As they consider reauthorizing the No Child Left Behind law, Democratic congressional leaders amazingly are giving serious consideration to reviving portfolio assessment as a primary way to gauge the effectiveness of the $23 billion program of federal school aid.

Actually, that is more than amazing. It is nearly incomprehensible.

Why? Because in the initial wave of standards-based education reform in the 1990s, highly respected testing experts found portfolio assessment to be close to a total flop when Vermont and Kentucky substituted it for standardized multiple-choice testing.

Portfolios are collections of students’ work (essays, art, research papers, and the like) during a year. To assess the relative effectiveness of all schools in a state using portfolios, special “raters” must be paid to comb through these compilations, or the teachers themselves must do it.

The two research panels independently found these failings, among others, in the portfolio assessments done in Vermont and Kentucky:

• A failure to yield reliable comparative data.
• Large differences in the way teachers implemented portfolios.
• Great differences in the difficulty of class assignments, rendering comparisons among students or groups of students highly misleading.
• Major differences in opportunities given students to revise their work, resulting in grossly misleading data when students’ collected work was compared.
• Variations in the degree of assistance that students amassing their portfolios receive from peers, parents, teachers, and other sources.

The RAND Corporation review in Vermont, authored by Daniel Koretz, found that any positive effects of portfolio assessment had come “at a steep price in time, money, and stress.”

The study for the Kentucky legislature’s Office of Educational Accountability found the program “highly unstandardized in almost all aspects of its operation,” and policies regarding revision of student work in the portfolios “highly permissive.”

So what has changed since the mid-1990s when those two studies were done? Nothing, really. There has been no follow-up research to show that portfolio assessment is valid for high-stakes assessment, though Kentucky alone still clings to the use of writing portfolios. (One perverse consequence is that teachers are penalized if they help students correct their grammar.)

So why is there a sudden interest in dusting off portfolio assessment and using it to assess how effectively states and schools are using their No Child Left Behind money to close student achievement gaps?

The primary reason seems to be that the 3.2-million-member National Education Association, which exerts a powerful influence on Democratic politicians in particular, is engaging in a full-court press to dilute NCLB testing of reading and math skills and to substitute so-called “multiple measures” to be pressed upon the states. Those would include graduation rates, attendance rates, enrollments in Advanced Placement or International Baccalaureate courses, and “performance or portfolio assessments,” according to the NEA’s priorities for NCLB.

Given that the NEA has kept up a steady drumbeat of criticism of objective testing, and has staunchly opposed tenure reform, merit pay, and parental choice, it is reasonable to assume the big teacher union loves portfolios because they would make it much harder to determine if schools are teaching kids to read and to do math.

In a National Press Club address July 30, House Education Chairman George Miller, D-CA, seemed to be conceding ground to the NEA when he said the renewed NCLB should include “multiple measures of success.” Earlier, Senate Education Chairman Edward Kennedy, D-MA, also praised that concept.

In a Q-and-A following his talk, Miller did balk at giving an outright endorsement of portfolio assessment, calling it “a very delicate subject.” However, he has implied the technique might be useful for evaluating schools’ progress in helping English Language Learners.

That raises the question whether portfolios would become a handy device to relieve schools of the obligation to teach immigrant children English promptly, in the way that bilingual education used to delay the process indefinitely.

Standardized tests are not perfect but they give communities the best value in terms of reliability, accuracy, ability to generalize results, ease of scoring, and cost. Without them, taxpayers could get all of the $23 billion in NCLB spending with none of the accountability.