**Year**2023**NSF Noyce Award #**1950292**First Name**Li**Last Name**Feng**Discipline**Other: Economics**Co-PI(s)**Mike Hansen, Brookings Institution; David Kumar, Florida Atlantic University; Hunter Close, Texas State University; Maria Fernandez, Florida International University; Ann Cavallo and David Sparks, University of Texas at Arlington; John Pecore, University of West Florida

**Presenters**Jieon Shim, Li Feng

## Need

During the COVID-19 pandemic, the national teacher shortage worsened, and the biggest hit was in high-poverty regions. This project is important that we investigate if the compensation or bonus could drive teachers to work in high-poverty and/or high-minority schools.

## Goals

The major research question is estimating the extra salary or bonus that would attract teachers to work in schools or school districts that have a challenging student population. Three specific research questions can be derived from this question: 1. Does the current teacher-pay schedule compensate teachers to teach in schools with a larger share of the low-income and disadvantaged minority students’ population? 2. How much do we need to compensate teachers to teach in these schools? 3. How much additional money do we need to pay teachers for them to stay at a high-poverty school versus an average school?

## Approach

The methodology we utilize to test if the current teacher-pay schedule compensates teachers for working in high-poverty or high-minority schools is the hedonic wage regression model. The following equation shows the initial specification. log(W_ijkt )= β_0+β_1 T_ijt+β_2 S_jt+β_3 D_kt+β_4 X_ijkt+ε_ijkt The outcome of interest is the teacher annual base salary in nominal terms denoted as W. It represents teacher i working in school j at year t. Control variables are teacher characteristics Tijkt, school characteristics Sjt, and school district characteristics Dk. Student body characteristics Xijkt are important that they include our parameters of interest. The parameters of interest are the coefficients in front of the variable that describes the proportion of free lunch eligible students at a school, FRPLijkt, and the variable that describes the proportion of non-white students at a school, Minorityijkt. We predict that β4 will be statistically significant and positive and that there will be positive compensating differentials. εijkt is a random disturbance term.

## Outcomes

We apply the hedonic wage regression model to examine if the teacher-pay schedule compensates teachers working at schools with a higher share of low-income and/or disadvantaged minority students’ population. The key variables of interest are the free lunch eligible student percentage and the minority student percentage. The first model shows the ordinary least squares model with only free lunch eligible student percentage, and the second model includes state and year fixed effects. On average, if a teacher works in a school with one percentage point higher free lunch eligible student population share, the log base salary decreases by 0.031, when keeping other variables constant. This effect is greater in the model with a fixed effect. The third model illustrates the ordinary least squares model with only minority student percentages, and the fourth model includes state and year fixed effects. We find that, on average, a teacher’s log base salary increases by 0.097 when working at a school with one percentage point higher share of minority students. The coefficient is lower in Model 4 with state and year fixed effects. Furthermore, we investigate the interaction effect between free lunch eligible student percentage and minority student percentage on log base salary. The correlation between free lunch eligible student percentage and minority student percentage is 0.627 showing some positive correlation. The ordinary least squares regression in Model 1 shows a negative association between free lunch eligible student percentage and log base salary, and a positive association between minority student percentage and log base salary. Including state and year fixed effects produces a similar result. When we interact the two continuous variables, the positive coefficient on the interaction term shows that the higher free lunch eligible student percentage, the higher the effect of minority student percentage on log base salary. Similarly, the higher minority student percentage, the higher the effect of free lunch eligible student percentage on log base salary. Adding state and fixed effects in Model 4 leads the coefficient of the interaction term to not be significant. In addition, we split the sample and compare schools in rural areas and schools in urban, suburban, or town. The coefficients for the free lunch eligible percentage and minority student percentage in the first column describing schools in rural areas correspond to negative compensation wage differentials for working in schools with a higher poverty percentage and positive compensation wage differentials for working in schools with a higher minority student percentage in the previous results. We find no significance in the coefficient for free lunch eligible student percentage representing schools in urban, suburban, or town. This result implies that the negative compensation wage differentials for working at a higher percentage of poverty schools come from the negative compensation wage differentials at schools located in rural areas. Therefore, we find that a wage penalty exists at high-poverty schools in rural areas, and that continuous disadvantage in compensation might lead to more inequality. Limiting teachers with STEM bachelor’s degrees gives a similar result. Positive compensating wage differentials are found for working at schools with a higher percentage of the non-white student population. In STEM fields also there exists negative compensating wage differentials in working at schools with a higher percentage of low-income students in rural areas. Since STEM fields lack teachers, the wage penalty in high-poverty schools in rural areas might worsen the teacher shortage. The coefficient is not significant for the coefficients for schools in urban, suburban, or town.

## Broader Impacts

In this paper, we provide unique and historical perspectives on how teachers’ salaries have changed over the last 30 years and whether we observe more or less compensating wage differentials for teachers teaching in more challenging schools or school districts. We find that positive compensating wage differentials exist for teaching a higher percentage of the minority student population. That is, a wage premium exists for teaching at schools with a higher percentage of the disadvantaged non-white student population. However, negative compensating wage differentials exist for teaching a higher percentage of the low-income student population. When we split the sample into schools in rural areas and schools in urban, suburban, or town, the negative compensating wage differentials are significant in rural areas. This dominant negative compensation wage differentials for teaching at a higher poverty school in rural areas leads to the negative compensation wage differentials in the overall sample.