Sunday, September 7, 2014

Stop Giving Students Laptops

Should have seen this coming:
Concerned that poor youngsters aren’t learning basic computer skills, some school districts have begun purchasing laptops and distributing them to every high school and middle school student.

A new study suggests the policy may be doing more harm than good. It finds public school students in North Carolina who gain their first regular access to a home computer between the fifth and eighth grades experience “a persistent decline in reading and math test scores.”

For disadvantaged youngsters, the positive impact of having access to online instruction “may be negated by counterproductive use of computers, particularly by students in unsupervised home environments.”
Anyone who ever owned a computer could have predicted this. A computer is a great tool for the disciplined, but a unending distraction for the undisciplined. One thing that consistently helps poor students is setting up a quiet place where they can do their homework without distraction; why would anyone who had seen such a study think that adding a powerful distraction device (i.e., a laptop) would help?


G. Verloren said...

I find it highly problematic that you seem to equate "poor" with "undisciplined".

That said, however, I'd like to address this study's findings - after all, a single study is merely an anecdote, and always requires further scrutiny and followup.

I found the study in question online and I'm not impressed with the arguments being drawn from the data.

G. Verloren said...

First, it's important to distinguish what the study is actually saying. The study arrives at two different conclusions - which do not necessarily influence one another.

The first conclusion is that they were able to "corroborate earlier surveys that document broad racial and socioeconomic gaps in home computer access and use".

This merely means that yes, as you might imagine and as the data has suggested for some time, poorer families are more likely to not have access to computers, and by extension, as a consequence of racial disparities of wealth, this also means ethnic minorities are more likely to not have access to computers. That part isn't surprising, I'll grant you that.

It's the second conclusion where we get off the rails a bit. The authors themselves describe their conclusions thusly: "the introduction of home computer technology is associated with modest but statistically significant and persistent negative impacts on student math and reading test scores". (Bolded emphasis mine.) That key phrase, "modest but statistically significant" gets right to the heart of things very neatly. It means "we measured a very tiny correlation, and we know it isn't merely a sampling error".

So they've shown that, yes, with reasonable validity there is a real correlation between the introduction of computers into a home, and a drop in math and reading test scores. They have not, however, shown the causation for this. Please also note, there is no wealth or racial factor to this second conclusion - it is merely "introduction of home computer technology", with no mention of race or income levels.

These findings are based on a questionnaire that students taking their standardized tests in North Carolina are required to fill out - which seems like a poor metric for study. Asking middle school students to independently quantize the amount of time they spend using computers for various tasks every day (and doing so on a test day, where they're otherwise occupied with greater concerns and really would rather the day just be over as soon as possible) seems like a recipe for inaccuracy in the data you collect.

But assuming that all of the data is correct and coincides with reality, the results themselves are lackluster. They make two different comparisons - one based on a straight comparison of test scores compared to home computer usage, the other based on a "Fixed Effects" model that treats all explanatory variables as non-random: in this case, ostensibly "to purge the estimates of any bias associated with time-invariant unobservables" - which is basically their way of saying "our study was not constructed in a way that can measure all the factors involved, so we're ignoring or re-weighting most or all of them".

The straight comparison finds that "students with access to home computers tend to score about 2% of a standard deviation higher on both reading and math test scores, conditional on a range of covariates". So just on a direct view of "do you have a computer" and "how did you do on your test", people with computers did better overall - but, I'll admit, at 2%, only very slightly better.

The fixed effects comparison find that "students who obtain access to a home computer sometime between 5th and 8th grade tend to score between 1% and 1.3% of a standard deviation lower on their subsequent math and reading tests". Please note the exact wording there - between 1% and 1.3% of a standard deviation lower.

G. Verlorn said...

What's the standard of deviation, here? They tell us themselves: "The dependent variable in each specification has been normalized to have mean zero and standard deviation one, using the entire population of students for whom we have test scores in a given grade, year, and subject." So they're not actually using the real standard deviation of the actual test scores, but they're instead using it as a normalized value of "1".

So they find that across all students in a direct comparison, those with computer access score 2% of the average difference of any individual student's score from the mean value score of all students better overall, while in a fixed effects comparison, those who had computers first introduced to them between 5th and 8th score 1% - 1.3% of the average difference of any individual student's score from the mean value score of all students lower overall.

Ultimately, they don't tell us what that standard deviation actually is, so we're left to guess. But one thing is certain: 2%, and 1% - 1.3% of that mystery value is going to be a very small value indeed. The average variance between an indivual's score and the average score of all tested students is not going to be very large typically - perhaps an average difference in score of 10% in extreme cases. Meaning even then, students introduced to computers late in life at worst are only scoring 0.13% lower than their peers.

This is a laughably small amount. And, it only correlates to the introduction of computers - a causal link has not been established, so we we can't with any certainty state that the computers themselves are the influencing factor.

And, even if we assume a causal link, the study only concerns itself with grades 5 through 8 - so we don't get to compare to things like introducing students to computers even earlier, such as in elementary school or before, rather than inmiddle school; or examine developments later in life, seeing how exposure to computers in middle school influence performance in high school and college.

And then there's the question of how many students are actually only being exposed to computer usage for the first time during 5th and 8th grade? The study actually tells us this, with some fascinating figures.

"Across all observations in all years, nearly 85% of students reported having access to a computer at home. This access rate differs by race and socioeconomic status. Almost 90% of white students have a computer at home, compared to 75% of black students. The disparities across free or reduced price lunch participation categories are larger; 71% of recipients have access to a home computer, versus 92% of non-participants. Disparities between the extreme categories of parental education, students with high school dropout parents and those whose parents have postgraduate education, are strongest: 98% of students in the highest parent education category have access to a home computer, versus 63% of students in the lowest parent education category."

So this tells us some interesting things about how being poor and poorly educated makes you less likely to own a computer (which makes sense, just in terms of wealth available), but it also tells us something even more interesting - despite these inequities, most kids already have access to computers. Even at the most disadvantaged levels, odds are good that most children have already been introduced to computers at home. What's more - the study finds these numbers are growing all the time.

G. Verloren said...

So is it reasonable to be concerned that introducing students to computers is going to ruin their academics?

Most kids already have them. In the future, they'll be even more ubiquitous, and children will be introduced to them earlier in age. Overall, currently only 15% of students are being first introduced to computers late in their childhoods. And they're scoring only marginally worse than the other 85%. And we're not even certain why, as we've only established a correlation, not a causal link.

So no, the fancy complicated computers aren't making the "undisciplined" poor children score lower by distracting them. They're scoring a miniscule amount lower - and we have no clue why. That could just as easily be due to factors like, or linked to, the ones which resulted in them not having a computer in the first place.