Are the Recent Declines in Solar PV Module Prices Sustainable?
Once dismissed as unrealistically expensive, solar photovoltaic (PV) power has been on a remarkable run lately: according to Bloomberg Professional Service, in comparison to the start of the decade, the annual installations of new solar PV power capacity have doubled. This expansion reflects in large part a dramatic decline in module prices from over $4 per Watt (W) at the beginning of 2008 to under $0.8/W by the end of 2013. In 2010, Richard Swanson, the founder of SunPower, documented that over the previous three decades the average sales prices of PV modules had conformed closely to a so-called 80% learning curve, that is, module prices declined by close to 20% with every doubling of the cumulative number of modules produced. However, recent solar module prices have not exactly followed the “Swanson curve”, as Figure 1 illustrates. In 2008 and 2009, module prices were above the benchmark predicted by the learning curve, a phenomenon widely attributed to a temporary shortage of polysilicon. Then, beginning in 2011, the situation reversed: price declines have been noticeably steeper than those predicted by the learning curve. This observation has raised several questions among industry analysts. Did these steep declines reflect accelerated cost reductions or potentially excessive additions to manufacturing capacity in recent years? Further, what price dynamics can be expected going forward, and what do these trends imply for public policy support for solar power?
Figure 1. Predicted and observed average sales prices (ASPs) of solar PV modules, 2008–2013. The red line in Figure 1 shows the log-linear 81% “Swanson curve” for those years. All prices are in 2013 U.S. dollars.
To tease apart cost reductions from excessive additions of manufacturing capacity, we establish a benchmark price for each time period. The benchmark in each period is the equilibrium price that would have been observed had it reflected cost reductions alone. Since the solar PV module industry is highly competitive, one can infer equilibrium prices from production cost data. In a recent working paper, we adopt this approach to estimate the equilibrium price, or Economically Sustainable Price (ESP), for the twenty-four quarters between 2008 and 2013. If module manufacturers had been able to sell their output at the ESP, the industry would have been in equilibrium and firms would have broken even on their investments in production facilities. The ESP at each point in time reflects not only the current full cost but also the anticipated lower costs of production in future periods. By comparing the ESP with the average sales price (ASP) at any given point in time, we can test whether capacity additions had been “excessive” so as to drive the market out of equilibrium.
Figure 2. ESPs generally track ASPs closely, though the two trajectories diverge after Q1-11. All prices are in 2013 U.S. dollars.
To estimate production costs, we rely on financial data from a sample of 10 firms with a combined market share of 35% and data provided by industry analysts. Figure 2 shows that the resulting ESPs generally tracked ASPs closely, though the two trajectories diverge after Q1-11 (i.e., first quarter of 2011). The close agreement between ESPs and ASPs before Q2-11 implies that we do not see evidence of overcapacity from Q1-08 to Q1-11. Comparing our estimates of ESPs and ASPs in each period, we find statistically significant differences between the two measures in Q1-12, Q3-12, Q1-13, and Q2-13. These results confirm the widely held belief that the industry indeed experienced overcapacity during those time periods. Moreover, a comparison between the green and blue lines in Figure 2 allows us to quantify the extent to which price reductions stem from overcapacity as opposed to continued cost reductions.
Despite the evidence of excess capacity in the industry, our analysis suggests strong continued cost reductions throughout the sample period of 2008–2013. Among other factors, these decreases could reflect improved polysilicon utilization rates and reductions in the costs of upstream steps, such as the preparation of silicon ingots. Our estimates of module production costs imply a 73% learning curve for solar modules. Thus, with every doubling in cumulative module production, we observe on average a 27% reduction in production costs, implying that ESPs have recently declined even faster than suggested by the 81% “Swanson curve”. This rate of cost reductions can be used to derive a trajectory of future equilibrium prices. The dashed yellow line in Figure 3 illustrates this projection and represents a trendline to which actual prices should converge over time. The blue line again documents market ASPs, and we regard the slight increase in module prices towards the end of 2013 as consistent with a gradual return of prices to the fundamental cost dynamics reflected in the yellow curve. While our analysis does not yet cover 2014, Bloomberg reports an average price of $0.76/W in 2014, a figure that is only slightly lower than the ASPs observed in 2013. We foresee a continued convergence between ASPs and ESPs.
Figure 3. Projected ESPs through 2020, assuming constant annual production of 40 GW. All prices are in 2013 U.S. dollars.
Provided the industry continues to produce 40 GW of modules annually in the coming years, our trajectory points to ESPs of ~$0.70/W and ~$0.61/W for 2017 and 2020, respectively. Notably, these prices remain above the $0.50/W level targeted by the U.S. SunShot program for 2020. Without a significant increase in annual production or a technological breakthrough, our point estimates suggest that the SunShot goal is unlikely to be achieved until 2025, at least if firms in the industry were to earn zero economic profits.
Our projections for solar module prices have immediate implications for the current debate about public policy support for solar energy. It is widely believed that the recent boom in solar installations in the U.S. is in large part attributable to the federal policy of allowing investors in solar installations a 30% investment tax credit (ITC). Under current law, this credit is scheduled to be reduced to 10% by the end of 2016 and to remain at that level in perpetuity. In recent work, Comello and Reichelstein (2015) examine the likely impact of this change in tax policy on the cost competitiveness of solar power for different regions of the U.S. and the different segments of the industry, that is, residential, commercial and utility-scale installations. Because future solar PV module prices are unlikely to decline at rates as dramatic as those observed between 2008 and 2013 (Figure 3), Comello and Reichelstein (2015) project that a step-down in the federal ITC would leave solar power uncompetitive by early 2017 in virtually all segments and locations. Given this projection, the authors evaluate a more gradual phase-down of the federal investment tax credit between 2017 and 2024. The gradual reduction in tax incentives would be calibrated so as to offset the anticipated cost reductions in new solar installations, with a commitment to end all federal tax support by 2025.
Cover image "Wayne National Forest Solar Panel Construction" by Wayne National Forest is licensed under CC BY 4.0.