1 Late Career Teacher RetentionDongwoo Kim Cory Koedel Shawn Ni Michael Podgursky Weiwei Wu Department of Economics, University of Missouri CALDER Annual Conference Washington, DC Feb 3, 2016 The authors wish to thank the Missouri Department of Elementary and Secondary Education for allowing use of their teacher data, and the John and Laura Arnold Foundation and the Center for Analysis of Longitudinal Data in Education Research (CALDER) for research support. The usual disclaimers apply.
2 Overview This project uses administrative data from two states to analyze effect of current and alternative pension plan designs on labor supply and workforce quality. Findings today STEM teachers in MO Evaluate fit of simulations using structural estimates of teacher preferences Simulate effect of reducing late career “push” incentives for high quality / high need teachers Selective retention bonuses Selective DROP plans
3 Overview Rising Costs for Teacher PensionsTeachers generally, and high performing teachers in particular, seem somewhat unresponsive to price incentives. It’s difficult to get high quality teachers to move to low performing schools (Glazerman, et. al. 2012) and they are expensive to retain ( Clotfelter, et. al., 2006).
4 Source: Costrell. http://www. uaedreform
5 Source: Costrell. http://www. uaedreform
6 Overview Final Average Salary DB plans Pension Plan Rules MO TennesseePSRS Replacement Factor 2.5 Exp ≤ 30, 2.55 Exp 31+ 1.5 Eligibility - Regular Age 60, Exp 30, A+E=80 Age 62, Exp=30 Eligibility -Early Exp 25 Exp 25 (age penalty) Social Security No Yes The table above presents key parameters of the Missouri and Other State DB plans. (The Other State rules are for incumbent teachers, not new hires.) An important difference is that Other State teachers are in Social Security (as teachers) and Missouri teachers are not.
7 Overview Accrual of Pension Wealth by Age
8 Overview Considerable upside potential Mean Retirement AgeBoth states are below the average for other college-educated professionals.
9 Overview Turnover of young teachers not out of line with similar professions (Harris and Adams, 2007) Exit of senior teachers is higher, and exit rates are responsive to pension plan incentives Furgeson, Strauss, Vogt (2006) Brown (2009) Costrell & McGee (2010) Ni and Podgursky (2016) Knapp, et. al. (2016) On the other hand, a small but growing literature finds that teachers are responsive to retirement plan incentives in the timing of their retirement. Retirement rates tend to spike at predictable points (age, experience) defined by rules of the pension system – “25 and out”, “rule of 80”, age 60). When these rules change the spikes move in response. This suggest that a more promising (and cost effective) strategy for increasing exposure of low performing students to high quality teachers is to postpone retirement for targeted groups of highly effective (or STEM) teachers. That is, rather than entice high performing teachers to move to high need schools, entice the better teachers who are already at those schools to delay retirement.
10 Why a structural modeling approach?Large and capricious variation in age/experience retirement incentives in current DB plans permits identification of behavioral parameters Effects of many policy-relevant reforms will not be seen for decades Temporary pilot programs will understate long-term effects Cooperation of pension plan administrators is lacking No “regulatory space” for experimentation Exception: charter schools in some states
11 Model basics (Stock-Wise model)In each year t, teacher chooses m (retirement year) to maximize expected utility Future is uncertain and retirement is irreversible. New information and circumstances arrive each year. Continuing to work and postponing retirement leaves open the option to retire in the future Dynamic “option value” model Estimated on full panel of MO teachers Ni and Podgursky (2016) Structural models used to forecast MO STEM retirements 2131 STEM teachers aged 58-65, exp ≥ 5
12 Model Basics (Stock-Wise Model)The period t expected utility of retiring in period m: 𝐸 𝑡 𝑉 𝑡 (𝑚)= 𝐸 𝑡 𝑠=𝑡 𝑚−1 𝛽 𝑠−𝑡 𝑘 𝑠 1−𝑐 𝑌 𝑠 𝛾 + 𝜔 𝑠 + 𝑠=𝑚 𝑇 𝛽 𝑠−𝑡 𝐵 𝑠 𝛾 + 𝜉 𝑠 denote: 𝑣 𝑠 = 𝜔 𝑠 − 𝜉 𝑠 and 𝜀 𝑠 = 𝜀 𝜔𝑠 − 𝜀 𝜉𝑠 ; assume: 𝑣 𝑠 = 𝜌𝑣 𝑠−1 + 𝜀 𝑠 ; 𝛽=0.965 𝛾=0.716 𝜌=0.643 𝜅=0.640 𝜅 1 =0.976, 𝑘 𝑠 =𝜅(60/age) 𝜅 1 σ=5000.0 The parameter estimates are obtained using PSRS data for (Ni and Podgursky, 2015) with a more precise measure of retirement status. The simulations reported below are out of sample in two ways: For PSRS it’s a different group of teachers in For Other State it’s both a different group of teachers in a different time period, and a different set of pension rules. The sigma parameter reflects the unobserved preference shock proportional to salary and is scaled up from the PSRS estimate (sigma=3660 there) to reflect an increase in salary and measurement error. Beta is the subjective discount rate, gamma captures risk-aversion and intertemporal substitution, kappa is the disutility of working. Estimated on sample of MO teachers For details see Ni and Podgursky (2015)
13 Missouri STEM Employment Survival FunctionConfidence bands commonly seen in other studies concern the standard errors of the estimator. The band in the plot assume no errors in parameters and only pertains to sampling errors stem from the draws of preference errors for each teacher. We assume aggregation of the retirement decision for each draw of the preference error for each teacher yields a survival rate. Accounting for standard errors in the parameter should make the band much wider.
14 Missouri STEM Age Distribution Retirees
15 Missouri STEM Age Distribution Non-Retirees
16 Policy Simulation DB plans have powerful “pull” “push” incentivesRelax the “push” incentives for selected high value teachers
17 Policy Simulation Simulate effect of selective retention bonuses$5000 and $10,000 one year bonuses to senior STEM teachers in “retirement windows” Deferred Retirement Option Plan (DROP) for STEM teachers There have been numerous studies simulating the effect of DC or CB-type plan versus traditional teacher plans. All find that conventional DB plans shorten work careers. Friedberg and Web (2005), Costrell and McGee (2010), Ni and Podgursky (2016).
18 Policy Simulation Effect of Various Retention Bonuses for Missouri STEM Teachers After Five and Ten Years Exp=32 Targeted Untargeted Size of Bonus Program Duration $5,000 $10,000 Five Years Additional Teaching Yrs 22.85 52.83 Gross Cost/Year $83,868 $82,274 $837,084 $788,765 Net Cost/Year $44,709 $43,115 $797,925 $749,606 Elasticity 1.57 1.60 0.14 0.15 Ten Years 60.41 137.90 $79,768 $78,384 $699,873 $670,787 $28,942 $27,559 $649,048 $619,961 1.79 1.82 0.16 gross cost=the increase in the cost of the senior teachers for extra years of teaching. net cost= gross cost-cost associated with needing fewer novice replacements.
19 Policy Simulation Effect of Various Retention Bonuses for Missouri STEM Teachers After Five and Ten Years Exp=32 Targeted Untargeted Size of Bonus Program Duration $5,000 $10,000 Five Years Additional Teaching Yrs 22.85 52.83 Gross Cost/Year $83,868 $82,274 $837,084 $788,765 Net Cost/Year $44,709 $43,115 $797,925 $749,606 Elasticity 1.57 1.60 0.14 0.15 Ten Years 60.41 137.90 $79,768 $78,384 $699,873 $670,787 $28,942 $27,559 $649,048 $619,961 1.79 1.82 0.16 2011 cohort. Aged 48-65, 5 or more years of experience. N=2131
20 Policy Simulation Effect of Various Retention Bonuses for Missouri STEM Teachers After Five and Ten Years Exp=32 Targeted Untargeted Size of Bonus Program Duration $5,000 $10,000 Five Years Additional Teaching Yrs 22.85 52.83 Gross Cost/Year $83,868 $82,274 $837,084 $788,765 Net Cost/Year $44,709 $43,115 $797,925 $749,606 Elasticity 1.57 1.60 0.14 0.15 Ten Years 60.41 137.90 $79,768 $78,384 $699,873 $670,787 $28,942 $27,559 $649,048 $619,961 1.79 1.82 0.16
21 Policy Simulation DROP PlansRetire and Keep Working, collect some or all of pension No further accrual of service years Examples Arkansas Teachers Up to 10 years, roughly 70% of retirement annuity Florida Teachers Up to 4 years, full annuity
22 Policy Simulation DROP Plan for STEM teachers 2 years.4 or .7 of regular annuity No further accrual of service years Results Very promising
23 Conclusion Late career retention policies hold promise as a way to raise teaching workforce quality Retention bonuses / Selective DROP plans Targeted to high-need fields/schools/high quality teachers Structural modeling is a useful and practical tool in analyzing pension plan reform