American 15-year-olds performed better in reading and science than most students around the world, but lagged behind in math, according to the latest results of an oft-cited international exam.

The results offer a glass half full and half empty, and a somewhat rosier picture than the recent federal exams for fourth- and eighth-graders, which showed declines in reading. Since the last international test in 2015, American students have gained relative to other countries in all three subjects. And since 2006, scores have risen in science.

But the longer-run trend lines in math and reading scores are flat. And the latest scores show large disparities in performance between affluent and poor students — even larger than those seen in most other countries.

“Maybe we can convince ourselves that we’re doing ok with reading,” said Peggy Carr, associate commissioner at the National Center for Education Statistics at the Department of Education in a press call with reporters. “But in math there’s a clear message — we’re struggling in math in comparison to our peers around the world.”

The exam, known as PISA, was first administered in 2000. The exams are designed to measure 15-year-olds’ high-level understanding of each subject, and are administered to a nationally representative sample of public and private school students from roughly 70 countries every three years. The OECD,  an international organization, whose aim is to foster cooperation across different countries, gives the exams.

The U.S. compared favorably in both reading and science relative to the other countries in the OECD, a collection of about three dozen largely industrialized nations.

In reading, U.S. students clearly outscored 21 other OECD nations, including France and Israel, but scored lower than just four others, including Canada and Finland. (There was no statistical difference compared with the other 10.) Similarly, 14% of American students scored in the top tier, compared with an OECD average of 9%.

In science, the American 15-year-olds did better than 19 other countries, and worse than six. The U.S. was one of only a handful of countries that saw a significant improvement in scores between 2006 and 2018.

These findings might be surprising since Secretary of Education Betsy DeVos — as well as one of her predecessors, Arne Duncan — often cite these international comparisons to argue that the country’s schools are mediocre and need substantial reform. In 2015, the U.S. scored at about average in reading and science, so the 2018 results represent a relative improvement, though that partially reflects declining performance in other countries.

Results were much bleaker in math on PISA, though. Scores of American high schoolers were statistically worse than those in 24 other OECD nations — including Japan, Latvia, and the Czech Republic — and clearly better than just six. More American students were considered low performers in math (27% vs. 24% in the average OECC) and fewer were high performers (8% vs. 11%).

Scores were closely linked to poverty and socioeconomic status. There was a moderate correlation between a nation’s poverty level and its test scores — those with a lower level of relative poverty generally did better, but there were many nations that bucked the trend in either direction. The U.S. has a high relative poverty rate, but also has a very high average household income.

And within countries, poorer students did worse on average than their more affluent peers. Across subjects, this gap was somewhat larger in the United States than most other countries, which may reflect greater economic inequality. However, even advantaged American students scored worse in math than affluent students in most other countries.

Research has found that low-income students in America do better academically when their families receive additional resources, including through government programs.

The results follow recent U.S. federal exams showing that fourth- and eighth-grade reading and math scores have been flat — or in some cases declining — for over a decade. The last federal exam for high schoolers was from 2015; these scores have generally been stagnant for many years. On a separate international test, TIMMS, which focuses on younger students, American fourth- and eighth-graders scored well above average in math and science, according to most recent results, from 2015.

High school graduation rates have increased sharply in recent years, but that may be due to credit recovery programs that allow students to rapidly make up courses, and changes in graduation standards, as opposed to increases in student learning.

There are reasons to take the PISA results seriously but also with a grain of salt.

For instance, some research has found that prior international tests, including PISA, are correlated with economic growth, suggesting that they test skills that may translate into economic prosperity for a country. But the U.S. has been something of an outlier in this regard — strong economic growth, even with middling international test scores.

Another issue is whether 15-year-olds who take these tests actually do their best. Perhaps these students — already subject to a battery of tests — simply don’t try that hard on these low-stakes assessments. There’s some evidence that’s the case.

One recent study found that about a third of the difference between countries is due to differences in student effort, as measured, for instance, by how well a student does by the end of the test versus the beginning. But these results don’t mean the results should be ignored, because student effort may matter in itself.

“Our measures of survey and test effort have been shown to capture important student noncognitive skills related to conscientiousness, diligence, and perseverance,” wrote the researchers.

Finally, other academics have warned that international test scores can’t — or at least shouldn’t — be used to make policy prescriptions. The reason is that it’s extremely difficult to say what policies explain better or worse scores between different countries. And even if we knew which policies worked in one country, it’s not clear they could be exported elsewhere.

“Education systems develop in social and political contexts and are inseparable from those contexts,” wrote three researchers in a 2015 report on international test scores. For this and other reasons they conclude, “It is challenging to learn about improving U.S. schools from comparisons based on international tests.”

That hasn’t stopped some countries, like Finland, from becoming educational darlings on the strength of their PISA scores. But Finland in particular has seen its scores fall substantially since the 2000s, and the latest round of results show the country has not recovered. Finnish scores are still well above average but have dropped more than any other country in the world in both science and math.