Chirine Mouharam explores evidence from the teacher labor market to examine the effects of occupational licensing requirements in her senior thesis.
Abstract
This paper examines the supply-side effects of increasing licensing requirements in the teacher labor market. I exploit within-state variation in subject test requirements over 2002-2015 and use the number of initial teacher licenses issued as a proxy for new teacher supply. I find that more stringent licensing requirements – in the form of subject test requirements – significantly reduce the supply of new teachers by 22% relative to the mean. I then investigate the mechanism by which subject test requirements affect new teacher supply, suggesting more stringent licensing requirements are associated with a decrease in the number of teacher preparation program completers and have no significant effect on the pass rate on required exams. Furthermore, my results suggest that subject test requirements disproportionately deter individuals from entering the teaching profession through the alternative route compared to the traditional route, and I find no evidence of a wage premium associated with subject test requirements. Overall, my results are consistent with the view that testing has acted as a barrier to entry.
I. Introduction
Occupational licensing laws, which require workers to obtain occupation-specific licenses in order to practice legally, have grown substantially more prevalent in the United States over the past three decades. The percentage of the American workforce subject to occupational licensing grew from around 5 percent in the 1970s to about 29 percent in 2008 (Kleiner & Krueger, 2013). A common justification for occupational licensing requirements argues that licensing decreases information asymmetry between consumers and suppliers, serving as a minimum guarantee of a service’s quality or safety (Leland, 1979; Shapiro, 1986). Opponents, however, emphasize supply-side effects: licensing increases the barriers to entering the labor market, drives away qualified individuals, impedes competition and raises prices faced by consumers (Ballou & Podgursky, 1998; Kleiner, 1990; Peltzman, 1976).
In this paper, I explore the supply-side effects of increasing licensing requirements for teachers. The teaching profession is one of the earliest-licensed occupations. States began to require prospective public-school teachers to pass competency tests in the 1960s (Angrist & Guryan, 2008). Although many states have required prospective teachers to pass competency tests since the mid-twentieth century, the stringency of those requirements has increased over the past few decades. Between 1983 and 2010, the percentage of states requiring some type of certification test increased from about 20% in 1983 to over 90% in 2010 (Larsen, 2015). Certification tests for teachers fall into three categories: basic skill tests, which assess reading, writing and general mathematics; professional knowledge tests, which assess pedagogical skills; and subject matter tests, which assess knowledge of a specific subject such as middle school social studies or high school physics. While the intent of these certification test requirements is to ensure that teachers meet a minimum quality standard, numerous empirical studies have found that certification tests do not significantly affect teaching quality (Angrist & Guryan 2004, 2008; Hanushek & Pace, 1995; Goldhaber & Brewer, 2000). On the other hand, additional testing and other certification requirements may deter qualified individuals with attractive outside options from entering the teaching profession if these more stringent requirements are too costly.
The teacher labor market provides a suitable setting for studying the supply-side effects of more stringent licensing requirements for multiple reasons. First, as described in more detail in Section II, teacher licensing requirements have changed in many states in recent years, creating variation that makes it possible to study their effects. For example, several states, such as Iowa, Michigan, and Wisconsin have recently relaxed certification test requirements in order to attract new teachers (DeGrow, 2018; National Council of Teachers of English, 2017; Pfannenstiel, 2018). This suggests that policymakers are concerned that stringent licensing requirements deter prospective teachers from entering the occupation.
Second, certification test laws are easier to quantify than other requirements, such as coursework or degree requirements, which have a greater range of differentiation (Larsen, 2015).
Third, the teaching occupation is sufficiently large to permit more powerful statistical tests. To my knowledge, this paper is the first study to take advantage of this by examining the effect of more stringent certification test requirements on the supply of new teachers. Previous literature on teacher licensure requirements has largely focused on the quality and wage effects of more stringent testing requirements, overlooking the supply side. Specifically, the wage literature finds that certification test requirements are associated with a wage premium (Angrist & Guryan, 2008). While the wage premium is assumed to be caused by a shift in supply, the wage literature does not formally test that hypothesis. On the other hand, the quality literature looks at the effect of licensing test requirements on the quality of teachers but does not examine whether licensure exams affect schools’ ability to fill new teacher vacancies (Angrist & Guryan, 2004; Larsen, 2015; Goldhaber & Brewer, 2000). This paper attempts to address this gap in the existing literature by answering the following question: What are the effects of more stringent licensing requirements on the supply of new teachers?
As further discussed in Section II, there was substantial variation in subject test requirements for teachers during my 2002-2015 sample period, while other kinds of test requirements were relatively stable. As a proxy for new teacher supply, I use state-level data on the number of initial teacher licenses issued from the U.S. Department of Education Office of Title II. I regress this outcome on a dummy variable encoding the presence of subject test requirements using a fixed-effects model that exploits within-state variation in subject test law implementation over the period 2002-2015. In addition, I seek to identify possible mechanisms by which subject test laws affect new teacher supply by testing two possible hypotheses. The first is that subject test requirements affect new teacher supply by decreasing the number of prospective teachers who pass all required exams. The second is that subject test requirements reduce the supply of new teachers by deterring individuals from enrolling in teacher preparation programs and entering the profession. Notably, prospective teachers can enroll in either a traditional teacher preparation program or an alternative preparation program. I therefore further explore whether subject test laws disproportionately deter people from completing the alternative route program, since those people hold non-education degrees and thus have more alternative employment options. Finally, this paper briefly considers the consequences of additional certification test laws on teacher wages.
Consistent with the view that more stringent licensing requirements increase the barriers to entry into an occupation, my results suggest that subject test requirements reduce the number of initial teacher licenses issued by 22% relative to the sample mean. Specifically, the results from my mechanism analysis suggest that more stringent licensing requirements reduce the number of teacher preparation program completers, which suggests deterrence from the occupation, especially from those entering the teaching profession through the alternative route. Finally, I find no evidence that subject test laws create a wage premium for teachers.
The paper is organized as follows. Section II provides a background on various forms of teacher testing and motivation for the research question; Section III presents a review of the literature on the supply-side effects of occupational licensing; Section IV describes the data used in my analysis; Section V outlines the identification strategy; Section VI presents the results and robustness checks; Section VII provides an analysis on the mechanism by which testing impacts the number of licenses issued; Section VIII extends the analysis by briefly looking at the effect of subject test laws on wages; Section IX presents a discussion of my findings.
II. Background and Motivation
Numerous reports released by the U.S. Department of Education have highlighted the importance of teacher quality. As President George W. Bush said at the 2006 State of the Union Address, “If we ensure that America’s children succeed in life, they will ensure that America succeeds in the world” (U.S. Department of Education, 2006). Over the past decade, federal and state governments have implemented a number of policies to ensure that teachers meet certain quality standards (U.S. Department of Education, 2006). These policies have included increased accountability of teacher preparation programs, tenure reforms, and increased use of teacher certification testing to demonstrate subject matter competency (U.S. Department of Education, 2010).
Currently, prospective candidates can receive initial teacher certification through two separate routes: the traditional route or the alternative route. The traditional teacher preparation program generally includes a four-year undergraduate degree in education. The alternative route to teacher certification, on the other hand, includes candidates who did not receive an undergraduate degree in education. Instead, these individuals are required to complete an alternative route preparation program that focuses on pedagogy instruction. While the requirements for initial certification vary across states, states generally do not differ in their required assessments for traditional and alternative route program completers (U.S. Department of Education, 2006).
States began to require tests for teacher licensure in the 1960s (Angrist & Guryan, 2008). Certification tests for teachers fall into three categories: basic skill tests, which assess reading, writing and general mathematics; professional knowledge tests, which assess pedagogical skills; and subject matter tests, which assess knowledge of a specific subject such as middle school social studies or high school physics. As a response to reports such as “A Nation at Risk” (Gardner, 1983) which advocated for educational policy reforms that increased teacher quality, the use of certification tests became more widespread (Larsen, 2015). The No Child Left Behind Act of 2001 further increased the prevalence of teacher testing as a signal for quality and content mastery. While the adoption of these testing requirements varies widely across states, the intuition behind required certification exams is the same: ensuring that teachers meet a minimum quality standard.
In recent years, improving the quality of teachers has turned out to be a challenge. Specifically, high public-school enrollment rates, increased teacher attrition and the retirement of baby boomers from the workforce have made high-quality teachers scarce (Boyd, Goldhaber, Lankford, & Wyckoff, 2007). At the same time, widespread media reports on state public school teacher shortages have made headlines (The Hamilton Project, 2017). A report from the Education Commission of the States highlights that teacher licensure requirements can affect a state’s ability to attract new teachers (Education Commission of the States, 2016).
States have approached the problems of teacher shortages and teacher quality in different ways. While some states have increased the stringency of certification requirements to ensure high-quality teachers, a number of states, such as Wisconsin, Iowa and Michigan, have relaxed certification test requirements as an attempt to deal with shortages and attract more teachers (Boyd et al., 2007; DeGrow, 2018; National Council of Teachers of English, 2017; Pfannenstiel, 2018).
One important issue in evaluating which of these two approaches is more effective is the existence and magnitude of a tradeoff between teacher quality and quantity. If more stringent certification requirements improve teacher quality and deter few people from entering the teaching profession, then increasing the stringency of licensing requirements would likely be net beneficial. However, if more stringent requirements have little or no effect on teacher quality and deter qualified individuals from entering the occupation due to the increased costs associated with meeting these requirements, then relaxing the stringency of licensing requirements would be better.
Empirically, evidence on the link between teacher testing requirements and teacher quality is mixed (Angrist & Guryan, 2004, 2008; Larsen, 2015) and this paper does not offer any new evidence on this question. But without substantial evidence on the supply-side effects of more stringent licensing requirements, the subject of this paper, the net effect of increasing stringency is unclear.
Public policy debates over the stringency of teacher certification requirements make the supply-side effects of more stringent licensing requirements in the teacher labor market particularly relevant. As further discussed in Section IV, I restrict my analysis to the 2002-2015 period due to limited data availability. Since basic skill tests and professional knowledge test requirements were relatively stable over my sample period, I choose to focus my analysis on subject test laws. As can be seen from Figure 1A and Figure 1B, the number of states requiring subject tests grew substantially over my sample period. I therefore decide to exploit the within-state variation in subject test law implementation to analyze the short- and medium- run supply-side effects of more stringent certification testing requirements. My methodology cannot study the longer-run effects which may be important in a full policy evaluation.
III. Literature Review
The existing literature on the supply-side effects of occupational licensing has approached the question from both the theoretical and empirical side. On the theoretical side, the arguments of Stigler (1971) and Peltzman (1976) on the effects of regulation center around the idea of self-interested supply and demand of regulation. Their arguments suggest that occupational licensing may restrict supply and decrease competition. In the case of teachers, Ballou and Podgursky (1998) argue that higher barriers to entry due to more stringent certification requirements may drive talented, high-quality individuals away from the teacher labor market. Their arguments suggest that more stringent licensing requirements may restrict supply, especially at the upper tail of quality in the teacher labor market.
While the aforementioned literature on occupational licensing is theoretical, the empirical strand of the literature on the supply-side effects of licensing is large and growing. The evidence on the supply-side effects of occupational licensing is mixed in the literature. On the one hand, Law and Marks (2017) and DePasquale and Stange (2016) find no evidence that licensing affects labor supply when looking at the labor market for nurses. On the other hand, a large number of studies find that occupational licensing actually lowers labor supply. Using a cross-sectional analysis, Hall and Pokharel (2016) find that the number of exams required to become a barber in a state is negatively related to the supply of barbers. By looking at the labor market for cosmetology, Adams, Jackson and Ekelund (2002) find evidence of a reduction in supply due to occupational licensing. Specifically, their findings suggest that more stringent statutes concerning cosmetology licensing requirements lead to higher average price and lower quantities consumed of those services. Cai and Kleiner (2016) find that occupational licensing is associated with a reduction in the supply of physical and occupational therapists. In addition, Jacob and Murray (2006) examine the effect of the implementation of a 150-hour education requirement for CPA licensure and find that the implementation of this requirement has been associated with a substantial decline in the number of CPA candidates. Similarly, using longitudinal data on a single cohort from the High School and Beyond (HSB) survey, Hanushek and Pace (1995) find that teacher training completion is significantly lowered by state requirements for courses and teacher tests, with no clear effect on teacher performance. However, Boyd et al. (2007) explain that the evidence presented by Hanushek and Pace (1995) should be interpreted with caution considering that their study evaluates the effect of differences in certification requirements at a single point in time, therefore limiting the internal and external validity of their analysis.
Consistent with the view that more stringent licensing requirements increase the barriers to entry into an occupation and restrict supply, several empirical studies have explored the effect of more stringent occupational licensing requirements on wages. Looking at the effect of licensing on wages of nurses, Law and Marks (2017) find that licensing requirements are associated with a wage premium. Similarly, Timmons and Thornton (2007) find a licensing wage premium for radiologic technologists. More generally, looking at the effect of occupational licensing across different occupations, Kleiner and Vorotnikov (2017) find that occupational licensing raises wages by about 11% after controlling for human capital and other observable characteristics. In the teacher labor market, Angrist and Guryan (2008) find that more stringent licensing requirements are associated with higher teacher pay. On the other hand, Larsen (2015) finds that more stringent licensing requirements have no significant effect on teacher wages.
This paper contributes to the existing literature on the supply-side effects of occupational licensing by exploring the effects of increasing licensing requirements in the teacher labor market[1]. Additionally, while the wage premium associated with more stringent certification requirements for teachers is assumed to be caused by a supply shift (Angrist & Guryan, 2008), the wage literature does not formally test this hypothesis. This paper attempts to address this gap in the teacher wage literature by evaluating the supply-side effects of increasing certification requirements.
IV. Data
To evaluate the effect of subject matter test requirements on the teacher supply, I construct an original state-by-year panel from 2002 to 2015 that combines measures from a range of datasets. I chose states as the unit of observation for several reasons. First, testing requirements for public school teachers are implemented at the state level. Second, I collect data on the number of initial teacher licenses issued by states for the period, providing me with a critical outcome variable. Third, I can use the state-level data to develop a fixed-effects analysis that allows me to examine how within-state changes in testing requirements affect new teacher supply.
New Teacher Supply
I would ideally measure state-level new teacher supply as the total number of individuals entering the teacher labor market each year. As such a measure is unavailable, I use data collected by the U.S. Department of Education under Title II requirements on the number of teachers receiving initial certification by state and year. The number of initial teacher licenses issued serves as a proxy for new teacher supply. I also collect data on the total number of students enrolled by state and year from the NCES. I construct my outcome variable as the ratio of the number of new initial teacher licenses issued per every 1,000 public school students so that the measure is comparable in small and large areas.
Subject Test Requirements
The data on subject test requirements from 2002-2010 were made available by Professor Bradley Larsen. I extend my sample size to 2015 by using data from the U.S. Department of Education National Center of Education Statistics (NCES). While the NCES provides information on subject test requirements, the data is not available for all states across my entire sample. In addition, the NCES does not provide the exact month in which subject tests were implemented. I fact-checked the NCES data and obtained months for these dates by referring to state laws and the Highly Qualified Teachers and Improving Teacher Quality State Grants (ESEA Title II Part A) monitoring reports. The monitoring reports come from U.S. Department of Education evaluations of states’ progress in meeting the No Child Left Behind Act’s requirements and provide useful information on the timing of subject matter test adoption. The start of the subject test policy period is defined as the fall of the academic year in which the policy was implemented, if within the first half of the academic year, or as the fall of the academic year following policy implementation, if implemented during the second half of the academic year.
State-Level Controls
I include a set of control variables in my model to account for state-specific changes in teacher benefits, testing requirements, and broader labor market and economic conditions. I collect data on average teacher salary and per-pupil expenditure from the NCES and adjust the measures to 1999 US dollars to control for real average teacher salary and real per-pupil expenditure. In addition to the NCES data on average teacher salary, I construct a second measure of average teacher salary by collapsing data on public school teachers’ income using repeated cross-sections of the Current Population Survey (CPS) Annual Social and Economic Supplement (ASEC). I adjust the CPS measure of average teacher salary to constant 1999 US dollars. I use the CPS measure of average teacher salary to complement the wage analysis in Section VIII. I include in my preferred model the two other types of testing requirements–basic skill tests and professional knowledge tests–to control for additional testing requirements that may have affected new teacher supply. These test requirements are represented by two binary variables for basic skill test and professional knowledge test. I compile these measures by combining a dataset made available by Professor Bradley Larsen with data from the NCES. Additionally, I collect state-level data on real GDP per capita from the Bureau of Economic Analysis, on population size from the U.S. Census Bureau, and on state unemployment rates from the Bureau of Labor Statistics.
Mechanism Analysis
In an attempt to better understand the mechanism by which subject test requirements affect teacher supply, I collect data from the U.S. Department of Education under Title II requirements on the pass rate on required assessments and on the number of alternative and traditional teacher preparation program completers. These measures are discussed in more detail in Section VII. Table 1 provides a summary of the descriptive statistics of the aforementioned variables.
Data Limitations
The U.S. Department of Education started collecting information on the number of initial teacher licenses issues under Title II requirements starting in 1999. Since the number of teachers receiving initial licensure is my main variable of interest, my sample is restricted in size. In addition, data on state testing requirements from the NCES are only consistently available starting in 2002, which further restricts my sample size. Consequently, within-state variation in subject testing laws implemented prior to 2002 is not captured by my model. In addition, the NCES data on basic skill test and professional knowledge test requirements, which are included as controls in my preferred specification presented in the next section, is not available for all states over my sample period. Due to the lack of data on these variables for some years and state, my panel will be unbalanced in the specifications that include these two controls.
V. Empirical Framework
Fixed Effects Model
To investigate the effect of required content knowledge examination on teacher supply, I estimate the following regression equation:
where for each in state s, in year t: Yst is the ratio of the number of new initial teacher licenses issued per every 1,000 public school students, Tst is a binary indicator that equals to 1 if a subject test was required in state s in year t, Xst is a vector of state-level control variables including state unemployment rate, a binary indicator that equals 1 if a professional knowledge test was required, a binary indicator that equals 1 if a basic skill test was required, real average teacher salary from the NCES, real per pupil expenditure, and real GDP per capita, is state fixed effects, is year fixed effects and represents unobservable factors. State fixed effects capture factors that vary across states but not over time (e.g., if some states have better quality of education and therefore attract more teachers across the entire sample). Year fixed effects capture factors that vary over time but not across states, such as the No Child Left Behind act that affected all states. The parameter of interest is , which is identified from variation in subject test laws within states across time. I weight my state-level regression by the population size of each state. Standard errors are clustered at the state level. Since the level of clustering is often unclear, I also conduct inference using a non-parametric permutation test as in Chetty, Looney, and Kroft (2009) and Hjort and Poulsen (2019).
The identification strategy presented above relies on two assumptions. The first is the same parallel trend assumption as in a difference-in-differences approach. Namely, the identification assumption is that absent the policy change, the trends in new teacher supply in the control and treatment states would have been similar. This assumption would be violated if, for instance, subject test policy reforms responded to trends in new teacher supply. I test this first assumption in what follows by conducting an event-study analysis to test for the presence of pre-trends. The second assumption is that is uncorrelated to factors that affect both state certification laws and new teacher supply in that state. To test the validity of this assumption, I conduct a robustness check that tests for alternative explanations.
Event Study Model
As mentioned above, the concern in the research design is that states might be changing subject test requirements as a response to changes in teacher supply. A common approach to address this issue is to determine whether the policy appears to have an effect on the outcome before it is actually implemented. The presence of differential pre-trends would suggest that the strict exogeneity assumption of the policy change is violated. I consider including state-specific linear time trends in the model presented above in order to control for pre-trends. However, as Wolfers (2006) mentions, in cases where policy changes are modeled by a dummy variable, the inclusion of state-specific linear time trends picks up not only differential pre-trends across states, but also differences in the evolution of the outcome between treated and control states following the policy change. Furthermore, this problem is exacerbated when only a few observations are available prior to the policy change (Wolfers, 2006). Since my sample begins in 2002 and the largest wave of subject test implementation happened shortly after, in 2004-2006, this leaves only a few observations with which to identify preexisting state trends. Wolfers (2006) demonstrates that these problems can contaminate estimates of the effect of a policy change. Therefore, I choose not to include state-specific linear time trends in my preferred model presented in equation (1). Instead, I test for the presence of pre-trends by using an event-study model which imposes less restrictions on the policy response dynamics (as in Kraft, Brunner, Dougherty, & Schwegman, 2018). In order to test for the presence of pre-trends that could lead to spurious findings, I use the following event-study model:
where for each state s, in year t: Yst is the ratio of the number of new initial teacher licenses issued per every 1,000 public school students, for k < 0 correspond to pre-trends, and for k ≥ 0—to dynamic effects k periods relative to the adoption of the subject test requirement, Xst is a vector of state-level control variables including state unemployment rate, a binary indicator that equals 1 if a professional knowledge test was required, a binary indicator that equals 1 if a basic skill test was required, real average teacher salary from the NCES, real per pupil expenditure, and real GDP per capita, is state-fixed effects and is year-fixed effects. States that do not experience a change in subject test requirements over my sample period are included as control states. I weight my event study model by the population size of each state. I cluster standard errors at the state level.
VI. Results
Fixed Effects Model
Theory would predict that higher barriers to entry into an occupation would lower labor supply. Within this line of reasoning, we should expect more stringent testing requirements to decrease the number of initial licenses issued by restricting the supply of new teachers. My estimates from model (1) are reported in Table 2. The first column estimates equation (1) by only including the subject matter test binary indicator. Column 3 includes the two other types of testing requirements. Column 4 includes state-level controls. Column 6 shows the full model, including state controls and other testing requirements. Since I am interested in estimating the effect of implementing a subject test requirement on new teacher supply within a state over time, all specifications include both state fixed effects and year fixed effects.
Because the professional knowledge test and the basic skill test variables have missing observations for some states and years over my sample period, the sample size decreases once these two controls are included. In order to show that the change in the coefficient of interest across the different specifications is not driven by changes in sample size, I re-estimate the same model presented in column 1 and column 4 in columns 2 and 5 respectively by using the restricted sample from the specifications that include the other testing variables. Doing so enables me to compare the coefficients from columns 2, 3, 5 and 6. The simplest form of my model for my overall sample, presented in column 1, suggests that subject matter test laws are associated with a 1.08 decrease in the number of teachers receiving initial licensure per 1,000 public school students.
As can be seen from comparing columns 2 and 3, the estimated impact of subject test reforms on new teacher supply is robust to controlling for other testing reforms simultaneously. Reassuringly, basic skill test and professional knowledge test requirements do not predict significant decreases in the supply of new teachers. As shown in column 3 of Table 2, only subject test reforms predict decreases in the number of teachers receiving initial licensure per 1,000 public school students. While controlling for all three types of testing requirements restricts the size of my sample due to lack of data availability, it allows me to better isolate the effect of subject test requirements from the effects of basic and professional knowledge test requirements.
The estimated impact of subject test requirements is also robust to the inclusion of state labor market and economic controls as shown in column 4. Other testing requirements and economic conditions do not appear to explain much of the decrease in new teacher supply, given the stability and significance of my estimates across all specifications with the same sample size. As shown in Table 2, the coefficient on the subject test variable is consistently negative and significant across all specifications. Column 6 of Table 2 presents my preferred model that controls for both testing requirements and economic conditions. Specifically, my results from column 6 suggest that, all else constant, implementing a content knowledge test law is associated with a 1.27 decrease in the number of initial teacher licenses issued per 1,000 public school students. This represents a 22% reduction in the average number of initial licenses issued per 1,000 public school students in the post-policy reform years among treated stated, relative to the average number of licenses issued by states across my entire panel. The magnitude of the coefficient implies that adopting a subject matter test requirement reduces the number of teachers receiving initial licensure per 1,000 public school students by 0.5 standard deviations.
Event-Study Model
To rule out the presence of pre-trends that may be driving the results found in the fixed effects model above, I now turn to my event-study model. The estimates from my event-study model are presented in Figure 2 and reported in Appendix Table 1. The results suggest that subject testing has a significant and negative effect on the number of initial teacher licenses per 1,000 public school students in the first year in which the test is implemented and at more than four years after the law change. In particular, the main effect of the law change appears to occur in the first year in which the subject test law is implemented.
This effect may arise from a decrease in the number of candidates meeting teacher requirements due to failure to meet the new testing standards. The second significantly negative effect that’s observed several years post-implementation is more difficult to interpret due to the fact that the effect seems to fade, then reappear. However, this effect could potentially be explained by announcements of certification requirement changes affecting undergraduates’ decision to enter the teaching profession. Most importantly, the estimates from the event-study model suggest that new teacher supply was not already decreasing prior to the implementation of subject test laws. The pre-treatment coefficients are positive and statistically insignificant. Therefore, I can rule out the concern that new teacher supply was trending downwards prior to the policy change.
Robustness Checks
To test the sensitivity of my results, I perform three separate robustness checks. First, I start by testing the sensitivity of my approach by logging rather than scaling new teacher supply. Appendix Table 2 presents the results of equation (1) with the logarithm of initial licenses issued as an outcome variable. As shown by Table A2, the coefficient on the subject test law variable is consistently negative and statistically significant at the 5% level under this alternative specification. Specifically, the estimates suggest that subject test laws lead to a 21% decrease in the number of initial teacher licenses issued.
Second, as mentioned in Chetty et al. (2009), a concern in difference-in-differences analysis is that serial correlation can bias standard errors and lead to an over-rejection of the null hypothesis of no effect. To address this concern, I follow Chetty et al. (2009) and Hjort and Poulsen (2019) and conduct a non-parametric permutation test for I do that by permuting the subject test variable across states 1,000 times and assigning randomly chosen subject test implementation dates to each state in my dataset. The permuted dataset preserves the real timing of subject test implementation but simply reassigns them randomly across my sample. Figure 3 presents the empirical distribution of estimates resulting from permuting the subject test variable 1,000 times and running equation (1) on the permuted datasets. The vertical line represents the true estimate found in Table 2 column 6. If subject test implementation has a significant effect on new teacher supply, we would expect the estimated coefficient to be in the lower tail of the distribution of placebo estimates. As seen in Figure 3, the true estimate is indeed in the lower tail of the distribution, with an implied p-value of 0.02. The permutation test therefore confirms that subject test requirements led to an unusually low level of new teacher supply.
Third, I test the validity of my exogeneity assumption by testing for an alternative explanation. Specifically, the decrease in the supply of new teachers could be explained by demand shocks in the public-school teacher labor market that were concurrent with the timing of subject test implementation. To test this alternative explanation, I run the model presented in equation (1) and replace the outcome with a measure of teacher demand. I calculate a proxy for teacher demand by dividing the total number of students enrolled in public school by the student-teacher ratio to get the total number of teachers needed to accommodate public school students (Aaronson & Meckel, 2009). In Appendix Table 3, I report the results from running equation (1) using the proxy for teacher demand as my outcome. The estimates shown in Table A3 suggest that my results were not driven by demand shocks in the public-school teacher labor market that were concurrent with subject test implementation.
VII. Mechanism Analysis
As the results presented above serve as substantial evidence of the effect of subject test requirements on new teacher supply, I now extend my analysis by attempting to tease out the mechanism through which this effect occurs. The results discussed in the previous section suggest two possible mechanisms through which subject test requirements reduce the number of teachers receiving initial licensure. Subject test laws could reduce the number of initial teacher licenses issued either because less candidates are passing the required tests or because less candidates are taking the test. These may have different implications for policy. If subject test requirements decrease the supply of new teachers by lowering the pass rate on required certification test, then if the tests are effective in assessing quality this would suggest that the policy is effectively ensuring that teachers who do not meet the minimum quality standard imposed by the subject test do not end up teaching in the classroom. On the other hand, if subject test requirements decrease the supply of new teachers by reducing the number of candidates who are taking the test and the people deterred from taking the tests are not systematically lower quality potential applicants, then this mechanism would suggest that subject test laws are deterring potentially qualified individuals from entering the teaching profession. While determining which of the two mechanisms is at play in the case of teacher testing laws is especially difficult due to lack of data availability, I attempt to explore these potential channels in what follows.
Mechanism I: Pass Rate
To evaluate whether subject test requirements reduce the supply of new teachers by decreasing the pass rate on required exams, I use the specification presented in equation (1) and replace the outcome variable by the pass rate on state-required certification tests.[1] The main data limitation associated with using pass rates to isolate this potential mechanism is the fact that pass rates data for pre- and post- subject test implementation is only available for states that required at least one of the two other types of certification test prior to subject test implementation, thus restricting the exploitable within-state variation in pass rates before and after the requirement was put in place. Additionally, following the 2008 reauthorization of the Higher Education Act, the Tile II reporting system improved the structure and guidance regarding the collection of pass rate data so pre-2008 data was reported less consistently.
The results from this specification are presented in Table 3. As shown in Table 3, while the coefficient on the subject test variable is negative as expected across all specifications, the statistical significance of the coefficient on the subject test law variable is not robust to the inclusion of state economic controls. Thus, the estimates from my preferred model in column 4 suggest that there is no evidence of a significant decrease in certification test pass rates following subject test implementation once we control for other testing requirements and state economic factors.
Mechanism II: Deterrence Effect
To evaluate whether subject test requirements affect the supply of new teachers by deterring individuals from entering the teaching profession, I use the specification presented in equation (1) and replace the outcome variable with the total number of teacher preparation program completers per 1,000 public school students. The total number of teacher preparation program completers is the sum of both traditional and alternative route program completers. Prospective teachers are expected to complete a teacher preparation program prior to certification testing and licensure. If teacher testing reduces the supply of new teachers by raising the barriers of entry and deterring individuals from entering the teacher profession, we should expect to find a reduction in the number of individuals completing teacher preparation programs. The Title II reporting system defines a program completer as an individual who has met all the requirements of a state-approved teacher preparation program, regardless of whether an individual has or has not passed state licensure tests. Therefore, using this measure allows me to isolate the deterrence effect from the effect of subject test pass rates. This approach, however, has limitations. Specifically, program completers may choose to teach in a state other than the one in which they complete their teacher preparation program. The number of teacher preparation program completers does not capture the deterrence effect associated with individuals who complete their preparation program in one state but obtain licensure in another state. These effects are probably small considering that states usually require prospective teachers to complete a state-accredited teacher preparation program, but some states have reciprocity agreements that enable out-of-state program completers to receive a license.
The results from this specification are presented in Table 4. As shown in Table 4, the coefficient on the subject test variable is consistently negative across all specifications. However, the coefficient on the subject test variable only becomes significant at the 10% level once broad economic and labor market controls are included. Specifically, my results from column 4 suggest that, all else constant, implementing a content knowledge test law is associated with a 3.67 decrease in the total number of teacher preparation program completers per 1,000 public school students. While the standard error on the point estimate is large, the 90% confidence interval does not include zero, suggesting a negative effect. Specifically, the magnitude of the point estimate implies that subject test requirements are associated with a 1.6 standard deviation decrease in the number of teacher preparation program completers per 1,000 public school students. Another interesting finding from this specification is that basic skill test requirements seem to also be associated with a significant decrease in the total number of teacher preparation program completers. The coefficient on the basic skill test variable is negative and significant at the 5% level.
As mentioned in Section II, prospective teachers can complete a teacher preparation program through two routes: the traditional route and the alternative route. If teacher testing affects teacher supply through the number of teacher preparation program completers, one should expect testing to have a larger deterrent effect on alternative program completers. Since alternative route teachers have completed a degree in a subject that’s not education, they have more alternative career paths. Higher entry barriers through required testing would be expected to especially deter candidates with skills that are valued in alternative labor markets from entering the teaching profession (Hanushek & Rivkin, 2010). To test this hypothesis, I re-run the above specification by replacing the outcome variable by the share of total program completers who have taken the alternative route. This specification estimates the effects of subject testing on the probability prospective teachers entered a preparation program through an alternative route. Table 5 presents the results of this specification. The results suggest that subject test requirements are associated with a 17-percentage point decrease in the share of teacher preparation program completers who have taken the alternative route. For instance, looking at Maine, my findings would suggest that subject test implementation is associated with a reduction in the share of alternative route completers from 34% on average pre-implementation to 17% on average post-implementation.
My mechanism analysis suggests that there is a potentially negative correlation between subject test requirements and the pass rate on required exams however the evidence is weak and statistically insignificant. My findings therefore suggest that the deterrence effect is the main way subject test requirements are reducing the supply of new teachers. In light of these results, it is puzzling why the immediate effect observed in the event-study is so large and this merits further research. In addition, a full assessment of the impact on teacher quality would require answering the question of whether the people deterred were self-screened lower quality or more randomly distributed across the quality spectrum.
VIII. Wage Analysis
The empirical evidence on the effect of teacher testing on teacher wages is mixed in the literature. While Angrist and Guryan (2008) find that testing requirements are associated with higher teacher wages, others such as Larsen (2015) find statistically insignificant effects on wages. The supply shift story would predict that an increase in the barrier to entry into the teacher profession reduces the supply of teachers and consequently increases teacher wages. To test the validity of previous findings and determine whether wages increase, I use the specification presented in equation (1) and replace the outcome variable with the natural logarithm of real average teacher wages. I test this hypothesis using two different measures of teacher wages. The first measure is the estimated average teacher wages from the NCES. The second measure of average teacher salary is constructed using the CPS.
The results from this specification are presented in Table 6. As can be seen from columns 1 and 2, while the coefficient on the subject test variable is positive as predicted, the coefficient is insignificant across both specifications. For the NCES wage measure, the 95% confidence interval for the estimate on the subject test variable is [-0.021; 0.029]. Therefore, the confidence interval suggests a very precisely estimated zero effect. For the CPS wage measure, the 95% confidence interval for the estimate on the subject test variable is [-0.011; 0.100]. The estimated effect of subject testing using the CPS measure of income is a less precisely estimated zero effect. The lack of precision in the estimate could be associated to the noise in CPS income reporting.
Thus, I find no evidence of a wage premium associated with licensing requirements. These results go against what the supply-shift story would predict. However, Hanushek and Rivkin (2010) explain how the institutional characteristics of the public-school teacher labor market such as tenure and unionization appear to have caused less variation in wages than would be found in more competitive markets. Thus, the absence of a wage premium associated with higher barriers to entry could be the result of the other wage-setting forces at play in the teacher labor market. Another possible explanation for the absence of a wage premium could be associated to the difference between the flow and stock of public-school teachers. My previous results suggest that subject test requirements decrease the flow of new teachers. However, I did not explore the relationship between subject test requirements and the stock of public-school teachers since a teacher who was licensed prior to test implementation is not required to take the test. Subject test requirements may have a large impact on the flow of new teachers but the impact on the stock will be more lagged. As a result, we should not expect to find a short-run wage premium associated with more stringent licensing requirements, but such a premium might show up over longer periods of time.
IX. Discussion
Since the early 2000s, states have been responding to concerns about teacher quality by increasing the stringency of licensing requirements. While proponents believe that certification test requirements guarantee teacher quality, opponents of occupational licensing argue that higher barriers to entry could deter qualified individuals from entering the profession (Ballou & Podgursky, 1998; Stigler ,1971; Peltzman, 1976). In this paper, I attempted to evaluate the barriers-to-entry argument by exploring the supply-side effects of subject test requirements on the labor market for teachers.
My investigation of the impact of subject test requirements suggests that subject test implementation reduces the number of initial teacher licenses issued by 22% relative to the sample mean. Consistent with the barriers to entry argument, the results from my mechanism analysis suggest that more stringent licensing requirements reduce the number of teacher preparation program completers and have no significant effect on the pass rate on required exams. Furthermore, my results suggest that subject test laws disproportionately deter individuals from entering the teaching profession through the alternative route compared to the traditional route. These findings are consistent with the view that higher barriers to entry particularly deter candidates who have skills that are valued in alternative labor markets (Hanushek & Rivkin, 2010). Finally, contrary to what the supply-shift story would predict, I find no evidence of a wage premium associated with subject test laws.
Before concluding, I must address some caveats of my study and avenues for future research. While this paper focused on the state-level supply-side effects of more stringent licensing requirements for teachers, future studies should seek out data at the district level. Existing empirical literature has explained how the effects of certification requirements for public school teachers vary across districts (Larsen, 2015), which suggests that more stringent licensing requirements may have heterogeneous effects across higher- and lower-income districts in the same state.
The difficulty of subject tests, in terms of content and cut scores, varies widely across states that have subject test requirements. While the Praxis II exam is the most commonly used, some states have developed their own content knowledge exams. The differential supply-side effects of different characteristics of subject tests, such as content difficulty and cut scores, is a potential area for future study.
Another caveat worth mentioning is the heterogeneous effects of subject test requirements across different subjects. We might expect more stringent licensing requirements to especially deter those who have majored in STEM subjects in which case alternative labor market options that value these skills are abundant. Due to lack of data availability on such a granular level, I am unable to test this hypothesis. However, future research might find a way to evaluate this subject-specific heterogeneity. This would be especially interesting considering the fact that most reported state shortages are subject-specific.
Despite these caveats, this paper sheds some light on how more stringent licensing requirements affect the supply of new teachers. Overall, my results are reasonably consistent with the view that testing has acted as a barrier to entry. From a policy perspective, my findings suggest that if there is little to no effect of increased licensing requirements on teacher quality, then relaxing testing requirements to attract more teachers could be net beneficial.
References
Aaronson, D., & Meckel, K. (2009). How Will Baby Boomer Retirements Affect Teacher Labor Markets? Economic Perspectives, 33(4), 2-15.
Adams, A.F., Jackson, J.D. & Ekelund, R.B. (2002). Occupational licensing in a “competitive” labor market: The case of cosmetology. Journal of Labor Research, 23(2), 261-278
Angrist, J. D., & Guryan, J. (2004). Teacher Testing, Teacher Education, and Teacher Characteristics. American Economic Review, 94(2), 241-246.
Angrist, J. D., & Guryan, J. (2008). Does teacher testing raise teacher quality? Evidence from state certification requirements. Economics of Education Review, 27(5), 483-503.
Ballou, D., & Podgursky, M. (1998). The case against teacher certification. Public Interest, 17-29.
Berger, M. C., & Toma, E. F. (1994). Variation in State Education Policies and Effects on Student Performance. Journal of Policy Analysis and Management, 13(3), 477.
Boyd, D., Goldhaber, D. D., Lankford, H., & Wyckoff, J. H. (2007). The Effect of Certification and Preparation on Teacher Quality. The Future of Children, 17(1), 45-68.
Cai, J., & Kleiner, M. M. (2016). The Labor Market Consequences of Regulating Similar Occupations: The Licensing of Occupational and Physical Therapists. Upjohn Institute Working Paper 16-259.
Chetty, R., Looney, A., & Kroft, K. (2009). Salience and Taxation: Theory and Evidence. American Economic Review, 99(4), 1145-1177. doi:10.1257/aer.99.4.1145
DeGrow, B. (2018). New Law Removes Obstacles to Classroom. Retrieved from http://www.mackinac.org/new-law-removes-obstacles-to-classroom
Depasquale, C., & Stange, K. (2016). Labor Supply Effects of Occupational Regulation: Evidence from the Nurse Licensure Compact. NBER Working Paper No. 22344
Education Commission of the States. (2016). Teacher Shortages: What We Know. Retrieved from https://www.ecs.org/wp-content/uploads/Teacher-Shortages-What-We-Know.pdf
Gardner, D. (1983): A Nation At Risk. The National Commission on Excellence in Education, Washington, DC.
Goldhaber, D. D., & Brewer, D. J. (2000). Does Teacher Certification Matter? High School Teacher Certification Status and Student Achievement. Educational Evaluation and Policy Analysis, 22(2), 129.
The Hamilton Project. (2017). Understanding and Addressing Teacher Shortages in the United States. Policy Brief 2017-05.
Hall, J. C., & Pokharel S. B. (2016). Barber Licensure and the supply of Barber Shops: Evidence from U.S. States. Cato Journal, 36(3).
Hanushek, E. A., & Pace, R. R. (1995). Who chooses to teach (and why)? Economics of Education Review, 14(2), 101-117
Hanushek, E. A., & Rivkin, S. G. (2010). The Quality and Distribution of Teachers under the No Child Left Behind Act. Journal of Economic Perspectives, 24(3), 133-150. doi:10.1257/jep.24.3.133
Hjort, J., & Poulsen, J. (2019). The Arrival of Fast Internet and Employment in Africa. American Economic Review, 109(3), 1032-1079. doi:10.1257/aer.20161385
Jacob, J., & Murray, D. (2006). Supply-side effects of the 150-hour educational requirement for CPA licensure. Journal of Regulatory Economics, 30(2), 159-178.
Kleiner, M., & Krueger, A. (2013). Analyzing the Extent and Influence of Occupational Licensing on the Labor Market. Journal of Labor Economics, 31(2), 173-202.
Kleiner, M. (1990). Are There Economic Rents for More Restrictive Occupational Licensing Practices. Industrial Relations Research Association Proceedings, 177–85.
Kleiner, M., & Petree, D. (1988). Unionism and licensing of public school teachers: Impact on wages and educational output. in When Public Sector Workers Unionize, ed. by R. B. Freeman, and C. Ichniowski. University Of Chicago Press.
Kleiner, M. M., & Vorotnikov, E. (2017). Analyzing occupational licensing among the states. Journal of Regulatory Economics, 52(2), 132-158
Kraft, M.A., Dougherty, S.M., Brunner, E.J., & Schwegman, D. (2018). Teacher accountability reforms and the supply of new teachers. Working Paper.
Larsen, B. (2015). Occupational Licensing and Quality: Distributional and Heterogeneous Effects in the Teaching Profession. SSRN Electronic Journal.
Law, M. T., & Marks, M. S. (2017). The Labor-Market Effects of Occupational Licensing Laws in Nursing. Industrial Relations: A Journal of Economy and Society, 56(4), 640-661.
Leland, H. E. (1979). Quacks, Lemons, and Licensing: A Theory of Minimum Quality Standards. Journal of Political Economy, 87(6), 1328-1346.
National Council of Teachers of English (NCTE). (2017). Teacher Shortage in Wisconsin Results in Softening of Teacher Certification Requirements. Retrieved from http://www2.ncte.org/report/teacher-shortage-wisconsin-results-softening-teacher-certification-requirements/
Peltzman, S. (1976). Toward a More General Theory of Regulation. Journal of Law and
Economics, 19, 211-240.
Pfannenstiel, B. (2018, March 07). Teacher licensing tests would be nixed under bill approved by Iowa House. Retrieved from https://www.desmoinesregister.com/story/news/politics/2018/03/06/teacher-licensing-tests-would-nixed-under-bill-approved-iowa-house/401043002/
Shapiro, C. (1986). Investment, Moral Hazard, and Occupational Licensing. The Review of Economic Studies, 53(5).
Stigler, G. (1971). The theory of economic regulation. Bell Journal of Economics and Management Science, II, 3-21.
Timmons, E. J., & Thornton, R. J. (2007). The Effects of Licensing on the Wages of Radiologic Technologists. Journal of Labor Research, 29(4), 333-346.
U.S. Department of Education, Office of Postsecondary Education. (2006). The Secretary’s Fifth Annual Report on Teacher Quality: A Highly Qualified Teacher in Every Classroom, Washington, D.C., 2006.
U.S. Department of Education, Office of Planning, Evaluation and Policy Development, Policy and Program Studies Service. (2010). Recent Trends in Mean Scores and Characteristics of Test-Takers on Praxis II Licensure Tests. Washington, D.C., 2010.
Wolfers, J. (2006). Did Unilateral Divorce Laws Raise Divorce Rates? A Reconciliation and New Results. American Economic Review, 96(5), 1802-1820. doi:10.1257/aer.96.5.1802