What If Everything You Thought You Knew About Teachers Unions Turned Out to be Wrong?

Iowa just became the latest state to limit collective bargaining rights for teachers. In other states, that’s meant big salary cuts for teachers…

Jennifer Berkshire: It’s a well-known true fact that teachers unions make it much harder to get rid of bad teachers. But you conducted a study that purports to find the opposite. In fact, you titled your study The Myth of Unions’ Overprotection of Bad Teachers. Tell us about what you found.

Eunice Han: What I found is that the facts are the opposite of what people think: that highly unionized districts actually fire more bad teachers.

Berkshire: That sound you just heard was of jaws collectively dropping. While we give readers a chance to re-combombulate themselves (and arm themselves anew with anecdotes), can you walk us through your argument? And feel free to use a formula. 

Han: It’s pretty simple, really. By demanding higher salaries for teachers, unions give school districts a strong incentive to dismiss ineffective teachers before they get tenure. Highly unionized districts dismiss more bad teachers because it costs more to keep them. Using three different kinds of survey data from the National Center for Education Statistics, I confirmed that unionized districts dismiss more low-quality teachers than those with weak unions formula2or no unions. Unionized districts also retain more high-quality teachers relative to district with weak unionism. No matter how and when I measured unionism I found that unions lowered teacher attrition. This is important because many studies have found that higher quality teachers have a greater chance of leaving the profession. Since unionized districts dismiss more bad teachers while keeping more good teachers, we should expect to observe higher teacher quality in highly unionized districts than less unionized districts – and this is exactly what I found. Highly unionized districts have more qualified teachers compared to districts with weak unionism.

Berkshire: You’re an economist and your study takes a basic tenet of microeconomic theory, that as the price of labor rises, less labor will be employed. We hear this argument constantly, except, as you point out, in the case of teachers unions, which have somehow figured out how to have their expensive cake AND not get fired for eating it on the job.

14641658051.jpg (1300×865)Han: We know that unions increase salaries and benefits, but people also argue that unions make it harder for teachers to get fired. Rarely do you see those two things happening at the same time. Based on microeconomic theory, as salaries go up, employment goes down because the employers can’t afford both unless there is some dramatic increase in revenue. This is especially true in districts that are under intense financial pressure. Think about the economic argument that you hear being made against raising minimum wages for fast food workers, that paying employees more will trigger higher unemployment. But for some reason, when it comes to teachers unions the claim is made that they are getting both: higher salaries AND higher employment. I thought that something was missing so I decided to investigate. This is the first study to rigorously test this assertion.  

Berkshire: In 2011, four states essentially eliminated collective bargaining for teachers, which gave you an unusual opportunity to test your argument in a real-life laboratory. What did you find?

Indiana, Idaho, Tennessee and Wisconsin all changed their laws in 2010-2011, dramatically restricting the collective bargaining power of public school teachers. After that, I was able to compare what happened in states where teachers’ bargaining rights were limited to states where there was no change. If you believe the argument that teachers unions protect bad teachers, we should have seen teacher quality rise in those states after the laws changed. Instead I found that the opposite happened.

Han: Indiana, Idaho, Tennessee and Wisconsin all changed their laws in 2010-2011, dramatically restricting the collective bargaining power of public school teachers. After that, I was able to compare what happened in states where teachers’ bargaining rights were limited to states where there was no change. If you believe the argument that teachers unions protect bad teachers, we should have seen teacher quality rise in those states after the laws changed. Instead I found that the opposite happened. The new laws restricting bargaining rights in those four states reduced teacher salaries by about 9%. That’s a huge number. A 9% drop in teachers salaries is unheard of. Lower salaries mean that districts have less incentive to sort out better teachers, lowering the dismissal rate of underperforming teachers, which is what you saw happen in the those four states. Lower salaries also encouraged high-quality teachers to leave the teaching sector, which contributed to a decrease of teacher quality.

Berkshire: You conclude that teachers unions are a net positive for educational quality. (Note that I’m looking over my shoulder even as I type these words!) But I think what a lot of people reading this will want to know is what impact unions have on student achievement.

Han: Since there’s currently no data on student performance by school district levels with nationally representative samples, I use high school dropout rates as a measure of student achievement. My study found that unions reduce the dropout rates of districts. This is where my study differs from some earlier ones that found that unionism either had no impact or had a negative effect on the dropout rate. I define unionism more broadly than those earlier studies. It’s not just collective bargaining that matters, it’s the union density of teachers in a district that’s important. Union density measures the strength of the union, because even when teachers can’t engage in collective bargaining they can use their collective *voice* to influence the educational system. What I found was that union density significantly decreased the high school dropout rate, even in districts without collective bargaining agreements. This is important because, as the research of Raj Chetty and others has found, the upward mobility of an area is higher when the dropout rate is lower. So, when unions, via high union density, reduce the dropout rate, they improve the educational attainment as well as the welfare of all children in the area. 

Berkshire: Your study upends so many assumptions people hold about teacher unions, and I’m helpfully including another link here in hopes that they will peruse your findings in detail. If there’s one thing you’d like them to take away from your research, what would it be?

Han: I hope that people open their eyes to these results and move beyond their prejudice. I used to share that prejudice before I did this study. Obviously, if people can accept the findings of my paper, the direct policy implication is that we should be promoting union-friendly environments.  

Eunice Han is an assistant professor of economics at the University of Utah this fall. She has a PhD in economics from Harvard University.



  1. I was really disappointed. As someone who defends teachers on a regular basis, I often find myself trying to disabuse people of the myth of the unfirable teacher. The subtitle of this post gave me hope that some great ammunition was on its way, but this is really pretty weak. As I read the study, the basic thought process seems to be: Union teachers make more, so districts have more to gain by firing them. Since the district is incentivized for firing them, it makes sense that they would do so at a higher rate.

    The people I argue with will tear that apart immediately, because it is basically conjecture. What I was hoping for was a study that showed that unionized teachers actually were dismissed for incompetence at a higher rate than non-unionized teachers.

    1. It’s not just conjecture. It was supported by Data showing that unions cause retention of highly qualified teachers. It’s not just that she thought it would happen; it did happen.

      1. In the beginning. There is no account for the long-term effects of tenure and the bureaucratic structure that unions protect.

  2. It is not, and never has been difficult to “fire bad teachers” — when and IF administrators do their jobs.

  3. A true dropout rate, that is periodically audited, is still too elusive to draw any significant conclusions. Perhaps politicians prefer things that way.

  4. I’d really like to agree with the conclusion here, but I have to admit I’m a little lost on the opening arguments. What, exactly, is a “low quality” teacher or a “high-quality” teacher? And how did you determine which kinds each district was firing more of?

  5. It’s not difficult to reduce drop-out rates and increase graduation rates. You just make the choice to pass students, regardless of academic performance. It is therefore a very suspect standard to use in assessing the effectiveness of a teacher population.

      1. Which is a joke. The bar for proficiency 10th grade is way lower than I younger grades. On math a proficient 8th grader could probably
        Pass the tenth grade test.

  6. Once again it is a pleasure to see serious economic research taken seriously. Even a citation of Chetty’s work that suggest it has merit. This would never happen on some other blogs.

    Let me suggest, however, an alternate explanation for why districts with relatively high pay are more willing to fire low performing teachers before tenure: districts that pay more can be confident that a new hire will have a higher likelihood of being a strong teacher than a district that pays less. I also suspect that high paying districts are more likely to be urban districts, and it is simply easier to find candidates for teaching positions in more urban settings. It is not really anything to do with the unionization, but rather how salary structure of a job impacts the applicant pool. The same is really true for retention: pay high salaries and you will retain more people, pay low salaries and people will find more lucrative employment elsewhere.

    As for dropout rates, was the evidence for this based on comparing the states with changes in union bargaining power to dropout rates in control states without changes in union bargaining power? It is not clear from the article if the stated impact of unionization on drop out rates was based on a difference in difference analysis or correlation between union density and drop out rates.

    1. TE, the point is that higher paying districts are much more likely to be strong union districts.

      1. Rebecca,

        Higher paying districts are also much more likely to be more urban, coastal, and have a higher state per capita GDP. Which of these factors do you think actually causes higher pay?

        A difference in difference approach is an attempt to address situation with multiple different causal explanations. I hope that the econometric model used to make this claim is made clear.

        1. I’m sending a list of questions to Eunice Han as several people have expressed interest. May I ask if you read the paper? Because she goes into her model in great detail… I’ll share her responses when I get them.

        2. I took a closer look at the paper, and here are my thoughts:

          1. The theoretical model is based on the idea that schools discover the productivity of a teacher in the first period and make employment decisions based on what they learn. The measure of teacher quality used in the empirical section is that a teacher has certain certifications. That is known to the school when the teacher is hired at the beginning of the first period, so the measure of quality used in the empirical section of the paper is inconsistent with the theoretical model.

          2. About 6% of the schools in the data set are Charter Schools, but unionization is measured at the district, not school, level. Few of the schools are unionized even if they are located in a highly unionized district.

          3. Footnote 16 is the important issue. Here is alternative story to go with the empirical observations. Highly unionized districts tend to have a labor vs management view of schools. Finding and dismissing low quality teachers is managements job, so teachers there will not “council out” low productivity teachers and management will have to terminate low quality teachers. Districts with low unionization rates take a more cooperative view of the enterprise and thus teachers will see it as a duty to “council out” low productivity teachers. Fewer low quality teachers are fired because they quit after their fellow teachers intervene. In this explanation, the view of a labor/management distinction drives both unionization rates, teachers being fired by management and teachers being convinced by their fellow teachers that they are low productivity workers.

          If unionization simply changes how low productivity teachers leave the teaching, it does not improve teacher quality. I suspect that the higher teacher retention rate with unionization reflects a higher retention of low productivity teachers.

          1. As promised I shared your questions with Eunice Han. Here’s what she had to say:

            On the question of whether geography is more likely to play a role in higher wages than unionism: “I took these factors into account in my analysis. I control for locality differences such as living cost of the area, urbanism, regional differences (Southern, Northeastern areas, etc).”

            And on your alternative theory to go with Han’s empirical observations: “This is an interesting theory. However, I have another paper coming up and it shows that teachers in highly unionized districts shares this same view (as teaching as a cooperative enterprise), and basically feel that they are working as part of a team. This finding would seem to contradict this alternative theory.”

    2. Very true to some extend albeit unions also can provide protection against arbitrary and unreasonable management behaviour.

      However think it out. Why would anyone want to crush a union : hint, it let’s them pay teachers a lot less. See Kansas. Of course Kansas does not have any teachers left.

      Think of “unionization” as an intervening variable.

      1. jrkrideau,

        I took a look at teacher salaries in 2016 and in terms of starting salary, Kansas pays more than Iowa, Ohio, Utah, Arkansas, West Virginia, South Carolina, Colorado, North Dakota, New Mexico, Arizona, Maine, Oklahoma, Mississippi, Idaho, Nebraska, North Carolina, Missouri, South Dakota, and Montana. In terms of average salary, Kansas pays more than North Dakota, Arkansas, Florida, North Carolina, West Virginia, New Mexico, Oklahoma, Mississippi, and South Dakota.

        If Kansas has no more teachers left, what about these other states?

        Source of salary data: http://www.teacherportal.com/teacher-salaries-by-state/

        1. Just making sure I understand your argument(s) before I pass them along. So higher attrition in districts with low or no unionism can be explained by teachers acting as quality checks on one another and forcing their low-performing colleagues out of the profession. Meanwhile, higher salaries in union districts are a result, not of unions demanding higher salaries but of the fact that these districts are urban, coastal and in states with higher per capita incomes. Did I miss anything?

  7. I served as a teacher and union negotiator for 35 years. Teachers do not want to support consistent ,weak performers. Poor teachers make the environment in a school difficult for all teachers, administrators, students, and parents. We would show them a path out and most appreciated our effort because they are well aware of their ineffectiveness. Our union news paper of about eight pages dedicated about one page to salaries and benefits and seven pages to improving teacher performance to help students. Be assured that a teacher who challenges students does not always win a popularity contest, until later.

  8. TE – Do you have any evidence for your alternative story? Do you have data that shows the actions taken by staff towards less competetent colleagues actually differ based on unionization? BTW – I think you meant “counsel out”.

  9. Dr. Han writes: “Critics claim that teachers unions overprotect the job security of ineffective teachers and that this practice is detrimental to educational outcomes. At first, this claim appears legitimate because teachers unions may seek to protect the job security of teachers, as any other workers associations will. However, the job security of public school teachers is addressed through the tenure system in most states, and tenured teachers are not easily dismissed, regardless of their union status.”

    That “tenured teachers are not easily dismissed, regardless of their union status” doesn’t gainsay the possibility that dismissal procedures are more arduous and expensive in locations that have high teacher union membership density and collective bargaining compared to locations that don’t. Significant variation along those lines might permit an additional alternative explanation, other than salary, for any higher rate of pre-tenure teacher dismissal where unions are strongest.

    Does Dr. Han have any findings that speak to that possibility?


    1. Passed this question along to her. And promised it would be the last one as I’ve been rather overwhelming her with them!

    2. As promised!

      Han writes: As long as there is some form of economic incentive (this can include salary, benefits, or any other non-pecuniary compensation), my theory will stand. I find that highly unionized districts do not dismiss fewer tenured teachers than less unionized districts. This implies that potentially “more arduous and expensive tenure procedures” (which should be considered a form of non-pecuniary compensation) in highly unionized districts do not make a significant difference in actual dismissal decision.”

      1. Thanks, Jennifer. After posting the question, I found this related material:

        “Staff from the National Council on Teacher Quality conducted an analysis of collective bargaining provisions related to teacher dismissal and found that ‘a third of the nation’s 50 largest districts prescribe the procedures that must be followed in order to dismiss a weak teacher.’ Provisions in collective bargaining agreements include where the hearing is held, the documentation and assistance process that must be followed to pursue a dismissal, and the appeals process—all of which have implications for the time, cost, and nature of the dismissal process.”

        “…The New Teacher Project found in one district they studied it could take 10 percent to 15 percent of a principal’s time over several months to bring one dismissal case to a hearing… In New York state, dismissal cases can take from 6 months to 18 months. It costs about $250,000 to fire an incompetent tenured teacher in New York City, and dismissal cases can take between two and five years just to be heard. The hearings themselves can take between 40 and 45 hearing days over a period of nine months or more…. An investigation of dismissal cases in California conducted by The Los Angeles Times found that ‘as a case winds its way through the system, legal costs can soar into the six figures.’ Los Angeles Weekly conducted a recent analysis of the costs of dismissing teachers for poor performance in the Los Angeles Unified School District and found ‘in the past decade, LAUSD officials spent $3.5 million trying to fire just seven of the district’s 33,000 teachers for poor classroom performance.’..

        Han: “I find that highly unionized districts do not dismiss fewer tenured teachers than less unionized districts.”

        That’s true where she’s just comparing gradations of CB districts. However, her chart on page 31 shows non-CB districts with a higher dismissal rate for tenured teachers than for any of the CB districts.

        “As long as there is some form of economic incentive (this can include salary, benefits, or any other non-pecuniary compensation), my theory will stand”.

        I think some aspects of the theories she presents could survive if teacher salaries are not the principal factor, others in precarious shape to begin with would be further weakened. She is quite explicit about salaries and wages: “The economic intuition that is overlooked in teacher dismissal is that school districts have a strong motivation to dismiss low-quality teachers if they must pay the higher salaries that unions demand. Particularly, during the probationary period, districts will carefully evaluate new teachers’ performances, as they must pay even higher wages once these teachers receive tenure.” To successfully argue whether or not “Unions’ Overprotection of Bad Teachers” is a myth, I think one would need to start with a far more complete and nuanced view of the costs (including the costs described above related to dismissal procedures) and benefits than are addressed in Han’s article with its narrow focus on teacher salaries, dismissal rates, and graduation data.

        But in any event, I’m grateful to have had a chance to look over Han’s interesting, provocative, thought-provoking paper (even if after around page 6 some of it was Greek to me). Thanks for letting us know about it. No need for a reply.

Comments are closed.