immigration implies an increased labor supply which in turn leads to a net increase in national income. Since the market wage rate also falls, this opposing force could explain the small coefficient value.

            The results of the regression analysis suggest that average insolvency time has no predictive significance with respect to GDP/Capita.

            Finally, the coefficient sign was correctly hypothesized as positive and significant for the explanatory variable Logistics Performance Index. The marginal effect of this variable on GDP/Capita is as follows: each increase in Logistics Performance Index by .001 leads to an increase in GDP/Capita by $24.56. As stated in the model explanation, infrastructure is somewhat of a pre-requisite for economic growth. A weak infrastructure makes economic development significantly more difficult. This is a likely explanation for the positive relationship among these two variables.

College Enrollment Model

            In discussing quality of life, education is rarely a topic that will go unmentioned. Furthermore, most people would regard increased education opportunities as an increase in the standard of living. The indicator used to measure post-secondary enrollment is called Gross Tertiary Enrollment Ratio, which is calculated by dividing the total number of students enrolled in post-secondary school by the number of citizens in the five-year age group following completion of secondary school.

            The first variable used to predict college enrollment is preprimary enrollment rate (denoted PREPRIMARY). This indicator is calculated by total number of students enrolled in preprimary education by the total number of citizens in the official preprimary education age group. A higher preprimary enrollment rate could have a few possible implications. First, a higher proportion of students enrolled in preprimary education could imply greater access to education opportunities. Second, the higher proportion could imply a higher societal value placed on education – since parents who more highly value education would likely enroll their children at a younger age. Because of these two factors, a positive relationship and positive coefficient is expected for preprimary enrollment rate.

            The second explanatory variable used in this model is Female to Male Secondary Enrollment Ratio (denoted FEMALE:MALE). This indicator is calculated by dividing the number of females enrolled in secondary education by the number of males enrolled. A higher female to male ratio is likely an indicator of increased gender equality with respect to education. A possible implication of increased gender education equality is a higher societal value placed on education. Also, a larger proportion of women in secondary school will likely result in a larger proportion of women enrolled in college. Because a low value for this indicator could imply gender inequality with respect to women, a high value likely implies more women enrolled in college and thus a higher Gross Tertiary Enrollment Ratio. The coefficient for this predictor is expected to be positive.

            The third and final predictor used in the post-secondary education model is prevalence of HIV (denoted HIV). This indicator refers to the percentage of citizens aged 15-49 infected with the HIV virus. An increase in the proportion of a population infected with HIV would likely result in a reduction in post-secondary enrollment, since sickness and death would likely affect many enrolled in post-secondary school. It makes intuitive sense that these two variables would be inversely related, giving this explanatory variable an expected negative coefficient.

Descriptive Statistics

Descriptive Statistics	Dependent Variable	Variable 1	Variable 2	Variable 3
	GTER	PREPRIMARY	FEMALE:MALE	HIV
Mean	46.142	68.741	96.793	68.933
Standard Deviation	28.463	35.971	12.133	97.919
Observations	75	75	75	75

Analysis Results

 = 8.524 + .451(PREPRIMARY)* + .129(FEMALE:MALE) + (-.086)(HIV)*

                                           (.073)                                  (.229)                      (.025)

F=46.20, p<.05; adj-R2=.554; dfbetween=3; observations=75

In the Post-Secondary Enrollment model, two of the three explanatory variables had predictive significance. Explanatory variables omitted from the model may be found in the appendix, footnote 1.

             The regression results returned a positive and significant coefficient value for Pre-Primary Enrollment Rate, which matched expectations. Calculating the marginal effect of this variable gives us: each percentage increase in pre-primary enrollment yields an increase of .451% of Gross Tertiary Enrollment Ratio.

            Regarding female:male secondary enrollment ratio as an explanatory variable, the analysis suggests no significance in predicting life expectancy.

            As expected, the coefficient value for HIV prevalence was negative and significant. The marginal effect of HIV prevalence on Post-Secondary Enrollment is as follows: each 1% increase in HIV prevalence yields a reduction in Gross Tertiary Enrollment Ratio of 8.6%.

Conclusion

            As it turned out, GDP/Capita was the easiest quality of life variable to predict. The Life Expectancy and Post-Secondary Enrollment offered less explanatory ability than did the earnings model. While all R Square terms are above .53, the analysis results should be interpreted with caution. Rather than establish significant explanatory variables as causal factors for quality of life, this paper merely lays groundwork for a further analysis of determining causality. Possible improvements that could have been made given more time include adding more explanatory variables to increase predictive ability and increasing size of data sets to include more countries and reduce bias in the data.