The Effects of Different Education Levels on GDP Per Capita


According to economic theory, GDP per capita of a country or a state can be increased in many ways. Education here plays a very vital role in the increase of GDP per capita. In general, the higher the education level, the higher the productivity and the higher the income. Affects the scale of the magnified state. The following applies: The higher the education level of a country, the better the economic outlook for that country. Education is an important variable, but other factors such as gender discrimination, quality of education, religion, and even the area in which the person lives play an important role in a person’s income. Therefore, further interpretation is needed. It can be expressed in many scales, including: number of people with a bachelor’s degree or higher / population ratio, enrollment rate in secondary education, etc.

Our dissertation is based on per capita income at various levels of education, as well as the other factors mentioned above. If we look closely, bachelor’s and master’s degrees, including higher education, have an impact on our income. In this paper we mainly focused on how GDP per capita in all the states of USA is affected by higher level of education.

In the United States 79 percent adults are educated in the year 2022. Following that 21 percent adult population of United States are illiterate in the same year. There are 54 percent adult whose level of education is below sixth grade. US have to bear the cost of low-level literacy up to 2.2 trillion dollars every fiscal year.

Literature review

For economists and politicians’ educational economy impact has been a controversial topic. In the configuration of this paper, various journals and publications are reviewed as a reference.

Among them one that explores the economic impact of universities that was published by Valerio and Van Reenen (2018). Authors of this paper pointed out the hyped trend of an increasing number of universities and the increasing number of universities per million people and mean growth in GDP per capita. A fluctuating plot that plots the number of universities per million people and the mean growth of GDP per capita income and is utilized to represent the trend. They represented the hypothesized relationship that has positive and strong correlation between GDPs per capita and universities Further results suggested that if in a region there is a 10% increase in the number of universities then GDP per person increases by 0.4 percent.

Mamit Deme and Ali M. A Mahmud (2020) stated that statistically primary and secondary education positively impact per capita real GDP. But the correlation between economic growth and education is not robust and weak. They conclude to a result that policy makers aim both the quality and quantity of education to achieve growth in real GDP per capita and primaryeducation for everyone needs to be ensured.

Another study that contributes to this field of study is Aghionet. Al. (2009) research is written by Aghion et. (“The Causal Impact of Education on Economic Growth”, 2009) This paper focuses on the impact of all education sectors, not universities. Introduced effects such as the transition to a model of skilled workers built a complex model explaining the impact of investment in education on GDP and introduced impacts such as the movement of skilled workers into the model.

Interesting insights into Odit et. al. (2010) The driving force of economic growth is what they see as an educational qualification that contributes to the quality of human capital. Harvest on scale. They used a constant CobbDouglas production function. Here, the human capital augmentation growth model uses human capital as an independent element of production. In conclusion, the results of the calculations suggest that education is productive and convincing, not “a way for individuals  to inform their employers of their skill level.”



The purpose of our paper is to detect how real GDP per capita of the states of USA  is associated with its level of education which is measured with population percentage that hold a bachelor’s degree or more than that. In our paper data we used here for GDP per capita is calculated in chained 2012 dollars. The advantage it has is that it is inflation adjusted. On the other hand, for the level of education the measure is population percentage of states who are 25 years old or older and who has bachelor’s degree or more than that. 2019 is the year both of the data sets are obtained from. Both data are taken from BEA (U.S. Bureau of Economic Analysis). There are some other variables that affects GDP per capita of sates across the country. These variables are total labour force, urban population, unemployment, and labour participation rate. Total number of sates of United states with Washington D.C is the sample size of the data set which is 51. The data of unemployment rate is taken from FRED. Urban population is taken from U.S Census Bureau. Total labour force and labour participation rate both are taken from U.S Bureau of labour statistics.


Variables Observations Mean Standard Deviation
gdpp 51 56789.9 20163
educ 51 32.6 6.6
urbanp 51 73.8 15.1
un 51 3.6 0.8
labf 51 3209000 3590838
labpar 51 63.8 3.9


The model in this paper satisfies the assumptions of Gauss Markov. They are:

1) Linear in Parameters:

For a linear regression model, y= 0+ 1x1+ 2x2+………… kxk+  is the basic form. Here, loggdpp is the dependent variable which is represented by y. Then educ is represented by x. However, 0 here is the intercept. And the coefficients are from 0 to k.  here is the error term.

2) Random Sampling:

51 is the sample size of this data which includes all the states of United States of America. Both low and high GDP per capita of the states and also low and high education level degree rates are included in this data set. And the available data is sampled.

3) No Perfect Collinearity:

When no perfect linear or connection or relationship exists among the independent variables the theory or assumption of no perfect collinearity is satisfied. To test no perfect collinearity R and R studio is used.

loggdpp educ un urbanp labpar loglabf
loggdpp 1.0000
educ 0.7612 1.0000
un 0.1408 -0.1436 1.0000
urbanp 0.5324 0.4997 0.1557 1.0000
labpar 0.6015 0.6321 -0.3835 0.2613 1.0000
loglabf -0.0544 0.0469 0.0588 0.4299 -0.1631 1.0000

4) Zero Conditional Mean:

Zero conditional mean asses that suppose all the values of independent variables are given in that case expected value of  which is the error term will be zero. The residual plot can be used to test zero conditional mean.

5) Homoskedasticity:

The assumption of homoskedasticity basically is that the variances of different samples are indifferent. So, the variance of  will also be the same for all.



Model 1:

In the first model a simple linear regression model is conducted. The hypothesized model is:

log(gdpp)= β0+ 1log(educ)+

The estimated equation is:

log(gdpp) = 4.298 + 0.0133educ

Number of observations 51
F (1, 49) 67.49
Prob > F 0.00
R squared 0.5801
Adjusted R squared 0.5709
Root MSE .07175



Source SS df MS
Model .347606012 1 .347606012
Residual .252285384 49 .005148681
Total .599891396 50 .011997828


loggdpp Coefficient Standard error t value P > |t| [95% Conf. Interval]
educ .013254 .0015381 8.22 0.00 .0095469 .0157286
_cons 4.297786 .051176 84.52 0.00 4.222752 4.428436


The adjusted R-squared value of this simple regression model is 0.5801. This means that the association that exist among the two-variable log(gdpp) and log(educ) is moderate. In the table we can see that the coefficient of education is positive. It means that log GDP per capita increases when there is an increase in education. But this hypothesized relationship is not enough or adequate to conclude.

Model 2:

The model 2 is multi regression model which includes all the controlled variables that were omitted in the first simple regression model. The form of the equation is

log(𝑔𝑑𝑝𝑝) = 𝛽0 + 𝛽1𝑒𝑑𝑢𝑐 + 𝛽2𝑢𝑛 + 𝛽3𝑢𝑟𝑏𝑎𝑛𝑝 + 𝛽4𝑙𝑎𝑏𝑝𝑎𝑟 – 𝛽5log (𝑙𝑎𝑏𝑓)+

And the equation that is estimated is:

log(𝑔𝑑𝑝𝑝) = 3.7914 + 0.0081𝑒𝑑𝑢𝑐 + 0.0517𝑢𝑛 + 0.0014𝑢𝑟𝑏𝑎𝑛𝑝 + 0.0094𝑙𝑎𝑏𝑝𝑎− 0.0412log (𝑙𝑎𝑏𝑓)

Number of obs 51
F (5, 45) 25.7
Probability > F 0.00
R squared 0.7398
Adjusted R squared 0.7118
Root MSE .0588


Source SS Df MS
Model .444307301 5 .08886146
Residual .155584095 45 .003457424
Total .599891396 50 .011997828


loggdpp Coefficient Standard error t value P > |t| [95% Conf. Interval]
educ .0081546 .0018195 4.59 0.000 .0046845 .012014
un .051692 .0117609 3.63 0.001 .0190303 .0664058
urbanp .0014537 .000757 1.92 0.061 -.000071 .0029784
labpar .0093673 .0031247 3.00 0.004 .0030738 .0156607
loglabf -.0411538 .022121 -1.43 0.160 -.0761332 .0129749
_cons 3.79138 .2552162 14.92 0.000 3.292792 4.320856


This model adds more variables that has connection with both variables GDP per capita and education. After adding more variables, the value of R squared has increased to 0.7398. This increase is significant. This is an indication that new model is much better in representing the connection than previous model. Except log(labf) all the coefficients found in this regression are positive.

Model 3:

log(𝑔𝑑𝑝𝑝) = 𝛽0 + 𝛽1𝑒𝑑𝑢𝑐 + 𝛽2𝑢𝑛 + 𝛽4𝑙𝑎𝑏𝑝𝑎𝑟+

Number of observations 51
F (3, 47) 39.81
Prob > F 0.00
R squared 0.7201
Adj R-squared 0.6996
Root MSE .06004


Source SS Df MS
Model .430481883 3 .143493961
Residual .169409513 47 .003604458
Total .599891396 50 .011997828


loggdpp Coefficient Standard Error T value P > |t| [95% Conf. Interval]
educ .0095253 .0016768 5.59 0.000 .006152 .0128986
un .0497983 .0114319 4.36 0.000 .0268004 .0727963
labpar .0107416 .0030555 3.52 0.001 .0045946 .0168885
_cons 3.564069 .186281 19.13 0.000 3.18932 3.938818

The estimated

This model removes the two variables urbanp and log(labf). These two variables are relatively insignificant comparing to the other variables. This is the reason these two variables are omitted in the third new model. The R square value of this model is also much lower. The value is 0.7201 Heteroskedasticity is not detected in the data set used for this paper.



This paper studied on the fact what relationship GDP per capita has with level of education. And the paper detects a positive connection between the two variables which means that a state which has more population of bachelor’s degree or higher than that has higher GDP per capita. The coefficient of the variable education here is positive. Other factors or variables that effect the GDP per capita of a state are countless and very hard to mode.

In conclusion, the impact of education on the economy is positive, whether in terms of innovation and research and development, or labour productivity. It suggests that a person’s standard of living depends on how higher education affects overall economic outcomes. This paper aims to represent this existing correlation among GDP per capita and higher levels of education.



1) The Causal Impact of Education on Economic Growth. (2009). Retrieved 8 May 2022, from

2) Odit, M. P., Dookhan, K., & Fauzel, S. (2010). ‘The impact of education on economic growth: The case of mauritius. International Business & Economics Research Journal’

3) Mamit Deme and Ali M. A Mahmud (27 July, 2020) ‘Effect of quantity and quality of education on per capita real-GDP growth: evidence from low- and middle-income African countries’,is%20weak%20and%20not%20robust.

4) Elizabeth Appiah ‘The Effect of Education Expenditure on Per Capita GDP in Developing Countries’

5) Łukasz Goczek and Ewa Witkowska ‘How Does Education Quality Affect Economic Growth?’

6) Laura Marquez-Ramos ‘Education and economic growth: an empirical analysis of nonlinearities’

Note: This article was submitted to Professor Farzana Munshi of Brac University of Bangladesh as a part of the writer’s course study.

Leave a Comment