I-Introduction

The paper’s main objective is to analyze the relationship between GNP per Capita and Infant

Mortality rates. The primary goal is to explain infant mortality rates with income, GNP per Capita will be used as the proxy for income. The intuition that the paper is interested in shedding some

light on is to see if GNP per Capita is a good estimator of infant mortality. The paper hypothesizes that as GNP per capita increases, the infant mortality rate would decrease

II-Data

The data used in this paper is a sample from 1990 from 97 different countries from 6 different

regions, namely, Eastern Europe, Latin America, Western Europe, Middle East, Asia, and Africa.

The sample contains information on GNP per capita, birthrate per 1,000, death rate per 1,000, and infant mortality rate per 1,000. These variables are the main source of interest. The dataset also

presents information on the life expectancy for males and for females.

Table 1: Summary Statistics

Notes: Birth Rate, Death Rate, and Infant Mortality Rate are in 1,000 lives.

Birth

rate     Death

rate

Infant   mortality

Life

Expectancy

male

Life

Expectancy

Female

GNP/capita

III- Empirical Framework

The estimated model is below:

(1)     Infant MOTtalityi  = a + β ∗ GNP/capita i  + εi

Based on the estimation output, GNP/Capita is statistically very significant at any reasonable  significance level. As GNP/Capita increases by 1 dollar, infant mortality rate will decreases by .00344 per 1,000 lives. However, it is apparent that the R2 value is very low, meaning that

GNP/Capita is only explaining 36.2% of the variability in infant mortality rates across the 91

countries. The model should also include some other variables to capture more of the variability in   infant mortality rates. Possible control variables are the birthrate and life expectancy for males. The paper hypothesizes that as the birthrate increase infant mortality will also increase. Adding male life expectancy would be a general control for the general health situation of the country. The paper

hypothesizes that as the life expectancy increases the infant mortality rate would decrease. The paper will use robust standard errors for the multivariate regressions due to heteroskedasticity at the 5%

significance level (White testis used to determine heteroskedasticity).

(2)  Infant MOTtality i    =  a + β ∗ capit(GNP)ai  + Y ∗ BiTtℎ Rate i  + ε i

Adding birthrate in the model as a control decreases the effect of GNP/Capita, GNP/Capita is still significant but only at the 5% significance level. The results also show that birthrate has a positive    effect on infant mortality, as the birthrate increases by 1 per 1,000, infant mortality increases by

2.67, holding GNP/Capita constant. It is also apparent that the birthrate variable is adding extra   explanation to the variability of infant mortality. This is captured by a very large increase in the R2. This multivariate model (model 2) explains 74% of the variability in infant mortality.

(3) Ln (Infant Mortality) i    =  a + β ∗ capit(GNP)ai  + Y ∗ Ln(Birtℎ Rate)i+ ε i

(4)  Ln (Infant Mortality) i   =  a + β ∗ capit(GNP)ai  + Y ∗ Ln(Birtℎ Rate)i+ δ ∗ Male Life Expectancy i+ εi

The paper extended the initial model (model 1) by using a different functional form, namely,

estimating the log of infant mortality. Other functional forms could be tested as well, such as the quadratic form for GNP/Capita (presented in model 5 and 6).

IV-Results

Table 3: Estimation Results for Different Models

Notes: Model 2 uses robust standard errors

Missing observations were dropped by STATA

Standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

In the first model use, GNP/Capita is statistically very significant at the 1% significance level. An  increase in GDP/Capita would result in a decline of .0034 in infant mortality. The model explains 36.2% of the variability in infant mortality. The second model explains 74% of the variability in

infant mortality with the linear relationship with GDP /Capita and birthrate. GDP/Capita is

significant at the 5% significance level, which means that once birthrate is included in the regression GNP/Capita adds extra explanation to the variability of the infant mortality rate. An increase of 1

dollar in GDP/Capita would decrease the infant mortality rate by .000595, holding the birthrate constant. Adding birthrate adds extra explanation to the variability of infant mortality once

GNP/Capita is already in the model; birthrate is significant at the 1% significance level. An increase in birthrate by 1 per 1,000 would increase the infant mortality rate by 2.67, holding GNP/Capita

constant. The second model explains the variability of infant mortality better compared to the simple model.

Model 3 uses the log functional form. It is the better fit model because the log (natural) adjusted R2     is the largest within the six models tried. Model 3 explains 86.7% of the variability in infant mortality

with the linear relationship with log (natural) birthrate and GNP/Capita. GNP/Capita is significant at the 1% significance level, which means that once the log of the birthrate is included in the

regression, GNP/Capita adds extra explanation to the variability of the natural log of the infant

mortality rate. An increase in GNP/Capita by&nb