The 2014 Brown Center Report on American Education (2014 BCR), released last week, included a study of homework. The study revisits a question investigated in the 2003 BCR: how much homework do American students have? Recent stories in the popular press have featured children burdened with an enormous amount of homework, three hours or more per night. Are these students’ experiences typical or rare?
They are rare. According to 2012 NAEP data, only five percent of nine-year-olds, seven percent of 13-year-olds, and 13 percent of 17-year-olds had more than two hours of homework the day before filling out the student questionnaire.[i] MetLife’s 2007 survey of parents and children reports similar figures. Three percent of parents with children attending elementary schools estimated three hours or more of homework on a typical school day. For parents of secondary school children, the share was five percent. In the student surveys, only two percent of kids in grade 3-6 said they had three hours or more of homework, and only eight percent of kids in grades 7-12 said they had that much.
A Very Different View
So only a small sliver of the overall population has as much as three hours of nightly homework. And yet on March 21, 2014, CNN reported dramatically different findings from a survey of 4,000 students conducted by researchers at Challenge Success (Stanford Graduate School of Education). The students in that survey had an average homework load of three hours, with some doing as much as five hours per night. The researchers also found that excessive homework was correlated with high levels of stress and health problems. Not only is the homework load onerous, the study concluded, it is also unhealthy for kids.
As a gauge of the national homework load, the study is profoundly flawed. It’s impossible to say whether the findings are even generalizable to the 10 high schools that the students attended. The schools (comprising the study’s sampling frame) are not representative. They consist of four public and six private schools, all in well-to-do suburban neighborhoods in California (median household income of $90,000). The schools are extraordinarily high-achieving, with 93 percent of graduates going to college. Fifty-four percent of the students who answered the survey are female. Only six percent are black or Hispanic. These statistics diverge significantly from statistics for the U.S. as a whole.
The sampling strategy was also vulnerable to skewing. Students were invited to fill out the questionnaire. They were not selected randomly. Self-selection can bias a sample by making a group of highly-motivated subjects appear larger than it really is. Students who were unhappy with homework were probably more motivated to fill out the survey. Those who were content or indifferent were probably less motivated. The authors of the study do not report how many students were initially invited to respond, making the response rate incalculable. The average public high school in California has more than 1,200 students—many suburban neighborhoods have larger schools—so the non-respondents probably outnumber the respondents. Four of the schools offered the questionnaire online, adding another opportunity for self-selection.
A non-representative sampling frame followed by self-selection of respondents can produce misleading survey results. A famous example of this is the Literary Digest’s 1936 public opinion poll on the presidential election. The magazine mailed out 10 million postcards to its subscribers, automobile registrants, and telephone users. More than 2 million responses led to a clear prediction: The Republican candidate, Alf Landon, would receive 57 percent of the vote and easily defeat incumbent President Franklin Roosevelt, the Democratic candidate. But that didn’t happen. Roosevelt won in a landslide, getting 61 percent of the vote.
How could the survey be so wrong? The ten million voters who were surveyed leaned strongly Republican. The Digests’ subscribers tended to be from wealthier households, as did car owners and telephone users in 1936. Combine that with the propensity of disgruntled voters to return such a survey, and the ingredients for misleading results are in place.
What Should Be Done?
Look, it’s a fact that some students have too much homework. And it’s plausible that in ten high-powered high schools one can find hundreds of students who are stressed out to the point of missing sleep, which is one of the study’s indicators of poor health. But that is not the norm. Stress from academic expectations is not the experience of the average American teen.
Laurence Steinberg, a psychologist at Temple University, has devoted his career to studying adolescents, including their cognitive and emotional development and health. His research has drawn on databases designed to be nationally representative (e.g., National Longitudinal Study of Adolescent Health). The tenth edition of Steinberg’s book, Adolescence, was published in 2013. Writing recently in Slate (February, 2014), Steinberg calls our high schools a “disaster,” primarily because they ask so little of students. Steinberg recommends “classes that really challenge students to work hard.”
Parents and teachers should monitor students for taking on an academic workload that is too stressful. But to conclude from anecdotal reports and case studies that American high school students are overworked would be wrong. There also is an element of the Stanford study that makes those students’ situation puzzling. Recall that six of the schools in the Stanford study are private schools. High-powered, academically-focused high schools are not for everyone. School officials in such schools typically make the homework load clear to prospective students and their parents. AP classes aren’t for everyone either. It’s like going to a steak house for dinner and then getting sick from the menu because you’re a vegetarian. You made a bad choice. My advice is to eat somewhere else.
[i] Asking students for the amount of homework on the day before taking the NAEP test presents both benefits and risks The benefits are a more precise timeframe for the homework estimate (as opposed to “usually”) and reliance on short term memory. A risk is that teachers, on the day before the test, may assign less or no homework to students selected for NAEP. The Brown Center Report compared responses to this item with responses to a separate NAEP question, posing the question as “usually,” that is now discontinued but asked until 2004. There is evidence of under-reporting at the low end of the homework load (e.g., the no homework group), but the percentage of students with a heavy homework load appears unaffected. See the discussion on pages 20-21.