Engineering360: "Immigrant students build STEM skills more than US-born students"

Following the links to supporting data leads one down a rabbit hole of unbelievable assumptions, substitutions, estimations, obfuscations and downright vagueness which is compounded by the degrees of hierarchal footnoting. It is a house of cards built on sand. The first footnote is an example "In the data-preferred model, there is a small but significant degree of imperfect substitutability between natives and immigrants which, when combined with the other estimated elasticities, implies that in the period from 1990 to 2006 immigration had a small effect on the wages of native workers with no high school degree (between 0.6% and +1.7%). It also had a small positive effect on average native wages (+0.6%) and a substantial negative effect (−6.7%) on wages of previous immigrants in the long run." If that is not enough try reading footnote 22 Levenshtein VI (1966) Binary codes capable of correcting deletions, insertions and reversals. Sov Phys Dokl 10:707–710. OpenUrlGoogle Scholar

This study is differentiating correlation and causation using conditional probability. As any statistician knows correlation does not imply causation (Statistics 101). The assumption that language skills or lack thereof is a causal effect in the choice of STEM vs Language courses is extremely narrow and does not address more pervasive possibilities such as culture base importance of education and the current trend in US schools of social promotion.

By their own statistics comparing age at time of immigration they show that the longer someone is in another countries school system the better they do at the more rigorous study of STEM, which requires lots of reading to understand the mathematical relationships if you are to understand what they represent. Anyone who has had an exposure to higher level Engineering studies can verify this.

Before I can take any of this as being remotely causal the authors need to examine not just the language but also the country of origin and the value that they place on education. My reasoning being that lumping in 3rd world countries where survival outweighs education to a country such as Hungary (whose language has no Latin origins) which places a high value on education is an obfuscation that is polluting the data. Why they would overlook this obvious multivariable approach is a question to be asked, as any investigator worth there salt uses multivariable analysis. My experience in the past have shown that when single variable analysis is used there is a preconceived hidden agenda that they wish to support. Or it could be naivety or a requirement for tenure or dissertation where complex words, long sentences and many footnotes imply rigor.

"There are liars, damned liars and statisticians." - Mark Twain

"Do not put your faith in what statistics say until you have carefully considered what they do not say." - William W. Watt

“Against stupidity the gods themselves contend in vain.” ― Frederich Schiller