Many SEAs collect teacher-level data on compensation as well as on a variety of other teacher characteristics (e.g., age, experience, demographic characteristics). However, not all states collect such data. Moreover, while such data are generally more accurate than household survey data, if some variables are not routinely audited, they may contain reporting errors. In addition, there may be differences in the way states collect data on certain variables, how those variables are defined or created, and what data are included in data files. Thus, it is useful to consider alternative sources of administrative teacher data.
One such alternative that might be used to construct measures of teacher compensation, especially in cases where SEAs do not collect teacher-level data, is teacher pension fund data. Therefore this section provides a brief examination of the feasibility of using such data in teacher labor market studies. In theory, teacher pension funds should track at least some of the variables collected in the SEA teacher compensation data. Since teacher pensions are tied to earned salaries, associated administrative data must also track teacher earnings. Moreover, because benefit formulas use experience to determine retirement benefits, some measure of tenure must be collected as well. For this report, several large state pension funds were surveyed about their data collection. Six states were chosen either because they participated in the CCD feasibility study but not the telephone discussion (Alabama and Washington), because they have pension systems specifically for teachers and other K–12 professionals (Arkansas and Missouri), or because their pension systems cover public employees other than teachers (Colorado and New Jersey).32
The administrators of these pension funds were asked about their data systems to determine whether they contained elements similar to those kept by SEAs. In addition to teaching experience and education background, the administrators were asked about job titles (i.e., whether teachers can be distinguished from other positions such as principals or central office administrators), total and base salary (as defined above), and length of contract. They were also asked whether they are willing to make their data available for research.
As compared to the data collected by SEAs, the results of the conversations with pension fund administrators suggested a number of significant limitations to the pension data as a data source for the type of research examined in this report. This is not surprising, given that pension fund data are collected for a relatively narrow purpose—providing retirement payments—and not the much broader purposes of the SEAs. For example, two of the pension funds did not collect data on job titles. None of the pension administrators reported that they collect information on teacher education levels, while only three reported collecting a measure of actual teaching experience.33
The type of salary is reported differently within and across state pension funds. Since state statute and local operating procedures affect salary and contract structures for both certificated and noncertificated staff, type of salary and length of contract are reported in many ways. For example, in Arkansas and New Jersey, pension funds collect data only on base salary (i.e., the amount specified in the contract). By contrast, in Colorado and Missouri, "includable salary" is legally defined and corresponding statutes prescribe the information required from employers. In Alabama, salary in the school system is defined by the county (and, thus, is not uniform within the state), and in Washington, the statutory definitions of salary for the purpose of computing pension annuities are complex and do not correspond well to the definitions of total, base, and supplemental pay. In short, direct "apples-to-apples" comparisons of teacher pay may not be possible. Finally, because pension systems often report to independent boards, they are not necessarily obligated to share data. All six pension systems indicated that they would share data depending on the type of request.
The analysis of this small sample suggests that, from a research perspective, data on teacher characteristics and pay from pension systems are likely inferior to data from systems maintained by SEAs. However, there is at least one interesting research question to which these data may contribute. Many state pension systems allow educators to return to teaching, either part or full time, after retirement (sometimes termed "double dipping"), while still collecting pension benefits. Merging pension fund data with SEA data may allow researchers to expand measures of total compensation to combine pension benefits with current teaching salaries for a more complete measure of total compensation.