February 23, 2002

U.S. ILLITERACY RATES ARE LOWER THAN YOU THOUGHT

People critical of the state of American education commonly invoke the image of a high school graduate clutching a diploma he can't read.

But the image might be misleading, and not just because there aren't as many high school graduates as people think. There aren't as many non-readers as people think, either.

The bad news about adult illiteracy hit the headlines in 1993, when the federal Department of Education released results of the 1992 National Adult Literacy Survey. As the News' banner headline put it at the time, ``Nearly half of adults illiterate, study says.''

The survey established five performance levels for adults, with those at the two lower levels judged to have ``reading, writing and math skills so limited they are deemed illiterate,'' according to the report.

The report placed 20 percent of adults at Level 1, the lowest, meaning they could do everyday tasks such as totaling a bank deposit or locating the time and place of a meeting. Twenty-seven percent were at Level 2, able to determine the difference in price between two items or find an intersection on a street map.

Given the description of the tasks, the results seemed unnecessarily pessimistic. As a skeptical News editorial writer asked, ``Can we be this dumb?''

And the answer is no, apparently not.

About a year ago, the National Center for Educational Statistics published its final technical report on the 1992 survey. Not very many people plowed through its hundreds of forbidding pages. But Thomas Sticht, who did, was pleased to discover that Andrew Kolstad, the director of the original survey, had concluded that the criteria for assigning adults to each of the five levels were set too high.

To score at a given level, a survey respondent had to be able to do 80 percent of the tasks at that level. But even someone at Level 1 could do nearly half the Level 2 tasks, and one in six of Level 5 tasks. So the 80 percent ``response probability'' exaggerates the number of false negatives, that is, the number of people who can do what the survey results suggest they can't.

Setting the response probability too low, on the other hand, exaggerates the number of false positives, people who can't do what the survey results suggest they can.

With the response probability set at 50 percent, however, ``one can have the same degree of confidence in statements about what adults can't do as in statements about what adults can do,'' Kolstad wrote. ``(It) minimizes the total misclassification error.''

Using the 50 percent criterion, 9 percent of adults score at Level 1 and 13 percent at Level 2 - hardly good news, but decidedly less awful.

So where were the headlines? Well, The Washington Post published a story last July after Sticht wrote about Kolstad's new findings.

Kolstad told the Post that 70 percent of the people judged illiterate in 1993 said they read a newspaper once a week. ``This is not illiteracy,'' he said.

In fact, in the original survey 93 percent of respondents said the could read well or very well. Even allowing for people's natural tendency to overrate themselves, that's hard to square with the allegation that ``nearly half'' of them were illiterate.

Altering how skill levels are reported wouldn't alter the fact that those at the lowest levels often struggle economically. Those at Level 1 in the 1992 survey reported working only 18 to 19 weeks on average in 1991. Those at Level 3 or higher worked 34 to 44 weeks on average, and their incomes were nearly three times higher.

Setting performance levels is a policy choice, not a right-or-wrong answer. As Sticht observes, one cynical answer to the question ``How many functionally illiterate adults are there in the United States?'' is ``As many as you would like.''

Too high an answer generates headlines, all right, but it may not generate the most effective response. The literacy survey is due for updating in 2002; let's hope that Kolstad's research leads to better policy choices.

(675 words)