Crude Birth Rate in Relation to Economic Factors

Malcolm Cellard-Farrall, Garret Hokr

ECN 410

1. Introduction

This project will clearly explain the significance between various forms of economic factors and their effects on a nation’s crude birth rate. Crude birth rate is measured as the number of successful births per 1,000 people in a nation and is an accurate way to measure a nation’s population growth. The concerning questions we hope to answer are, “How do these factors affect crude birth rate? And if they do, is it better to increase or decrease these factors?” To find the answers, we will analyze the economic factors that influence this statistic and find how much they impact crude birth rate. The factors used in our test will include unemployment, GDP per capita, national inflation rate, national real GDP growth, education (measured with literacy), and net trade. Because a low literacy rate can impede economic growth, we are considering education as an economic factor.

The purpose of this project is to find information that could guide policy makers around the world when they are making decisions concerning either the promotion or discouragement of certain factors that can affect crude birth rate, especially countries that have birth rates that are relatively high or relatively low.

 II.    Data Used

The data provided for analysis has been gathered by indexmundi from over 130 countries around the world from the regions of North America, Central America, Caribbean, South America, Asia, and Europe. Africa was left out due to countries having missing data which skew the results of our regressions. Another reason African countries were left out was because many African governments manipulate data; just try and find Zimbabwe’s unemployment rate if you don’t believe us. This would result in unreliable data and is the reason why we chose to not include countries in Africa.

Since our main concern is how different variables affect birth rate, Crude Birth Rate is our dependent variable. The independent variables chosen for this project were selected because they are universally recognized as common factors in a nation’s economy with the exception of education. As stated previously, we are considering education as an economic factor. The data provided for the regression has been used only for statistical analysis and is free of any bias from either member in our group.

III.    Methods

After gathering our numbers, we imputed the data into the Stata columns. By doing this in Stata, we have created a controlled environment where crude birth rate can only be affected by the independent variables of unemployment, GDP per capita, inflation rate, GDP real growth, and net trade. With this information we ran the following regression:

CrudeBirthRate = b0 + Unemployment b1 + GDPperCapita b2 + InflationRate b3 + GDPrealGrowth b4 + NetTrade b5  + e

After running our first regression, it was clear that one more independent variable was needed to complete the project. It was then suggested that we add an additional variable related to education, which we did to the end of original regression line. Once that was done, we needed to test the assumption of independence to check if our independent variables were correlated to the error term. In simpler words, check if our independent variables that affect crude birth rate included in our regression correlate to independent variables that affect crude birth rate that were left out of the regression. To do this, we simply gathered information to create a new independent variable called GDPppp, which is a nation’s purchasing power parity, for each country. We then ran a regression with our new variable:

CrudeBirthRate = b0 + Unemployment b1 + GDPperCapita b2 + InflationRate b3 + GDPrealGrowth b4 + Education b5 + NetTrade b6 + GDPppp b7 + e

In order to check for independence, we compared the regressions to each other by storing their estimates (i.e. typing in estimates store reg1 after a regression). This is a common practice known as Forward Selection. Once stored, we looked at the estimates table for both regressions (estimates table reg1 reg2). What we found was that two of our variables, unemployment and net trade, had a drastic coefficient change when we added GDPppp. How we fixed these variables will be explained in a later step of the methods section. Fortunately, GDPperCapita, InflationRate, and GDPrealGrowth were about the same when we introduced GDPppp to the regression, meaning they are “robust”, which a sign of accurate measurement. This proves that these independent variables are correlated to our dependent variable of crude birth rate.

The next step was checking for multicollinearity. To check for this, we used the variance inflation factor test, or VIF for short. This runs regressions for all independent variables to compare their r-squared to check the variables are too strongly correlated  to each other. All that is needed is a simple command in stata written as estat vif. All of our independent variables were shown to be in the acceptable threshold for the VIF test meaning multicollinearity is not an issue. We will go in more detail on the actual results of this test and how to interpret them in the ‘Results’ section of the project.

The final test we ran in stata is called the hettest. This test can be used to check for heteroskedasticity or equal variance. The results of this test show that the null hypothesis of constant variance is present (HO: constant variance). This means that by accepting this hypothesis, our assumption of equal variance is not violated (our Chi2=0.00) . The results also show that there is indeed heteroskedasticity present in the data. To rectify this, we used robust standard errors. This is done by adding a command to our original regression, vce(robust), to create our last regression:

CrudeBirthRate = b0 + Unemployment b1 + GDPperCapita b2 + InflationRate b3 + GDPrealGrowth b4 + Education b5 + NetTrade b6  + e

IV.    Regression Results & Interpretations

Listed above is a table summarizing the data collected for the six independent variables we used as well as our dependent variable, crude birth rate. Certain values, like the minimum value for unemployment rate, are listed as low as .0001, which were rounded down to 0%. This data, which seems almost too good to be true, is the rate given by that nation’s government officials. Be wary that some of the data may not accurately represent the nation as a whole, but for the purpose of our regression, we used the data that was given. Crude birth rate is, again, the amount of successful births per 1,000 people in a population. With the exceptions of net trade and GDP per capita, all other values should be interpreted as percentages, and later, our coefficients will further explain this.

After running our regression for the first time, we noticed that two variables, those for unemployment and net trade, had significantly high p values. Both variables had p values well over .05, so we had to then test for multicollinearity. We would not want this multicollinearity in our model because it could possibly increase the variance of our coefficients, making them less accurate. This is when we decided to run the test on our regression to check and see how closely correlated our variables were to each other. We were able to do this by performing the Variance Inflator Factor (VIF) test, as mentioned earlier. If our variables had VIF values greater than 10, we would know that there is a serious problem of multicollinearity. Ideally, a VIF should be around 1 to show no multicollinearity, and thankfully, all of our variables’ VIFs are less than 1.4. The fact that our mean VIF is at a confident 1.16 shows that multicollinearity was not a problem in neither our data nor our regression model.

After running our regression, we ran the hettest and, as stated earlier, found that our regression almost definitely had heteroskedasticity within the data. To fix this problem, we re-ran our regression model. This time we ran it so it included the robust standard errors in our model to offset the heteroskedasticity. All the variables besides unemployment and net trade have p values less than .01, and the two that were over this threshold (unemployment and net trade) could still be included because of their extremely low VIFs. After looking at our results for our variable coefficients, we decided to multiply the GDPperCapita coefficient by 1,000, so that every $1,000 increase in GDP/Capita would result in a .0619 crude birth rate decrease, which can be seen in the chart above. A similar approach was done with all of the variables that are represented by a rate. This means that the variable coefficients for education, unemployment, GDP Growth, and inflation were all multiplied by 100. This way, for every one percent increase in any of these rates, the new coefficients (which is the coefficient in the regression but whose decimal point is moved two places) would affect the crude birth rate. Again, our sample size could have been more inclusive, but we found that our 131 country observations made sense and worked with the model. It would have just been too difficult to accurately determine certain variables in nations with little-to-no data. Doing this would impact our data set with values of “0” that we’d have to put in place for them in Stata; and it would be unreasonable for any of our independent variables to have a value of zero.

V.    Conclusion 

After running and analyzing our regression with the aforementioned variables, we believe that it is safe to say that the economic and demographic factors we have chosen have a direct effect on a nation’s crude birth rate. For example, many economists point out that less developed (or up-and-coming) countries are having more children, and that more developed countries are having less. Our regression model backs up this statement perfectly, which can be seen with the coefficient for GDP per capita and GDP growth rate. Many of the richer nations around the world have lower birth rates, which is perfectly represented by our negative coefficient for GDP per capita. As poorer nations begin to grow, they do so at a much quicker rate than more established nations. For example, countries like Mongolia, which have a GDP per capita of roughly $4,000, are growing at a rate of 12% while richer countries like the US, which have a GDP per capita of approximately $53,000, are only growing at a rate of 2%. Again, this explains why the coefficient for GDP growth rate is positive. Overall, one could look at most of our variables and their respective coefficients and agree that they are statistically significant and logically make sense. Even outliers, like net trade or unemployment, have low and like-enough VIFs that prove their significance in the model.

In conclusion, one could use our regression model and make a fair guess on what their country’s crude birth rate would look like. Country leaders could use the figures they collect on GDP, unemployment, etc, to better predict what their birth rate would be like in the upcoming year and then decide whether or not they are happy with that figure. Ultimately, our model could be used to find ways to control the birth rate with the variables we tested for, and that makes their job one of just plugging in numbers to our regression model.

VI.    Works Cited 

http://www.indexmundi.com/map/?t=0&v=25&r=xx&l=en

http://www.indexmundi.com/map/?t=0&v=74&r=xx&l=en

http://www.indexmundi.com/map/?t=0&v=67&r=xx&l=en

http://www.indexmundi.com/map/?t=0&v=71&r=xx&l=en

http://www.indexmundi.com/map/?t=0&v=66&r=xx&l=en

http://www.indexmundi.com/map/?t=0&v=39&r=xx&l=en

http://www.indexmundi.com/map/?t=0&v=85&r=xx&l=en

http://www.indexmundi.com/map/?t=0&v=89&r=xx&l=en

http://www.indexmundi.com/map/?t=0&v=67&r=xx&l=en

Images

http://s3.amazonaws.com/hiphopdx-production/2014/09/readytodie.jpg

https://stocksaints.com/sites/default/files/styles/article_image_full_node/public/field/image/iMoney.jpg?itok=AOdOjDeI