Essays on subsidies, technology diffusion and the returns to education
MetadataShow full item record
This dissertation presents two lines of research on solar photovoltaic (PV) market and the returns to education respectively. The research on solar PV industry includes discussion of incentive pass-through for residential solar market, and the effects of government subsidies and a new business model (the third-party ownership) on consumers' solar demand. The research on the returns to education is about decomposition and comparison of linear estimators. The first chapter is about supply side of the solar industry. It measures incentive pass-through for third-party contracts in the California Solar Initiative (CSI). Previous studies found nearly complete pass-through for host-owned systems. In this chapter, the incentive pass-through rate for third-party contracts is estimated based on a unique dataset and we find that pass-through ranging from 40% to 45%, suggesting an incomplete pass-through. The second chapter estimates consumer demand of solar and discusses the effects of the third-party ownership (TPO) business model and government subsidies on the diffusion of PV. The demand model is estimated using the 2007-2013 PV installation data from the CSI program. Our counterfactual simulations indicate that TPO contributes 36% of the total installations. Of this, TPO products have gained 70% of the market share by attracting new customers into the market. If no incentives were provided, around 17% of the existing systems would not have installed. If TPO adoptions were not eligible for subsidies and the government allocates all subsidies to host-owned adoptions, there would be 16% more installations than under current policy. Based on our results, if the goal is to increase PV adoption, solar incentive programs should subsidize less on TPO products than on host-owned products. My final chapter is about linear estimators of the returns to education. This chapter examines the implications of the linearity restriction for estimating returns to education when the marginal effects of education on earnings are nonlinear. A reweighted OLS estimator, which would be more comparable across samples, is proposed. We show that the difference between OLS and IV estimators can be decomposed into two parts: the difference due to the bias in OLS marginal effect estimates and the difference due to different weights. This decomposition method could help to explain the puzzle that the linear IV estimates of return to education are generally higher than OLS estimates. By decomposition of the difference between OLS and IV estimates, we find that the difference is mainly driven by selection bias.