Clay turner dissertation

Doctoral Dissertation Three Essays on the Economics of Education A dissertation submitted to the School of Public Policy and Management, H. John Heinz III College and Tepper School of Business Carnegie Mellon University in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Economics and Public Policy by Abby Clay Turner Dissertation Committee: Lowell Taylor, Carnegie Mellon University Maria Marta Ferreyra, Carnegie Mellon University Dennis Epple, Carnegie Mellon University David Deming, Harvard University August 14, 2013 1 Introduction This thesis is composed of three essays on the economics of education. Chapter 1 analyzes the effectiveness of a battery of formative assessments, 4Sight, which are broadly aligned with annual assessments required by NCLB. These formative assessments are designed to provide teachers with feedback on student performance throughout the year in order to raise end of year student test scores.

Methodologies for evaluating the effect of using 4Sight on test score outcomes are drawn from the program evaluation literature, and include individual and school- level OLS, quartile regressions, and probit regressions, as well as matching at the school-level. Micro-econometric results show that 4Sight had no discernible effect on math scores and a small negative effect on reading scores in its first year in Pennsylvania. Policy recommendations include continuing a smaller-scale trial period of 4Sight for several more years in conjunction with careful, improved alignment between 4Sight and PSSA, ongoing empirical analysis of its effects, and incorporating more opportunity for feedback to teachers and students to improve 4Sight as a formative assessment. Perhaps the most surprising finding of this paper is that while 4Sight has a very small effect on student outcomes, students receiving tutoring under NCLB are significantly and between four and nine percent less likely to pass their exams, even after controlling for all covariates, including eligibility for tutoring.

This implies that the publicly- funded tutoring is actually disadvantaging those students who receive it. These results should serve as a cautionary example to school, district, and state-level policy-makers when choosing interventions designed to improve student performance on tests tied to NCLB. Teacher strikes and the right of public employees to collectively bargain are topics of frequent and heated debate in the public sphere, with little research available to inform the debate. In firms, the negative relationship between labor unrest and reduced productivity is well- documented; the purpose of this study is to explore whether there exists a similar, measurable relationship between labor strife and productivity in public schools.

In Chapter 2, I use regression analysis to analyze data that includes teacher strikes and expired contracts over a seven-year period in Pennsylvania, and I find that the pass rates on a district-level cohort’s math tests decrease by about 1-2% in the year of a strike and by about 0.5% during a year that teachers work under an expired contract. Additionally, cohorts experiencing a strike during their 11th- grade year realize about a 2% decrease in their graduation rate. In addition to improving upon 2 the methodologies of previous teacher strike papers, this paper distinguishes between productivity loss due to strikes and that due to lengthy ongoing labor disputes that do not necessarily end in strike. Policy implications include making administrators aware of the possible effect of a strike on graduation rates and the need for better collection of data on collective bargaining by state agencies.

The past decade has seen enormous growth in the for-profit higher education industry, and along with it, enormous debate over the relative costs and benefits of such an education. Utilizing the rich data from the NLSY97 Geocode merged with institutional data from IPEDS, in Chapter 3 I empirically analyze data on individuals with two-year degrees, estimate the average marginal earnings gain from a two-year degree, and compare the effects of degrees across institutional sector and across major area of study using OLS with family background and extensive demographic controls. I find evidence of selection at three levels: selection into college, selection into type of college, and selection into major area of study. Any estimates of labor market returns to these degrees will be biased until future research unravels and models these selection mechanisms and processes.

This chapter provides a first look into the differential inputs and outputs of for-profit and public two-year degree programs. I find that a two-year degree results in an 8.1 percent average marginal earnings gain over a high-school diploma, and that the sector of the degree-granting institution alone does affect this gain. I also find that earnings gains vary greatly by major; individuals with “academic” degrees experience no significant earnings gains while individuals with “vocational/technical” degrees on average experience a 32.7 percent earnings gain. I find statistical differences in the marginal earnings gains across institutional sector within major fields of study, suggesting that attending a for- profit does matter when major field of study is taken into account.

Policy-makers should take note that this preliminary analysis of the returns (which can be thought of as the private benefits) to public and for-profit degrees does not provide unambiguous evidence in favor of one sector over another, but rather a first look into the “black box” in which students, institutions, and major areas of study come together and jointly determine labor market outcomes. 3 Acknowledgements Chapter 1 utilizes confidential student achievement information obtained under a signed confidentiality agreement with the Pennsylvania Department of Education. The authors wish to thank Jinxiang Liu for programming and database support. The authors benefited from discussions with Maria Ferreyra, Margaret McMacken, John Garrow, Dave Davare, Brian Junker, participants at the October 3, 2008 seminar on applied microeconomics at the Tepper School, participants at the May 23, 2009 seminar at Heinz College, and comments received at the 2009 Annual Meeting of the American Economic Finance Association and the 2009 Summer Meeting of the North American Econometric Society.

Financial support from the Heinz Endowments and the William Penn Foundation is gratefully acknowledged. Responsibility for the findings and interpretation of this paper rests solely with the authors. Regarding Chapter 2, I would like to thank Lowell Taylor, Maria Marta Ferrerya, and Mel Stephens, Heinz and Tepper seminar participants, Melanie Zilora, Billie Morrow Davis, John Gardner, and Abby Linn. I also thank Dave Davare, Anne Herald, and the staff of the PSBA.

Chapter 3 has benefited from the advice of Lowell Taylor, Melanie Zilora, Billie Morrow Davis, Aaron Turner, my committee, and comments received at my dissertation proposal. I also thank Stephanie Cellini and Latika Chaudhary for their help in replicating their paper. I gratefully invite the comments and feedback of any readers of this version. This research was conducted with restricted access to Bureau of Labor Statistics (BLS) data.

The views expressed here do not necessarily reflect the views of the BLS. 4 Table of Contents Three Essays on the Economics of Education .4 Table of Contents .5 List of Tables .6 List of Figures .8 Chapter 1: With the Best of Intentions: School-Level Interventions and the Pursuit of Proficiency .1 Introduction and Research Questions .2 Some Background on Student Achievement Testing and Testing in Pennsylvania .3 Overview of 4Sight and its Use in Pennsylvania .4 Evaluation Methodology and Data Requirements .6 Conclusions, Implications, and Future Research .8 Tables and Figures. 59 Chapter 2: Labor Strife in Public Schools: Does it Affect Education Production?.7 Conclusions and Policy Implications .9 Tables and Figures. 99 Chapter 3: Selection Bias and For-Profit Associate Degrees: Evidence from the NLSY Geocode .6 Replication of Cellini and Chaudhary (2012) .9 Tables and Figures.

128 5 List of Tables Table 1.1: Student-Level Descriptive Statistics by 4Sight Math Use .2: Mean Student-Level 2005-06 PSSA Scaled Scores by 4Sight Math Use .3: Student Descriptive Statistics by 4Sight Reading Use .4: Mean Student-Level 2005-06 PSSA Scaled Scores by 4Sight Reading Use .5: School-Level Descriptive Statistics by 4Sight Math Use .6: Mean School--Level 2005-06 PSSA Scaled Scores by 4Sight Math Use .7: School-Level Descriptive Statistics by 4Sight Reading Use .8: Mean School-Level 2005-06 PSSA Scaled Scores by 4Sight Reading Use .10: Individual-Level OLS Results: Math.11: Individual-Level OLS Results: Reading .12: Individual-Level Quartile Regressions: Math .13: Individual-Level Quartile Regressions: Reading.14: Individual-Level Probit Results: Math .15: Individual-Level Probit Results: Reading .16: Individual-Levle Tobit Regression Results: .17: Individual-Level Tobit Regression Results: .18: Individual-Level Tobit Regression Results: Math Reporting Category with Controls .19: Individual-Level Tobit Regression Results: .20: School-Level OLS Results: Math .21: School-Level OLS Results: Reading .22: School-Level Nearest-Neighbor Matching Model Results: ATT of Math 4Sight on Mean Math PSSA Scaled Scores .23: School-Level Nearest-Neighbor Matching Model Results: ATT of Reading 4Sight on Mean Reading PSSA Scaled Scores .1: District-Level Math PSSA Pass Rate Descriptive Statistics.2: District-Level Reading PSSA Pass Rate Descriptive Statistics .3: Descriptive Statistics: Striking vs. Non-Striking Districts .4: Percent of Districts with Expired Contracts or Striking Each Year .5: Cohort Regression Results: 8th-11th Grade‡ .6: Regression Results: Graduation Rate† .7: Cohort Regression for One-Year First Differences, 4th-8th Grades† .2: Descriptive Statistics for Two-Year Completers, by Institutional Status .3: Outcome Descriptive Statistics, Age 25+ .4: Outcome Descriptive Statistics for Two-Year Completers, by Institution Type, Age 25+ .5: Area of Major for Two-Year Completers, by Institution Type .6: Selected Descriptive Statistics for Two-Year Completers, by Major and Institution Type .7: OLS Regression Results: Effects on Log Weekly Earnings .8: OLS Regression Results: Effects on Other Outcomes .9: Two-Year Completers: Industry Working In At Age 26, by Major and Institution Type.10: Descriptive Statistics for Fixed-Effects Analysis .11: Pre- and Post-Education Summary Statistics .12: Fixed-Effects Analysis: Returns to College Attendance, Log Weekly Earningsa .13: Fixed-Effects Analysis: Returns to College Attendance and Completion, Log Weekly Earningsa .14: Fixed-Effects Analysis: Effects of College Attendance on Other Labor Market Outcomesa .15: Fixed-Effects Analysis: Effects of College Attendance and Completion on Other Labor Market Outcomesa .16: Fixed-Effects Analysis: Returns to College Attendance by Major, Log Weekly Earnings .17: Fixed-Effects Analysis: Returns to College Attendance and Completion by Major, Log Weekly Earnings .144 7 List of Figures Figure 1.1: Propensity Scores for Math 4Sight Use† .2: Propensity Scores for Reading 4Sight Use† .1: Time-Trend in Math PSSA Pass Rates.2: Time Trend in Reading PSSA Pass Rates .3: Map of Striking Districts .4: Test Score Data Available, By Cohort. 95 8 Chapter 1: With the Best of Intentions: School-Level Interventions and the Pursuit of Proficiency with Robert P. Strauss Abstract This paper analyzes the effectiveness of a battery of formative assessments, 4Sight, which are broadly aligned with annual assessments required by NCLB.

These formative assessments are designed to provide teachers with feedback on student performance throughout the year in order to raise end of year student test scores. Methodologies for evaluating the effect of using 4Sight on test score outcomes are drawn from the program evaluation literature, and include individual and school-level OLS, quartile regressions, and probit regressions, as well as matching at the school-level. Micro-econometric results show that 4Sight had no discernible effect on math scores and a small negative effect on reading scores in its first year in Pennsylvania. Policy recommendations include continuing a smaller-scale trial period of 4Sight for several more years in conjunction with careful, improved alignment between 4Sight and PSSA, ongoing empirical analysis of its effects, and incorporating more opportunity for feedback to teachers and students to improve 4Sight as a formative assessment.

Perhaps the most surprising finding of this paper is that while 4Sight has a very small effect on student outcomes, students receiving tutoring under NCLB are significantly and between four and nine percent less likely to pass their exams, even after controlling for all covariates, including eligibility for tutoring. This implies that the publicly-funded tutoring is actually disadvantaging those students who receive it. These results should serve as a cautionary example to school, district, and state- level policy-makers when choosing interventions designed to improve student performance on tests tied to NCLB.1 Introduction and Research Questions The enactment of the No Child Left Behind (NCLB) Act in 2002 has hastened the spread of a culture of testing in the K – 12 public schools of our nation.