“Native English” is not a hiring signal: What 82,000 reasoning tests taught us
This content submission is written by Alan Schultz, Chief Marketing Officer of Virtustant, a remote staffing agency that places professionals from Latin America with US companies. He writes about hiring and managing remote teams.
Most remote hiring in Latin America starts with an English filter. A job post asks for “fluent” or “native-level” English, applicants pick a level from a dropdown, and the shortlist gets built from the top of that list.
It is a reasonable instinct. For a role that talks to US customers or sits in US meetings all day, English matters.
The problem is what happens next. Once English is the first filter, it quietly becomes a stand-in for everything else: judgment, speed of learning, attention to detail. We test for those things separately, and our data says the shortcut does not hold, especially at the very top of the English scale.
What we looked at
When candidates apply to Virtustant, they report their English level (B1, B2, C1, C2 or “native”) and the monthly pay they want in US dollars.
Many also take our cognitive reasoning test, scored from 0 to 100. Between January 1 and October 30, 2026, about 82,000 candidates living in 18 Latin American countries took that test.
We limited the analysis to Latin America on purpose.
Across all regions, the “native” group is dominated by applicants from countries where English is a first language, which would turn any comparison into a comparison between countries. Keeping the pool to Latin America makes the question narrower and fairer: among people from the same region, what does the “native” label tell you?
The pattern
| Self-reported English | Candidates tested | Median reasoning score | Average reasoning score | Average monthly pay requested |
|---|---|---|---|---|
| B1 | ~9,000 | 57 | 53.6 | $1,542 |
| B2 | ~14,200 | 63 | 58.4 | $1,630 |
| C1 | ~29,900 | 67 | 62.2 | $1,783 |
| C2 | ~24,000 | 70 | 63.8 | $1,998 |
| Native | ~4,800 | 63 | 57.5 | $2,089 |
Pay averages exclude entries above $20,000 a month, which are almost always typos.
From B1 to C2, the two lines move together. Candidates who report stronger English also score higher on reasoning, by about 13 points at the median. That is not surprising.
Reaching C1 or C2 in a second language takes years of deliberate study, and the habits that get people there overlap with the habits a reasoning test rewards.
Then the “native” row breaks the pattern. Candidates who describe their English as native score like the B2 group, with the same median of 63 and an average about six points below C2.
At the same time, they ask for the most money of any group: on average about $90 a month more than C2 candidates and about $460 more than B2 candidates, who score the same.
We checked whether a single large country was driving this. Restricting the data to Argentina, Colombia and Mexico gives the same shape: C1 and C2 candidates have a median score of 70, while the “native” group’s median is 63, and the native group again asks for the most.
Why this probably happens
We can’t tell from this data why the native group scores lower, and we would be careful with anyone who claims they can. A few explanations are plausible, and none of them is a judgment about the people involved.
- “Native” is not a CEFR level. B1 through C2 describe proficiency against a framework. “Native” describes a background: a bilingual household, years in an English-speaking country, a heritage speaker. It says nothing about how far someone pushed themselves academically.
- Self-assessment works differently at the top. A candidate who studied English for ten years knows exactly where they fall short and often picks C1 even when they speak very well. A candidate who grew up with English tends to pick “native” without thinking about it. The label measures confidence as much as skill.
- Role mix. Candidates with native English may apply more often for roles where spoken English is the core of the job, such as customer service or sales, and the groups mix roles and seniority levels.
What the data does show clearly is that the label and the pay request rise together, while the reasoning score does not. If your process lets the English label set the price and the shortlist, you can end up paying a premium for a signal that does not carry the information you think it does.
What this means for your own hiring
None of this argues against hiring native speakers, and none of it means a reasoning score should decide who gets hired. The research on selection methods has moved in the last few years.
A 2022 re-analysis by Sackett and colleagues in the Journal of Applied Psychology found that earlier estimates for general cognitive ability had been overcorrected. Its predictive validity came down from the long-quoted 0.51 to about 0.31.
Structured interviews came out on top at about 0.42, ahead of job knowledge tests (0.40) and work samples (0.33). A reasoning test is one useful gate, not the answer.
The practical lesson is to stop asking one measure to do the work of three.
1. Test English for the job, not for the label
If the role is spoken and client-facing, run a live conversation that mirrors a real call. If it is written support, use a writing task based on a real ticket. Score what the role needs: clarity, comprehension, tone, how the candidate handles a misunderstanding.
2. Measure reasoning separately and early
A short reasoning test is cheap to run and catches problems that a fluent interview can hide. Use it to narrow the pool, then read it as a band, not a precise number. A 61 and a 64 are the same result.
3. Make the work sample the deciding round
Give the finalists a task drawn from the actual job, scored with a written rubric by more than one person, and let that carry the most weight.
4. Price the candidate on demonstrated skill
A pay request is a starting point. If two finalists perform the same on the work sample and one asks for 25% more because of how they described their English, that difference should be part of the conversation.
How we use it
In our own vetting process, out of every 100 applicants, about 22 clear the recruiter screen, 9 pass the cognitive, skills and English tests, 3 reach a live interview and 1 is hired.
Reasoning and English are separate tests in that funnel, so a candidate who calls their English native still has to show it, and still has to clear the reasoning bar.
If you build your own process, keep them separate too: don’t let a strong score on one average away a weak result on the other.
That separation is the whole point of this data.
Self-reported English is useful for deciding who to talk to. It is not evidence of how someone thinks, and it should not be what sets their price.
The limits of these numbers
These are applicant records, not job outcomes. The English level is self-reported, the pay figures are requests made at application, and the reasoning scores come from a single test. The groups also mix roles and seniority levels.
We have not linked these scores to on-the-job performance in this analysis, so the data shows an association between a label, a score and a price, not a cause.
It is still enough to question a shortcut that many remote hiring processes rely on.
Independent
