Building a Competitive Education Industry
A Weekly Column by Myron Lieberman

In the article, the authors conclude that the teacher unions do not hinder education performance.  In drawing this conclusion, they have relied upon the positive correlations between teacher unionization and higher SAT scores. That is, the states with higher SAT scores tend to be the unionized states.  Since the correlation is alleged to be rather high, the authors conclude that the teacher unions do not hinder educational performance. 

However, to demonstrate the impact of teacher unionization, it would be essential to provide data on educational performance before and after unionization.  Despite the complete absence of such data, the authors conclude that the high test scores in unionized states demonstrate that the teacher unions do not hinder educational performance.  This illustrates the poor logic that characterizes the article; for all we know, the test scores might have been much higher in the absence of unionization. 

To appreciate this point, consider the fact that the teacher unions are the strongest in large cities, such as new York City, Chicago, Los Angeles, Philadelphia, new Orleans, Miami, and so on.  This is true even in states with bargaining laws.  The test scores in the large urban districts are lower than the test scores of the remainder of the test-taking population.  Does the correlation between unionization and low SAT/ACT scores in large urban districts indicate that the teacher unions hinder education performance in these districts?  It does not, as the teacher unions would be the first to argue.  For all we know, the test scores might have been even lower in the absence of unionization.  Actually, the teacher unions do hinder educational performance in inner city schools but this conclusion is based upon such factors as the union emphasis on seniority in assignment which invariably results in the assignment of high proportions of new, inexperienced and/or substitute teachers in these schools.

A simple two-state example illustrates our criticism.  Connecticut is a state with high average SAT scores.  It is also a state in which teachers are highly unionized.  In contrast, North Caroline is a state showing much lower SAT scores.  It is also a state in which unionization is prohibited.  Can we deduce from these facts the conclusion that teacher unions have a positive impact on education achievement?  We cannot.  Connecticut had much higher average SAT scores than North Carolina before Connecticut teachers were unionized, and it is safe to say that if teacher unions were prohibited in Connecticut and allowed in North Carolina, SAT scores in Connecticut would still be much higher in Connecticut than in North Carolina.   And no matter how many states with high and low SAT scores we add to the mix, the conclusion remains the same.  We cannot - logically, that is - draw any conclusions about the impact of the teacher unions on educational achievement from the correlations data cited in the article.

The article is also seriously flawed in terms of who takes and who does not take the tests.  The percentage of high school students who take the SAT varies from four percent in Utah and Mississippi to 88 percent in Connecticut.  What is the educational impact of unionization on students who do not take the SAT?  The authors say nothing about it, but the omission undermines any conclusions about the impact of the teacher unions on achievement. 

The authors' reliance upon SAT scores is deficient for several other reasons as well: 

• The SAT scores include the scores of pupils who are homeschooled - a group that scores above average but is not subject to teacher unionization.
• The state SAT scores also include the scores of private school test takers, who average higher than public school test takers but are rarely educated in unionized schools.
• A large number of pupils have divided their K-12 years between public and private schools.  These considerations indicate that "average SAT scores" is a very unreliable guide to educational performance.

One of the most glaring weaknesses in the Review article is that it completely ignores the data about the effects of unionization outside of education, but gives no reason or explanation for this omission.  What is there about education that justifies the conclusion that teacher unions will not have the negative effects on productivity that have emerged in other unionized industries?  The authors do not raise this question, asserting only that how teacher unionization affects educational performance is a "mystery" worth of further study.  Indeed, the article does not discuss or even cite some of the best recent research on the impact of the teacher unions on educational achievement.  For example, the article does not cite a study by Sam Peltzman of the University of Chicago, in which Peltzman found that teacher unionization was the only factor that could explain the decline in SAT scores since 1962.  This omission is all the more surprising because Peltzman's findings are discussed in a book cited by the authors as one that criticizes teacher unionization without any empirical evidence to support its conclusions.

Perhaps the most telling commentary is the fact that the authors previously published the gist of the Harvard article in a different journal, where it languished without attention.  Once it was published in a journal with "Harvard" in the title, some policymakers automatically assume that the article must have merit.  Unfortunately, it does not.

Dr. Myron Lieberman, Senior Research Scholar, Social Philosophy and Policy Center, Bowling Green State University, Ohio, and Charlene K. Haar, President, Education Policy Institute, Washington, DC.

This analysis is adapted from a forthcoming article in School Reform News.

1. The Peltzman study is discussed briefly in Myron Lieberman, The Teacher Unions (New York: Free Press, 1997), p. 220. The Teacher Unions devotes seven full pages to the arguments pro and con regarding the impact of the teacher unions on educational achievement (pp. 217-225). The 2000 paperback edition does also (pp. 231-238).