This item is only available as the following downloads:
A Global Analysis of Infant Mortality By, Kavida Gobalasingham University of Florida College of Liberal Arts and Sciences Fall 2013 Economics Honors Thesis Advisor: Professor Lawrence Kenny
Page | 1 Abstract: analyzed to determine how we may improve the state of global health Using eight independent variables, I conduct an empirical analysis of the 2013 infant mortality rate s in 150 countries. I utilize a single multivariate regression with the following independent variables for each country: gross domestic product per capita (using purchasing power parity), the female literacy rate, the percent of the population with access to improved drinking water sources, physician density, percent of households using solid fuels as their primary energy source for cooking and heating, prevalence of HIV/AIDS (ages 15 49), prevalence of malaria, and the democracy index. I performed a one t ailed t test for every variable. GDP per capita (PPP) and physician density were not significant The other six variables were significant at the one percent level. The female literacy rate had the highest impact, which indicates that the female literacy rate has the greatest influence on infant mortality rates of the se variables. I. Introduction: The The infant mortality rate (IMR) is defined as the number of infants who die before they are one year old per 1,000 live births within the same year. 1 Since 1990, t he global infant mortality rate has dropped from sixty one to approximately thirty eight in 2013 2 Although this rate has been declin ing, infant mortality is still rampant particularly in sub Saharan Africa and South Asia. As of 2013, infant mortality is as high as 119.41 in Afghanistan and 106.49 in Mali. In most highly industrialized countries, the IMR is around two. 1 B y understanding which factors significantly a ffect the IMR, we can implement changes and policies to combat infant mortality and improve health care world wide. II. Significant Results from Previous Literature There have been a multitude of studies performed on infant mortality. For the most part, these studies have focus ed on a single country or geographic region (e.g. India, sub Saharan Africa, or Southeast Asia) There have been far fewer global analyses performed. Previous research has shown that economic factors, such as GDP per capita, average income, and income equality, have a significant and negative correlation with infant mortality rates. That is, as inco me rises, infant mortality rates 1 Infant Mortality Rate (accessed July 6, 2013); available from https://www.cia.gov/library/publications/the world factbook/rankorder/2091rank.html 2 Infant Mortality." WHO. World Health Organization, 2013. Web. 4 Sept. 2013.
Page | 2 should fall In a 2005 study Edward G. Stockwell, Franklin W. Goza and Jack L. Roach analyzed the relationship between income levels and infant mortality in the state of Ohio. Comparing five economically distinct regions of Ohio, they found that there was a disparity in mortality r ates between wealthy and poor families. Higher levels of income corresponded to lower infant mortality rates. 3 Additionally, in 2004, Oystein Kravdal from the University of Oslo published a stud y about child mortality rates in India. had a significant and negative correlation with child mortality at the one percent level. 4 Furthermore, Carl Otto Schell, Marie Reilly, Hans Rosling, Stefan Peter son, and Anna Mia Ekstrom did a study of infant mortality in 152 countries using data from 2003. They found that gross national income per capita, young female illiteracy, and income equality were significant and predicted 92% of the variation in national infant mortality rates. They found public health spendi ng and poverty rates to be insignificant 5 These three studies are among the most highly cited works on infant and child mortality. I have found minimal representation of non socioeconomic variables in infant mortality analyses. Therefore, i n addition to analyzing GDP per capita and female literacy I also assess the significance of environmental conditions with two variables: the perc ent access to improved drinking water sources and household solid fuel use Additionally, I include the prevalence of both HIV/AIDS and malaria in order to evaluate the effect of disease on mortality. In order to capture the effect of the amount of medical resources on mortality, I also include the physician density, which is the number of physicians per 1000 people. The data used in my regression are also quite recent (most data are from 2010 2013), which should serve to provide a very current analysis of the issue of infant mortality. III. Sample 3 Mortality in a Metropolitan Aggregate, 1989 Sociological Forum, Vol. 10, No. 2, 1995. Accessed Septemb er 4, 2013. Available from Jstor. 4 Population Studies, Vol. 58, No. 2, 2004. Accessed August 24, 2013. Available from JStor. 5 s of infant mortality: a worldwide study of 152 low middle and high income Scandinavian Journal of Public Health, Vol. 35, Issue 3, May 2007. Accessed September 9, 2013. Available from the National Center for Biotechnology Information, U.S. National Library of Medicine.
Page | 3 The sample for this regression is 150 countries, which includes high, middle, and low income countries. I included all countries with data available for each variable. IV. Methodology I use one multivariate regression to assess the significance and correlation of the independent variables with the IMR. For a variable to be significant in a two tailed t test, its t five percent confidence level. For a one tailed t est, the absolute value of the t statistic must be at least 1.67 at the ninety five percent confidence level. In this regression, I used a one tailed t test for each variable because the alternative hypothesis for each variable was in a single direction. I also calculated the impact of each independent variable by multiplying the standard deviation with the coefficient. By looking at the p values, we can assess at which level each variable is significant. The p values that were generated by E xcel are for a two tailed test. For a on e tail ed test, we divide the E xcel p values by two. For a variable to be significant, the halved p value must be less than .10 for a 10 percent significance level, less than .05 for a 5 percent level, and less than .01 for a 1 pe rcent level. Attached are the regression results and summary statistics of ea ch variable. As shown in the correlation matrix none of the values for r (the correlation coefficient) exceed an absolute value of 0.80, which means that multicollinearity should not be a concern. V. Dependent Variable : Infant Mortality Rate The data for the current infant mortality rates ( as of 2013) were obtained from the Central Intelligence Agency (CIA) World Factbook. The CIA World Factbook defines the
Page | 4 infant mortality rate (I MR) as the number of deaths of infants under the age of one per 1,000 live births in the same year. 6 VI. Independent Variable #1 : GDP per capita (PPP) in 2012 U.S. $ The first independent variable is the gross domestic product per capita (GDP per capita), wh ich uses a purchasing power parity basis (PPP) By using PPP instead of just GDP per capita, we are able to avoid being misled by fluctuations in exchange rates. Another advantage of using the PPP is each value is calculated by using the same set of prices (in this case, U.S. 2012 prices). This value gives the physical amount of newly produced goods and services per average citizen in a country. The values were given in terms of 2012 United States dollars, were obtained from the CIA World Factbook and were from the year 2012. A few countries without 2012 data have figures from either 2010 or 2011 The most recent figures were used for each country. Per capita GDP was used instead of the aggregate GDP because it gives a more a ccurate picture of the resources of an individual citizen. For example, India's GDP of $ 4,761,000,000,000 is greater than Germany's GDP of $ 3,250,000,000,000. 7 However, from this information, we cannot tell that the average German citizen has a higher st andard of living than the average Indian citizen. We know that the average German citizen is economically better off than the average Indian citizen because the GDP per capita of Germany ($39,100) is greater than that of India ($3,900). 8 6 "Infant Mortality Rate." The World Factbook. Central Intelligence Agency, 2013. Web. 6 July 2013.
Page | 5 We predict that a country with a higher GDP per capita will have a lower infant medi cations, clothing, shelter, adequate food and other resources for their children. Therefore, we wo uld expect a negative coefficient; as GDP per capita rises, the IMR should fall. In the regression GDP per capita (PPP) was not significant with a t statistic of 0.73 2 and a p value of 0.233 Surprisingly, this variable did not conform to the results of previous literature. The other studies I reviewed on this topic focused either only on economics variables (inequality, poverty rate, etc.) or only on economic and educational variables. I chose to use GDP per capita because of the lack of recent data on G ini coefficients and poverty rates for all 150 countries in my sample. Perhaps because I included a variety of variables pertaining to disease, democracy, physician density, and environmental issues, the effect of GDP per capita on infant mortality became negligible. This seems plausible because the United States is the world leader in aggregate GDP and has the fifth highest per capita GDP (PPP), yet it only has the thirty second lowest infant mortality rate. This analysis suggests that GDP per capita does not affect infant mortality rates. However, we simply cannot dismiss the fact that many studies have confirmed the si gnificance of income on the IMR. VII. Independent Variable #2: Female Literacy Rate (%) The next independent variable is the female literacy rate. This number is the percentage of women who are literate out of the population of women in a country. Most typically, a person must be able to read and write by the age of fifteen to be
Page | 6 9 A few countries use age sixteen, age t welve, or age eighteen Most of the data are from 2012. However, some countries have not updated their numbers for the past several years because they have already reached 99 100 percent literacy. For examp le, the United States, France, and the United Kingdom had a 99 percent female literacy rate in 2003 and have not updated their rates since then. It is reasonable to assume that literacy rates have not changed for these countries since literacy tends to imp rove over time, and 99 100 percent is the maximum level of literacy that can be attained. The latest available rate for each country is used in the regression. These data are also from the CIA World Factbook. Because literate women are more educated, they should have more knowledge about prenatal care, proper nutrition, and health. When women are equipped with this information, infants should be well cared for an d face a lower risk of mortality We predict that an increase in t he literacy rate of women wil l lead to a decrease in the infant mortality rate (negative coefficient). As predicted, this variable ha d a negative coefficient ( 0.366 ). This means that as female literacy rises by one percent, infant mortality falls by 0.366 deaths per 1000 live births. It was significant at the one percent level with a t statistic of 5.96 4 and a p value of 9.417 x 10^ 9 Ad ditionally, the impact was 8.37 7 which was the highest of a ll the variables. This means that the female literacy rate has the great est effect on i nfant mortality of these variables. Based on this result, if a country would like to reduce its infant mortality, then it should try to create more educational opportunities for women. 9 "Literacy." The World Factbook. Central Intelligence Agency, 2012. Web. 29 July 2013.
Page | 7 VIII. Independent Variable #3: Access to Improved Drinking Water Sources (%) The next independent variable is access to improved drinking water sources. This value is the percent of indiv iduals within a country who use water from improved drinking water sources. Unicef defines improved drinking water sources as sources whose constr uction prevents contamination from outside sources, particularly fecal matter. 10 Most of the data points were from 2011. However, in t he few countries where there were no data available for 2011, the percentages from 2000 were used. These data were obtained We predict that as the percent access to improved drinking water sources increases, the infant mortality rate will fall (negative coefficient). The rationale behind this hypothesis is that water from cleaner sources should contain fewer disease causing pathogens and fewer toxic chemicals, which should decrease the risk of health problems and infant death As hypothesized, this variable ha d a negative coefficient ( 0.32 2 ). It was also significant at the one percent l evel with a t statistic of 3.38 3 and a p value of 4.635 x 10^ 4 Of these variables, this had the third largest impact ( 4.971 ). Therefore, we can say that access to improved drinking water sources does significantly affect infant mortality. 10 "Progress on Drinking Water and Sanitation (2013 Update)." Unicef. Unicef and World Health Organization, 2013. Web. 15 Oct. 2013.
Page | 8 IX. Independent Variable #4 : Physician Density (number of physicians per 1000 people) The next independent variable is the physician density, which is the number of doctors per one thousand people in a country. These fi gures include both general and specialized physicians. Wi th most figures coming from 2010 2012, these data were taken from the World Bank. We predict that as the density increases, then infant mortality rate should fall (negative coefficient). The rationa le behind this hypothesis is tha t when more doctors are available, families can more easily make appointments, receive immunizations, and have a doctor deliver their baby. All of these benefits should lead to better infant health. In the regression, this v ariable was not significant. This may be because this variable captures neither the availability of other vital medical resources, such as hospitals, immunizations, and medications, nor the prices of medical resources. If a country has a costly healthcare system, the infant mortality rate may be high if poorer families cannot afford medical care, regardless of how high the physician density is There may also be a disparity between physician density in urban and rural areas. This variable measures average d ensity throughout an entire country, which means that this does not specifically pick up on shortages in rural areas. The rural physician density rates could not be used instead because they were only available for a small number of countries in this sample. X. Independent Variable #5: Household Solid Fuel Use (% )
Page | 9 Our next variable is the percent of households that use solid fuels as their primary energy source for cooking and heating purposes. These data are in percentage form, are from 2010, and come fr 11 According to the World Health Organization, indoor solid fuel use can lead to acute lower re spiratory infections (ALRI), chronic obstructive pulmonary disease (COPD), and lung cancer because they produce a large amount of carbon monoxide, carbon dioxide, and smoke. 12 Unfortunately, since solid fuels are cheap and readily available, rural or impove ris hed families may not be able to afford or access safer fuels. Because of these health risks, we predict that as the use of solid fuels increases, the infant mortality rate will increase too (positive coefficient). As predicted, this variable h ad a posi tive coefficient (0.18 0 ). Additionally, it was also significant at the one percent level. The t statistic was 3.66 3 and the p value was 1.758 x 10^ 4. The impact was 6.29 3 which was the second highest. Clearly, household solid fuel use exacerbates the infant mortality problem Therefore, substituting away from solid fuels and towards safer fuels such as natural gas, may help reduce infant mortality rates. XI. Independent Variable #6: Prevalence of HIV/AIDS (%) 11 "Population Using Solid Fuels, Percentage." UN Stats United Nations, 11 July 2012. Web. 9 Oct. 2013. < http://unstats.un.org/unsd/mdg/SeriesDetail.aspx?srid=712>. 12 "Indoor Air Pollution." WHO Household Energy Database World Health Organization, 2010. Web. 12 Oct. 2013.
Page | 10 The sixth independent variable is the adult pr evalence of the human immunodeficiency virus and acquired immunodeficiency syndrome, otherwise known as HIV and AIDS. Obtained from the United Nations AIDS Report, these data are from 2011. The adult prevalen ce of HIV/AIDS is the percent of individuals be tween the ages of fifteen and forty nine who are living with HIV/AIDS. According to the World Health Organization, around 1.7 million people died from HIV/AIDS in 2011 This is particularly of concern in sub Saharan African, Southeast Asian, and Caribbean countries. Fortunately, the number of new cases has fallen substantially over the past decade. 13 HIV /AIDS de stroys the immune system, can be passed between parents, can be passed from a mother to her baby, and is often terminal. According to the U.S. Department of Health & Human Services infants can get HIV during pregnancy, during vaginal childbirth, and through breastfeeding. There is a twenty five percent chance that an infant will contract HIV during pregnancy, labor, and delivery from an HIV positive mother. 14 HIV severely weakens the immune system, which can make indivi duals more susceptible to infections and other diseases. HIV can also develop into AIDS, which is terminal. With a higher prevalence, there is an increased likelihood of infants contracting and suffering from HIV/AIDS. Therefore, we predict that this variable will have a positive correlation with the IMR. As hypothesized, the coefficient was p ositive and had a value of 0.73 4 It was also significant at the one percent level with a p value of 1.363 X 10^ 3 and a t statistic of 13 "Number of Deaths Due to HIV/AIDS." WHO World Health Organization, 2013. Web. 20 Oct. 2013.
Page | 11 3.05 1 The i mpact was 3.008 which was only the fifth highest of the six significant independent variables. Still, given its significance at the one percent level, it is evident that the prevalence of HIV/AI DS does affect infant mortality. XII. Independent Variable #7: Prevalence of Malaria (0 1) The next independent variable is the preva lence of malaria. This value was calculated by taking a count number of reported ma laria cases in 2011 and dividing by its population. This value (from 0 to 1) represents the proportion of the population that contracted malaria during 201 1 This data was obtained from the Kaiser 15 Many count ries have a high risk of malaria because of the high taxes on insecticides and mosquito nets. Unfortunately, 660,000 people, who are mostly children under the age of 5, die each year of malaria. 16 For the most part, these deaths occur in sub Saharan African countries. According to the World Health Organization, malaria can be cured if an antimalarial medication is given twenty four hours prior to the onset of a fever. However, not all citizens have knowledge of or access to this medication, and it can be dif ficult to diagnose malaria early on Malaria is caused by the Plasmodium parasite which is spread by mosquito bites. Symptoms (vomiting, fever, headaches) occur typically from ten to fifteen days after exposure and easily spread to other people. Left untreated, it can become fatal very quickly. 17 From the perspective of a highl y developed country, the six dollar cost for antimalarial treatment appears to 15 "Fact Sheet: Malaria, A Global Crisis." UNICEF Unicef, 27 Aug. 2004. Web. 21 Oct. 2013.
Page | 12 be quite affordable. However, many families in less developed countries live on less than two dollars per day (the international pove rty line), which makes receiving treatment i mpossible for them. 18 It is possible that not all cases are reported especially in less developed countries For financial reasons, families may not be able to get medical help, which means that cases may go unreported. However, the other measure of malar ia prevalence, the number of malaria caused deaths, has a smaller margin of error but could not be used because it includes malaria related infant deaths which are counted in the dependent variable. Although this may not be a perfect measure, the number of reported cases is the best measure of malaria prevalence that we can use. According to Unicef, women who contract malaria during pregnancy have a greater risk of having a baby with a low birth weight, which can increase the risk of infant illness and mortality. Malaria is also very contagious and can become lethal very quickly if symptoms go undiagnosed or untreated. Therefore we predict that a higher prevalence of malaria will lead to a higher IMR (positive coefficient). As hypothesized, this variab le had a positive coefficient (37.07 2 ). Additionally, it was significant at the 0.01 percent level. The t statistic was 2.50 0 and the p value w as 6.784 X 10^ 3 With a value of 2 .706 the impact for this variable was the lowest of all of the significant v ariables. By comparing the values for impact, we know that the prevalence of malaria has the smallest effect on the IMR out of all of these 18 Cohen, Jessica. "Prices, Diagnostic Tests a nd the Demand for Malaria Treatment: Evidence from a Randomized Trial." CGDEV Center for Global Development, 29 July 2010. Web. 20 Oct. 2013.
Page | 13 independent variables. However, given its significance, it is clear that in order for infant mortality rates to fall malaria must be combated too. XIII. Democracy Index (0 100) The final variable of this regression is the democracy index of a country. These values were computed by The Economist Intelligence Unit for the year 2012. This index is measured on a scale of zero to ten, where ten represents a perfect democracy and zero represe nts an authoritarian regime This index includes information about a articipation, political culture and civil liberties. Countries receive points for meeting certain criteria in each of these categories. For example, one component is voter part icipation. A country receives 1 point for consistently having over 70% participation, 0.5 points for 50 70% participation, or a 0 for less than 50% participation. The total points are then converted into a democracy index for each country 19 Because personal freedom (which includes freedom for both men and women) is a component of t his index, a high value i ndicate s that women participate in political and social activities and can push for access to healthcare, schools, and other resources. With a democracy, people s hould be able to demand a sufficient healthcare system and adequate med ical facilities, and the government is held accountab le for the health and happiness of its citizens Therefore, we predict that a country with a higher democracy index will have a lower infant mortality rate (negative coefficient). 19 "D emocracy Index 2012: Democracy Is at a Standstill." EIA The Economist Intelligence Unit Limited, 2013. Web. 10 Oct. 2013.
Page | 14 As predicted, the coefficient for this variable was negative ( 1.94 5 ). Addit ionally, it was also significant at the one percent l evel. The p value was 1.315 x 10^ 4 and the t statistic was 3.7 4 4. The impact was 4.33 0 which was the fourth highest of the six independent variables. Therefore, we can say that the level of democracy of a country d oes have a significant effect on the IMR. XIV. Concluding Remarks With the exception of GDP per capita (PPP) and physician density, all of the independent variab les were significant at the one percent level. Additionally, all of the sign ificant variables had the predicted signs. The r squared was 0.842 which means that 84.2% of the variation in infant mortality rates can be explained by these variables. Because it has the highest impact, the female literacy rate has the greatest influenc e on the infant mortality r ate of these variables Therefore, if a country wishes to reduce its mortality rate, then it should focus on increasing the educational attainment and opportunity for women. Based on the results of this empirical analysis, buildi ng safer drinking water systems switching away from solid fuels, dispersing both HIV/AIDS and malaria education and medication, and adopting democratic policies are all ways through which a country may lower its infant mortality rate. With the results of this paper, I hope that we may understand the means through which we may create a healthier and happier world for all Acknowledgement: I would like to thank Professor Kenny, the economics undergraduate committee, and my ECO 4935 classmates for their invaluable suggestions during this research project.
Page | 15 XV. Data Table 1: Empirical Analysis Results Coefficients Standard Error t Stat (Excel Generated) P value Standard Deviation Impact (Coefficient X Std. Dev.) Expected Sign Actual Sign GDP per capita (PPP) 5.254E 05 7.178E 05 0.732 0.233 16914.478 N/A Negative N/A Female Literacy (%) 0.366 0.061 5.964 *** 9.417E 09 22.883 8.377 Negative Negative Improved Water Access (%) 0.322 0.095 3.383 *** 4.635E 04 15.437 4.971 Negative Negative Physician Density (per 1000) 1.195 0.939 1.272 0.103 1.502 N/A Negative N/A Household Solid Fuel Use (%) 0.180 0.049 3.663 *** 1.758E 04 34.939 6.293 Positive Positive Prevalence of HIV/AIDS (%) 0.734 0.241 3.051 *** 1.363E 03 4.099 3.008 Positive Positive Malaria Prevalence (0 1) 37.072 14.829 2.500 ** 6.784E 03 0.073 2.706 Positive Positive Democracy Index (0 10) 1.945 0.520 3.744 *** 1.315E 04 2.226 4.330 Negative Negative ***Significant at 0.01 Level, **Significant at 0.05 Level, *Significant at 0.10 Level
Page | 16 Table 2: Summary Statistics GDP per capita (PPP) Female Literacy Improved Water Access Physician Density Household Solid Fuel Use Prevalence of HIV/AIDS Prevalence of Malaria Democracy Index Infant Mortality Rate Mean 15230 80.129 86.787 1.623 35.347 1.656 2.626E 02 5.561 29.736 Standard Deviation 16914.478 22.883 15.437 1.502 34.939 4.099 0.073 2.226 26.640 Minimum 600 12.600 47 0.014 5 0.010 0 0.600 2.170 Maximum 103900 100 100 6.723 95 26 0.520 9.930 119.410 Table 3: Correlation Matrix GDP per capita (PPP) Female Literacy Improved Water Access Physician Density Household Solid Fuel Use Prevalence of HIV/AIDS Prevalence of Malaria Democracy Index GDP per capita (PPP) 1 Female Literacy 0.492 1 Improved Water Access 0.543 0.660 1 Physician Density 0.581 0.642 0.619 1 Household Solid Fuel Use 0.606 0.706 0.734 0.706 1 Prevalence of HIV/AIDS 0.180 0.077 0.226 0.314 0.264 1 Prevalence of Malaria 0.285 0.415 0.382 0.374 0.543 0.213 1 Democracy Index 0.511 0.459 0.569 0.464 0.487 0.033 0.191 1
Page | 17 References "10 Facts on Malaria." WHO World Health Organization, Mar. 2013. Web. 20 Oct. 2013.
Page | 18 Kravdal, Oystein. "Child Mortality in India: The Community Level Effect of Education." Population Studies 58.2 (2004): 177 92. JStor. Web. 24 Aug. 2013.