Human-induced climate change, with its potentially catastrophic impacts on weather patterns, water resources, ecosystems, and agricultural production, is the toughest global problem of modern times. Impeding catastrophic climate change necessitates the widespread deployment of renewable energy technologies for reducing the emissions of heat-trapping gases, especially carbon di-oxide (CO₂). However, the deployment of renewable energy technologies is plagued by various market failures, such as environmental externalities from conventional energy sources, learning-by-doing, innovation spillover effects, and peer effects. In efforts to begin to address these market failures, several governments at all levels—city, state, regional, and national—have instituted various subsidies for promoting the adoption of renewable energy technologies. Public resources are limited and have competing uses. So, it is important to ask: how cost-effective are renewable energy subsidies? Are the subsidies even reaching the intended subjects—the potential adopters of renewable energy technologies? In this empirically-driven dissertation, I analyze these important policy design and evaluation questions with a focus on the solar subsidy programs in California. All programs to incentivize the adoption of renewable energy technologies run into the same key question: what is the optimal (maximum capacity inducing) rebate schedule in the face of volatile product prices and the need for policy certainty? Answering this question requires careful attention to both supply-side (learning-by-doing) and demand-side (peer effects) market dynamics. I use dynamic programming to analyze the effectiveness of the largest state-level solar photovoltaic (PV) subsidy program in the U.S. – the California Solar Initiative (CSI) – in maximizing the cumulative PV installation in California under a budget constraint. I find that previous studies overestimated learning-by-doing in the solar industry. Consistent with other studies, I also find that peer effects are a significant demand driver in the California solar market. The main implication of this empirical finding in the dynamic optimization context is that it forces the optimal solution towards higher subsidies in earlier years of the program, and, hence, leads to a lower program duration (for the same budget). In particular, I find that the optimal rebate schedule would start not at $2.5/W as it actually did in CSI, but instead at $4.2/W; the effective policy period would be only three years instead of the realized period of six years. This optimal (i.e., most cost effective) solution results in total PV adoption of 32.2 MW (8.1%) higher than that installed under CSI, using the same budget. Furthermore, I find that the optimal rebate schedule starts to look like the actual CSI in a ‘policy certainty’ scenario where the variation of periodic subsidy-level changes is constrained. Finally, introduction of stochastic learning-by-doing as a way to better capture the dynamic nature of learning in markets for new products does not yield significantly different results compared to the deterministic case. Another, still-unanswered, redistribution question related to the CSI program is: to what degree have the direct PV incentives in California been passed through from installers to consumers? I address this question by carefully examining the residential PV market in California by applying multiple methods. Specifically, I apply a structural-modeling approach, a reduced-form regression analysis, and regression discontinuity designs to estimate the incentive pass-through rate in California’s solar program. The results consistently suggest a high average pass-through rate of direct incentives of nearly 100%, though with regional differences among California counties and utilities. While these results could have multiple explanations, they suggest a relatively competitive market and a smoothly operating subsidy program. Combining evidence from the optimal subsidy policy design and the incentive pass-through analysis, this dissertation lends credibility to the cost-effectiveness of CSI given CSI’s design goal of providing policy certainty and also finds a near-perfect incidence in CSI. Long-term credible commitment as reflected through CSI’s capacity-triggered step changes in rebates along with policy and data transparency are important factors for CSI’s smooth and cost-effective functioning. Though CSI has now wound down because final solar capacity targets have been reached, the historical performance of CSI is relevant not only as an ex-post analysis in California, but potentially has broader policy implications for other solar incentive programs both nationally and internationally. / text