Source: April 13, 2016   SPAP Academic Salon

Seen from the United Nations Intergovernmental Panel on Climate Change (IPCC) latest assessment report, for initiatives in response to global climate change, the development of new energy is a non-negligible force. Development of new energy, such as solar energy, in the current situation still requires a certain degree of government subsidies. Mainstream economic theory suggests that the new energy subsidies are helpful to overcome two main market failures: negative environmental externalities and spillover effects of knowledge production. In fact, the new energy subsidies are useful to increase employment and promote economic development as well. In addition, new energy subsidies in a way can achieve the purpose of poverty alleviation, such as the "PV Poverty Alleviation" project promoted by our government.

In the United States, the forms of subsidies for solar energy are various: not only from the federal, but also from the state government, even from internal power company. How to assess the cost-effectiveness of these subsidies, policy design and pass-through effect are very important issues. Taking the state government of California as an example, the largest state level subsidy program in USA (California Solar Initiative, CSI) spent about $ 2.2 billion in achieving approximately 2 GW of installed capacity, or an average subsidy of about $ 1.10 per watt. However, even so, the price of rooftop PV system was also up to $ 6-10 / watt at that time.

To assess the cost-benefit of a project means to perform Counterfactual Analysis for maximized cost-effectiveness in the same condition. For this reason, I have built a dynamic optimization model to analyze, when the level of subsidy changed, whether we can improve the objective function, i.e. maximizing the total installed capacity. The results show that under the influence of strong Peer Effects and strong Learning-by-Doing Effects, we can further improve the level of early stage subsidy and reduce the level of last stage subsidy, to increase 8.1% of installed capacity. However, if the decision-makers still have other policy objectives or constraints, for example, to maintain the relative stability of policy, then the results from optimization model will be similar to CSI's final policy design.

      CSI policy design is actually very interesting. From international experience, the biggest problem for design of solar subsidy programs is the budget will be spent out in a very short period, similar to the PV installation users' "crazy snapping." The fundamental reason is that the lag and inflexibility of policy design: Due to the rapid decline in the cost of photovoltaic products, if the level of subsidy policy is not adjusted accordingly, then the subsidies is at risk of being snapped. But to adjust subsidy policy requires cost, and to gather relevant market information also requires cost. In the case of future technology with unpredictable speed of advances, CSI took linkage mechanism with a method called capacity targets and the level of subsidies to solve this problem in one action. Specifically, CSI take subsidy gradient approach: subsidies for each change (i.e. lowered) depends on the achievement of corresponding capacity goals of previous subsidy's level. CSI ultimately adopted 9-step approach, each step corresponding to each level of subsidy and capacity targets. CSI policy design is not like that at the beginning. The authorities have considered five plans and determined this program. Once the program is determined, the action will be strictly enforced in the future without any change.      

Just due to CSI's gradient and stepwise mechanism to adjust subsidy, the researchers can use this feature to take the method of Regression Discontinuity to estimate the pass-through effect of CSI. From the time dimension, due to corresponding capacity targets corresponding to a certain subsidy level will be achieved finally one day, then the next day the subsidies will be reduced. This provides us with an opportunity of Regression Discontinuity in time. On the other hand, from the geographical latitude, the subsidy rates between two bordered power companies are inconsistent (because of the different capacity target), there is a breakpoint opportunity geographically on the power company boundaries, which means different subsidies will appear at adjacent places (even to the residents in the same zip code) under similar conditions of supply and demand conditions. These two regression discontinuity opportunities are helpful for us to estimate CSI's shift effect. Of course, we should consider other issues in the actual estimation process, such as consumer self-selection bias, different time trends in market prices from power companies on both sides of adjacent places and so on. The final estimated result showed that CSI subsidies mainly fell into the hands of consumers, i.e., pass-through effect of market is extremely high.      

In fact, from the perspective of policy design and implementation, CSI is not only to solve the cash flow problems faced by consumers, but also at the information level, the authorities have done a lot of work. For example, CSI entrust a third party to maintain a good data website. It records the basic data of all users who applied for CSI and all products (after desensitization), such as the user's zip code, city, affiliated power companies, and price, size, manufacturer and basic technical parameters of the PV system. The information is public, helping compensate for the lack of market information, and to some extent promoting competition in the market. In addition, CSI has established an approved items of PV system installation manufacturers to facilitate the consumers to select. From the developing process of CSI, the various stakeholders (power companies, consumers, environmental organizations, enterprises, researchers) all have provided good advices and data support. These contributed to CSI's success.      

While CSI has ended in 2014, but its experience is promoted to other parts of the United States (such as Massachusetts) and Europe. Similar new products (such as electric cars, storage battery) can also adopt this mode. Finally, from the success of this solar subsidy policy, a good policy must have the following characteristics: simple, clear, credible, predictable, flexible, capable to finish and incentive compatibility. According to these standards, the level of China's policy has yet to be further improved.