Education Level of Nations

Introduction

Arguably, the biggest benefit an individual and the society can gain is through education.

While per capita GDP tells how financially well of are the individuals in a nation, the Human Development Index (HDI) has been used to tell how much “developed” are the people.

United Nations Development Programme (UNDP) calculates the Human Development Index (HDI) based on four variables − life expectancy at birth, expected years of schooling, mean years of schooling, and per capita PPP gross national income. So HDI does cover some aspects of education.

However, in this paper, we are interested to understand the “Education of Nations” using a few more parameters than the ones used in HDI. Therefore, we will deep-dive into some of the parameters including the ones used in HDI, and try to draw general inferences from the same.

Section 1: Deep Dive into Various Parameters of Education

In this section, we will examine various other parameters of education and understand their association with Development (HDI) and Wealth (GDP). But before that let us examine the association between Development and Wealth itself.

As per capita GDP rises from low to medium, the HDI also rises. Once a GDP threshold is reached, the HDI also has reached its peak. There is a little flattening seen around $15k and a further flattening around $30k. At the $15k level, an HDI of 0.75 is achieved. So, $15k might be a desirable minimum level that a country could strive to achieve in Development terms.

Indicator: Mean years of schooling (years)

Mean years of schooling is one of the variables of HDI therefore there is bound to be an association between these. This happens to be the most correlated with HDI.

Most of Europe, North America, and Oceania are the best performers on this indicator, while some in Africa and a few in Asia are the worst performers.

As expected, there is a high degree of association between this indicator and HDI.

We can see this strong association across continents. However, in Europe where this indicator is on the higher side across countries, this association is not as strong.

The GDP per capita does seem to affect mean years of schooling and it tapers after achieving $50k.

In Africa the effect is strongest with a little rise in per cap GDP, mean years of schooling rises fastest. In Americas+Oceania, the effect is almost linear throughout. In Asia, the effect is very strong until GDP per cap of $25k after which the schooling tapers at around 10 years. The effect is not visible in Europe at all where this indicator is on the high side, mostly 10+, across countries.

Expected years of schooling female (years)

On a comparative scale, except for a few countries in Oceania, South America, and Europe, the expected years of schooling for a female is not very high.

Many countries in central parts of Africa and a few in Asia perform very low on this parameter. The rest of the countries are on an average level.

Being one of the variables in HDI, this indicator has a strong association with HDI.

This association is quite strong across the continents.

However, GDP per Capita does not seem to have a very strong effect on this indicator. There is some effect up to the GDP level of 25k achieving over 15 years of schooling and then tapering around this level.

In Africa, the effect is strongest. In Americas+Oceania, the effect is strong u to about $20k. In Asia, the effect is strong up to $25k. In Europe, the effect is far less.

Expected years of schooling male (years)

This indicator is quite like its female counterpart. Almost the same pattern is seen.

The association is like the female indicator. However, the max this indicator reaches for most countries is around 18 years as against 21 years for females.

The association across continents is strong.

Like the female indicator, the GDP per capita is quite flat.

There is some effect in Africa and Americas+Oceania, but much less in Asia and Europe.

Gross enrolment ratio, secondary (% of secondary school-age pop)

This indicator is calculated as all those in secondary schooling divided by the children of secondary age group (which makes this indicator go above 100% too).

Australia and a few European countries are high on this. Most countries are in the middle. While many countries in central parts of Africa and a few in Asia are on the low side.

There is a strong linear relationship of this indicator with HDI.

The relationship can be seen in varying degrees across continents.

GDP per cap does affect this indicator up to a level of $15k, where it becomes almost 100% and tapers off.

All continents show a similar pattern but with differing cut-off for plateauing of the curve. In Africa, the tapering happens at it a little lower level of GDP. In Europe, the effect is far less.

Pupil-teacher ratio, primary school (pupils per teacher)

Most countries have a good pupil-teacher ratio except for countries in Africa and the Indian Sub-continent.

There is a strong negative relationship of this indicator with HDI.

The negative relationship can be seen across all continents.

GDP per cap does affect pupil-teacher ratio up to a level of $15k to $25k after which the curve flattens.

In Africa that has the lowest level of GDP, the curve doesn’t flatten and the effect is the strongest. In Americas+Oceania and Asia, there is a flattening after $15k. In Europe, the effect is much weaker with the curve relatively flatter than in other continents.

Summary of Section 1

We can see that the additional indicators other than those used in the Human Development Index also show similar patterns. So the education aspect is largely covered in HDI.

GDP per capita between US$15,000 to US$25,000 at 2017 prices gives an inflection point for many indicators up to which GDP affects the indicator. This must be used by countries as a target to achieve.

Although Africa comes out with worse cases, it is not all of Africa and some Asian countries too show up in worse case scenarios. Therefore, the continents do not completely give the health picture. We need to cluster countries into different sets.

Section 2: Country Clusters

We intend to do country clusters at two levels. First, we form main clusters, and then within each, we will form sub-clusters. Since we desire to form a two-level hierarchy, we will use a hierarchical clustering algorithm. For linkage, we will use the “ward” method as it minimizes within-cluster variance which will help in forming more homogenous groups of countries.

Based on the indicators (scale standardized to give equal importance to each indicator), we have the following four clusters of countries. Central Africa and a few for Asia make the lowest cluster c3 (in red). Parts of the Indian Subcontinent, parts of South-East Asia, and some African and South American countries make the next cluster c2 (in yellow). Northern parts of LatAm, many from Asia, a few from Europe, a few from Africa make the next cluster c1 (in light blue). Northern America, Southern parts of South America, most of Europe, and Oceania make the highest cluster c0 (in green).

Countries of Cluster c0

There are 54 of 169 countries in this cluster. Out of these 34 of 38 European, 10 of 41 Americas+Oceania, and 10 of 42 Asian countries are in this cluster. No African country is in this cluster.

This splits into five sub-clusters.

Countries of Cluster c1

There are 31 of 169 countries in this cluster. Out of these 23 of 41 Americas+Oceania, 19 of 42 Asian, 6 of 48 African, and 4 of 38 European countries are in this cluster.

This splits into four sub-clusters.

Countries of Cluster c2

There are 32 of 169 countries in this cluster. Out of these 15 of 48 African 10 OF 42 Asian, and 7 of 41 Americas+Oceania are in this cluster. There are no European countries in this cluster.

This splits into four sub-clusters.

Countries of Cluster c3

There are 31 of 169 countries in this cluster. Out of these 27 of 48 African, 3 of 42 Asian, and 1 of 41 Americas+Oceania are in this cluster. There are no European countries in this cluster.

This splits into three sub-clusters.

Summary of Section 2

The following inferences we draw from the clustering analysis:

• Except for Africa which has many countries in the lowest cluster, continents do not define the level of education of countries.
• The four clusters represent the most educated (c0), less educated but not critical (c1), less educated a little critical (c2), and the least educated most critical (c3) nations.
• European countries never go into the lowest two clusters, and African countries never go into the highest cluster.

With these clusters, we can now analyze countries in these homogenous groups.

Section 3: Analysis of Indicator Patterns and Trends by Clusters

Having formed these clusters, we would now like to see how countries fare and compare within and across clusters. Since we have 169 countries, we will not look at all of these but take a subset. We have decided to take countries with a sizable population (in this case over 50 million) to be representative of their clusters. Thus, we have 32 countries to analyze in the section.

We will first look at patterns and trends with HDI and GDP and then further with each of the five indicators analyzed in section 1.

Indicator: Human Development Index

The pattern of HDI is very straight forward. As we go down the clusters the HDI steadily decreases.

Over the years, the clusters have witnessed an almost parallel progression. This means the gap between clusters on HDI has not been narrowing.

Cluster 1, and cluster 3 countries have been progressing in a narrow band of HDI, while there is a much larger band in cluster 0 and cluster 2.

Indicator: GDP per capita (2017 PPP $)

There is indeed heavy discrimination on GDP per capita. The highest cluster countries have a much higher level than the other three. However, the progressive decrease is still there.

Over the years the GDP gap appears to have increased. The highest cluster (c0) countries have had better growth, than the other three clusters. The bottom cluster (c3) has been on the verge of being stagnant.

Indicator: Mean years of schooling (years)

Mean years of schooling is a good discriminator of clusters, however, there are some overlaps.

Mean years of schooling indicator has been rising across clusters.

Almost all countries across clusters have been improving on this indicator.

Indicator: Expected years of schooling female (years)

Expected years of schooling for female discriminates the clusters well with minimal overlaps.

All clusters have been improving on this.

Most countries in cluster 0 had been historically high on this and a few that were not have dramatically improved. In cluster 3 some countries appear to be plateauing although these are still far lower than the other cluster countries.

Indicator: Expected years of schooling male (years)

Expected years of schooling for male discriminates the clusters well with little more overlaps than in females.

All clusters have been improving on this.

Like the female indicator, most countries in cluster 0 had been historically high on this, and a few that were not have dramatically improved. In cluster 3 some countries appear to be plateauing although these are still far lower than the other cluster countries.

Indicator: Gross enrolment ratio, secondary (% of secondary school-age pop)

Gross enrolment ratio, secondary discriminates the clusters well with a little overlap.

All clusters have shown varying degrees of improvement on this indicator.

Most cluster 0 countries were historically high on this so remained flat, but those that were not showed the most rapid improvement. Most cluster 1 and cluster 2 countries have been improving rapidly. Although there is an improvement in cluster 3 countries, the improvement is much slower causing the gap between cluster 3 and others to widen.

Indicator: Pupil-teacher ratio, primary school (pupils per teacher)

The pupil-teacher ratio discriminates clusters well but with some overlaps.

There has been an improvement in this indicator in all clusters except the worst cluster of c3.

Cluster 0 countries have improved and formed a narrow low band. Cluster 1 countries have improved a little slowly. Cluster 2 countries have improved a little more rapidly than cluster 1 countries. Cluster 3 countries have generally not improved, some have even become worse.

Summary of Section 3

The pattern and trend analysis has shown that all indicators discriminate the clusters quite well. We can also see that in some cases cluster 3 countries have not shown much improvement which should be a cause of worry.

Final Takeouts

In summary, we make the following points:

Per capita GDP on purchasing power parity between $15,000 to $25,000 at 2017 prices could be used by countries as a target.

Cluster 3 countries need special attention as these are far away from the above GDP target and have shown slow or lower improvement over the years.

The indicators that have not seen much improvement especially in cluster 3 countries should be targeted for improvement in countries that are weak on those indicators.

Credits

The source of data for this report is:

• UNDP Human Development Report at (http://hdr.undp.org/en/data). UNDP has shared their data under Creative Commons Attribution 3.0 IGO which allows free use if credit is given (as we have done here). [No changes to the data were made].

This report has been prepared by gralovis insights private limited, Mumbai — 400078, India.

Originally published at http://gralovis.com.