Where Does Teaching English Actually Make A Difference?
Foreign exchange programs that enable new or recent graduate programs to teach English abroad are a big business. By one count (admittedly by a biased source), over 100,000 positions for international English teachers open every year. EF Education First, which is a large company specializing in international exchange, including foreign English education, reportedly has a presence in over 5o countries and approximately 40,000 employees. Publicly reported revenue numbers do not appear to be available, but its founder Bertil Hult is estimated to have a net worth of $5 billion.
Given such high stakes, it makes sense to critically examine the effect of these programs. As designed, are they an effective tool for improving the livelihoods and outcomes of disadvantaged foreign students, or are they simply escapist vacations that actually increase global inequality?
Methodology
To identify those countries in which foreign English education may serve the sorest need, I compare the the English proficient population of a country to its per capita GDP on the basis of Purchasing Power Parity (GDP-PPP). This experimental design, therefore explores the correlation between the need for English education against the real, on-the-ground improvement inoutcomes as a result. Using GDP-PPP is an important aspect of this experiment because it normalizes the differences in purchasing power across different economic realities.
For this study, I find the CIA estimated GDP-PPP for those countries for whom data is available on English speaking population. I encourage discussion on the verity of the latter data source, specifically the proportion of the Chinese population who are proficient in English; while not optimal I could find none other that were as comprehensive.
Does English Actually Matter?
Before we can determine which countries stand to benefit the most from foreign English education, it is important to determine whether English proficiency is correlated with superior economic outcomes.
This graph compares the GDP-PPP of various countries against the proportion of the population that is proficient in English. The dark blue line is a line of best fit calculated using simple linear regression. The yellow dashed lines display the median values for each dimension.
To answer the question of whether English actually matters, we review the trend of the the line of best fit. As it has a positive slope, it is safe to say that the countries with a higher proportion speaking English tend to have a higher GDP-PPP than ones with a lower proportion. There are a number of reasons why this might be the case, but given a decision-making calculus using these parameters, it is sufficient to proceed at this time.
Where does English matter the most?
Now that we know that English proficiency does appear to make a difference in economic outcomes in terms of GDP-PPP, we can start to identify those countries that are best served by efforts to increase English proficiency. There are two ways that we can identify these countries using an econometric framework:
- By calculating the residual, or difference between the expected value output from the linear regression and actual value, we identify those countries that have the greatest deviation from where we expect them to perform given their level of English proficiency
- By calculating the Euclidean distance between a country’s actual proficiency and GDP-PPP and the median values for the population, we identify those countries that have the greatest deviation from the norm of the world population
It so happens the coefficient of determination (r-squared) for the model is rather low (0.12, or 12% of the variation is explained by the relationship described by the line of regression) because there are significant contradictory outliers. In other words, there are a number of countries that have a high English proficiency but still have a low GDP-PPP. They still need some kind of help, but English education is definitely not it.
This being the case, the preferred approach will be the latter. Using the Pythagorean theorem and feature scaling normalization, we can calculate the Euclidean distance between a countries actual location and the convergence of the two median lines. In this case, the median country in this model has a GDP-PP of $16,200 and an English proficiency of 42.31%. This method also allows us to subset the population to look at those that are of interest to us – namely those countries with a low GDP-PP and a low rate of English proficiency. The greater the distance between that country’s location and the median convergence, the greater the effect that higher English proficiency should, in theory, have on its GDP-PP.
Countries with sub-median GDP and English proficiency, sorted by distance to median
Country |
English % | GDP-PPP | Distance to Median Convergence |
Ethiopia | 0.22% | 1500 | 0.4510 |
Honduras | 0.44% | 4700 | 0.4377 |
The Gambia | 2.34% | 1700 | 0.4305 |
Tokelau | 2.86% | 1000 | 0.4285 |
Dominican Republic | 0.15% | 12800 | 0.4238 |
Nepal | 3.00% | 2400 | 0.4215 |
Malawi | 3.88% | 800 | 0.4200 |
China | 0.83% | 12900 | 0.4169 |
Swaziland | 4.38% | 7800 | 0.3908 |
Uganda | 8.09% | 1800 | 0.3770 |
Brazil | 4.94% | 15200 | 0.3744 |
Tanzania | 9.89% | 1900 | 0.3603 |
Yemen | 9.00% | 3900 | 0.3596 |
Algeria | 7.00% | 14300 | 0.3542 |
Cambodia | 10.53% | 3300 | 0.3481 |
Sri Lanka | 9.90% | 10400 | 0.3307 |
Bhutan | 11.40% | 7700 | 0.3232 |
India | 12.16% | 5800 | 0.3226 |
Rwanda | 15.00% | 1700 | 0.3160 |
Colombia | 10.99% | 13500 | 0.3150 |
Morocco | 14.00% | 7700 | 0.2983 |
Zambia | 16.02% | 4100 | 0.2946 |
Bangladesh | 18.00% | 3400 | 0.2807 |
Kenya | 18.37% | 3100 | 0.2792 |
Namibia | 17.24% | 10800 | 0.2579 |
Ghana | 21.30% | 4200 | 0.2479 |
Kiribati | 24.21% | 1600 | 0.2414 |
Cook Islands | 19.80% | 9100 | 0.2384 |
Fiji | 20.62% | 8200 | 0.2341 |
Lesotho | 27.86% | 2900 | 0.2050 |
Solomon Islands | 31.68% | 1800 | 0.1899 |
Tonga | 30.00% | 5000 | 0.1736 |
Zimbabwe | 41.58% | 2000 | 0.1552 |
Thailand | 27.16% | 14400 | 0.1530 |
Cameroon | 41.51% | 3000 | 0.1443 |
South Africa | 29.22% | 12700 | 0.1365 |
Egypt | 35.00% | 11100 | 0.0920 |
Iraq | 34.70% | 14100 | 0.0796 |
Botswana | 38.42% | 16000 |
0.0390 |
Is this need being met?
Unfortunately, there are precious few systematic resources available on the distribution of English teachers placed in foreign countries. One source suggests ten popular destinations for recent grads to teach English. Of these ten, only one is in the above list (China) and several are of laughable stock when it comes to making an impact – Japan, South Korea, Spain.
Another approach we could take could be to study the number of Americans living abroad as an analogue for the distribution of foreign English teachers. One paper authored by the UN takes a stab at this, noting that it is much harder to differentiate American citizens from the group at large, which may itself include immigrants. Another source, an interest group for Americans living overseas, posits the following numbers:
- Africa: 171,000
- East Asia and Pacific: 864,000
- Europe: 1,612,000
- Near East: 870,000
- South Central Asia: 212,000
- Western Hemisphere: 2,591,000
If 1% of each of these populations are serving as foreign English teachers, we see a disproportionate emphasis on the Asia-Pacific and European regions in terms of service, with Africa and South Central Asia being demonstrably underserved. The decision to incorporate Latin and Southern America into “Western Hemisphere” is unfortunate and obfuscates their numbers.
Let’s try one final technique to gauge whether the needs of countries that might ostensibly benefit from English proficiency are being met – comparing the number of Google search results for the query of “teaching English in [country]” for the top ten countries in our target list vs the ten most popular countries as suggested by the first source (these countries bolded; China is in both lists)
Country | Google Search Results (millions) |
Chile | 3.43 |
China | 62.60 |
Costa Rica | 2.87 |
Czech Republic | 4.15 |
Dominican Republic | 1.65 |
Ethiopia | 2.18 |
Honduras | 1.67 |
Japan | 42.3 |
Malawi | 1.70 |
Nepal | 2.62 |
Saudi Arabia | 5.19 |
South Korea | 6.54 |
Spain | 22.3 |
Swaziland | 1.43 |
The Gambia | 1.18 |
Tokelau | 0.87 |
Turkey | 18.30 |
Uganda | 2.25 |
Vietnam | 7.31 |
Conclusions
Without a doubt, the countries in which English education programs are being hosted are not being chosen for their relative need. Of ten ostensibly high-demand destinations for expatriate English teachers, nine were above the global median in either GDP per capita based on Purchasing Power Parity or English proficiency. Of those countries that fall below the median in both of these metrics, there does not appear to be substantial interest in expatriate English education, with the possible exception of China (this author is skeptical of the accuracy of the data provided for China).
If the purpose of expatriate English education programs is to improve economic outcomes in countries that have low English proficiency, it is apparent from this analysis that that purpose is not being met.