Background: Despite being an important component of teacher compensation, government-sponsored teacher pensions are only 72 percent funded on average and have total unfunded liabilities exceeding $600 billion nationally (McGee, 2019; Novy-Marx & Rauh, 2011; "Public Plans Data," 2020). Annual per pupil teacher retirement costs account for 11 percent of total per pupil education expenditures (Costrell, 2019). The soaring costs of pension benefits are straining education budgets and potentially crowding out other expenditures, such as teachers' pay (McGee, 2016; Nation, 2017). Research suggests that the current structure of pension benefits likely has serious drawbacks for the majority of teachers (Aldeman & Johnson, 2015; Backes et al., 2016; Costrell & Podgursky, 2009). As state and local governments work to address their funding challenges, they are also considering changes to benefit design. Purpose: Advocates claim that teachers prefer the current benefit design and that any change would be detrimental to schools' ability to recruit and retain a high-quality workforce. However, recent theoretical evidence suggests that teachers might actually prefer alternative retirement plan designs that are less back-loaded and more current versus deferred compensation (McGee & Winters, 2019). But there is very little empirical evidence on teachers' preferences for retirement benefits. Owing to the back-loaded nature of traditional pensions, experienced teachers who have accrued more in pension benefits likely have much stronger preferences for retirement benefits than beginning teachers who may be uncertain about if they will remain teachers for their whole careers. However, all teachers are likely to place higher values on other components of their compensation than the value they place on their retirement plan types, such as how much money they will receive in retirement benefits and when they can retire. Data: We collected original data for nearly 6,000 teachers in RAND's nationally representative American Teacher Panel. We administered a 15-minute survey focusing on teachers' retirement preparation, knowledge, and preferences and included survey questions designed to measure financial literacy, personality, numeracy, and risk tolerance. Research Design: This paper utilizes a discrete choice stated preferences experiment similar to those by Maestas et al. (2018) and Mas and Pallais (2017). Teachers are presented two hypothetical job offers and are asked which job they prefer. Jobs include eight characteristics: retirement plan type, replacement rate, retirement age, salary growth rate, class size, health insurance, Social Security enrollment, and salary. Job offers are the exact same except that salary and one other non-salary job characteristic randomly vary between the two job offers. Varying a non-salary job characteristic along with salary allows us to estimate how much teachers are willing to pay for the job characteristic, going beyond simply asking teachers which retirement plan types they prefer. Importantly, we provide a description of retirement plans instead of the common-place retirement plan labels in the experiment to get a clearer picture of how teachers value retirement plans; we do not want to measure how much teachers are attached to the plan's name. Econometric Approach: We assume that teachers pick their job based on a latent utility model where unobserved utility is a linear and additively separable function of the job's non-salary components (retirement plan type, Social Security enrollment, etc.) and the job's corresponding salary. Teachers choose among the hypothetical jobs based on which job has the higher latent utility. We use a series of logits to estimate the probability that a teacher selects one job with a set of non-salary components and salary instead of another job with its own set of non-salary components and salary. We use a nonlinear transformation of the marginal utility estimates for each non-salary job component and the marginal utility estimate for salary from the logits to generate a measure of willingness-to-pay for each non-salary job component. Results: On average, teachers prefer traditional pensions to alternative retirement plan options at less than 3 percent of salary, but early-career teachers are indifferent to the type of retirement plan they are enrolled in. The most experienced teachers who are the most likely to have accrued the most retirement benefits value traditional pensions at less than 5 percent of salary. We find that teachers value other job components more than they value their retirement plan type. Teachers are willing to pay 11 percent of salary to be enrolled in Social Security and would be indifferent between switching from a traditional pension to an alternative retirement plan if the alternative plan allowed them to retire one year earlier. We find evidence of heterogeneity in plan type preferences based on cognitive ability and financial literacy. Conclusions: We show that teachers do not place very high valuations on their retirement plan types, and value other pieces of their compensation more than their plan types. We also show that plan type valuations differ substantially by experience where early-career teachers are indifferent to plan types and late-career teachers value traditional pensions at less than 5 percent of salary. The main limitation of our analysis is that we estimate stated preferences but are unable to track behavior. However, discrete choice experiments, such as ours, do well at predicting actual behavior using stated preferences in non-teaching populations (e.g., Hainmueller et al., 2015; Wiswall & Zafar, 2018; Wlömert & Eggers, 2016). We conclude that teacher retirement reform is possible. While most retirement reforms only take effect for newly hired teachers, our results show that current teachers would be willing to enroll in a different retirement plan if teachers are compensated. Policymakers may find that alternative retirement plans can alleviate future budgetary pressures and that giving teachers modest salary raises, lowering retirement ages, or increasing expected benefits to teachers can offset potential negative labor market consequences.