Count the hours a semester abroad actually buys

Living somewhere supplies four things: volume of contact, stakes, a variety of speakers, and no opt-out. The fourth one you cannot buy at home and I am not going to pretend otherwise.

The first of the four, volume of contact, you can count. And counting it is useful, because a semester's price tag and a semester's hours are two very different-looking numbers once you put them side by side.

The three lines of arithmetic

No model, no fitted constant, nothing hidden:

The comparison bars divide those contact hours by 552 and by 690, the Foreign Service Institute's Category I band. Spanish, French, Italian and Portuguese all sit in that category, and FSI lists Spanish and French at 30 weeks.

One caution on the 552 and 690 denominators, since this site has picked at them elsewhere. FSI's published hours and its published weeks do not all divide cleanly: three categories give 23 class hours a week and the fourth gives 25. Category I is one of the three that agree, so 552 and 690 are the internally consistent end of that page. Even so, FSI's band is class hours for full-time diplomats with a curriculum and a teacher, and your hours in a bar in Arequipa are not the same unit. The bar is a scale, not a score.

One case, worked by hand

An 18-week semester, 16.7 hours a week in Spanish, $18,000 all in, against a $16 tutor.

Type 18, 16.7, $18,000 and $16 in and the tool prints 301, $59.88 and 1,125. The 16.7 and the 18 are not invented either: they are where a semester in Peru buys about 300 hours of Spanish lands, working from Slagter and Pyper's reported 25.7 speaking hours a week minus about 9 hours of English.

1,125 hours of tutoring is an absurd figure and I want to say why before anybody quotes it at me. Nobody books three hours of one-to-one Spanish a day for a year. The comparison is a unit-price comparison and not a plan, and the honest reading is that the money buys a great deal of speaking time in either format, spent very differently.

Does studying abroad improve your language?

Studying abroad improves your fluency, though no study has shown the trip itself is the cause, and that's the part most pages about this get wrong in one direction or the other.

The fluency gain is real. Kim, Lee and Lee (2024) pooled 14 between-group studies of second-language fluency and found Hedges' g = 0.65, 95% CI [0.44, 0.87]. That interval sits clear of zero. Yang (2016) got d = 0.75 across 11 studies. In Llanes and Muñoz (2013) the students abroad gained fluency at d = 0.77 across a semester while those at home moved at d = 0.10. Something happens to your mouth when you spend three months using it, and Moore's team, having catalogued every weakness in this literature, still concluded that going abroad "can have a positive impact on linguistic ability".

The causal claim is another matter. Moore, Torgerson and Beckmann (2021) searched nine databases, screened 2,548 records looking for a randomised controlled trial of study abroad, and found none. Six eligible studies, all quasi-experimental. Yang's meta-analysis: zero randomised. Kim, Lee and Lee: "all based on quasi-experimental designs". Across 31 study slots in three separate syntheses, not one random allocation, and every row of Moore's quality table says the same seven words in the grouping column: students self-selected into learning context.

So the pooled effect is estimated precisely and it is estimating a confounded quantity. How much of that 0.65 is Seville and how much is the kind of 20-year-old who applies to spend a year in Seville, nobody can say from these designs. The reason is decent, incidentally: randomly assigning a student to a year abroad or a year at home raises ethical problems no university will take on. A missing trial is still missing.

And the study-abroad gain is front-loaded. Most of the measured proficiency advantage arrives in the first three months, after which the study-abroad variable stops predicting outcomes in the regression. That shape is the practical finding hiding in a literature everybody quotes for its headline effect size. The full walk through all nine outcome domains has the rest, including the parts that cut against me.

That open attribution is why this tool counts hours instead of grading the decision. Hours are countable, the effect is real, and the attribution is open. Three claims that all fit together.

What this does not tell you

Contact hours are not interchangeable. An hour ordering coffee, an hour in a grammar class and an hour half-listening to a television in the corner all count as one hour here. They are not one thing. Input you cannot parse teaches you very little, and sitting in a café where the conversation runs past your comprehension is acoustically similar to sitting at home with a radio on.

Your weekly hours figure is a guess until you log it. The 16.7 default is arithmetic on one study's averages. Slagter and Pyper's students reported between just under 6 and 13 hours a week of speaking English inside a program where every class ran in Spanish, so the spread around any average here is enormous, and the students who reported less English showed the largest gains.

The tutoring column buys a different product. One-to-one hours give you correction, a curriculum and someone who notices what you keep getting wrong. They do not give you stakes, a bus driver, a ten-year-old, or the language arriving when you are tired and would rather it did not. That last one is the thing no software and no tutor sells.

Nothing here prices the rest of the trip. The friendships, the food, the year of your life. This tool divides money by language hours because language hours are the part with a number on them, which is a limitation of the tool rather than a claim about what matters.

My conflict of interest. I sell software to people learning a language without moving abroad, so an argument against going flatters me. I have tried to give the other side its strongest numbers, with its own effect sizes, and to keep the sentence that cuts against me: the review that found no randomised trial still recommended sojourning as a catalyst for language development.

Sources

Moore, I., Torgerson, C., & Beckmann, N. (2021), the systematic review that screened 2,548 records and found no randomised trial. Kim, Lee & Lee (2024) for the pooled g = 0.65 across 14 between-group studies. Yang (2016) for d = 0.75 across 11. Llanes & Muñoz (2013) and Segalowitz & Freed (2004) for the oral-fluency comparisons. Slagter, C., & Pyper, M. for the 25.7 speaking hours a week and the English-use range, across semester programs in Spain, Peru and Honduras.

FSI's Category I band and its 23-and-17 week come from the National Foreign Affairs Training Center page, read 3 September 2026. All of it is worked through in the review of the review and in the 300-hour arithmetic.

If you want the other half of this, what the four things living somewhere supplies cost to replace at home, that post itemises them, including the one that cannot be replaced.

Questions people actually ask