Skip to content

Is Solar Really the Reason for Negative Electricity Prices?

Solar gets all the attention, but another explanation is at least as important.

Even during extreme promotions like “Black Friday” and “Cyber Monday”, it is rare to see anything sold for a negative price. Low prices, sure. But truly negative prices are rare. You only see this when a seller is really desperate to get rid of something. 1-800-GOT-JUNK made $200 million in revenue last year getting paid to haul away junk. Turns out all that stuff in your attic isn’t priceless after all — it has a negative price.


Source: Look how happy this guy is about negative prices.

This Spring, California electricity generators have been doing the electricity market equivalent to calling 1-800-GOT-JUNK. Between March and July, there were over 100 hours in which wholesale prices were below zero. These negative prices have received lots of attention, and the discussion has almost universally attributed negative prices to California’s ramp up in solar generation.

Solar is indeed part of the story, but another explanation is at least as important. Over the last decade, hydro, not solar, has been the primary driver of negative electricity prices in the United States. The year 2017 is no exception. Spring 2017 was among the rainiest in history, and it is this combination of hydro and solar that has pushed prices below zero.

Let’s Look at the Data

So far in 2017, California wholesale electricity prices have been negative in 2.5% of all hours, more than 130 total hours.


Note: This figure was constructed by Lucas Davis (UC Berkeley) using hourly wholesale prices from SNL Financial. The underlying data are complete for NYISO, MISO, PJM, and ISONE, but start in 2009 for CAISO, and in 2010 for ERCOT so there may have been negative prices in those markets prior to those years. Also, for 2017 data is only available until mid-August, so the percentage is calculated over only part of the year.

Increased solar generation is definitely part of the explanation for what’s happened in California in 2017. Solar capacity in California, including both distributed and utility-scale systems, has grown from less than 1GW in 2007, to 14 GW today. Also, the timing of negative prices in 2017 points squarely at solar. Negative prices peaked in March and April between noon and 5pm, on sunny days with high levels of solar generation.

Make it Rain

But for me the big surprise in the figure above is 2011. Although it didn’t receive as much attention, there were over 100 hours during 2011 with negative electricity prices in California. But why?  Back in 2011 there was just a fraction of the solar generation that we have today. So what pushed prices below zero?

The answer is hydro. The two wettest years in the last decade? 2011 and (probably) 2017. During just the first 5 months of this year California has already generated 22 million MWhs from hydro, about equal to typical hydro generation for an entire year. If the rest of 2017 matches hydro generation from last year, 2017 will end up being the second highest hydro year in the last decade.


Note: This figure was constructed by Lucas Davis (UC Berkeley) using annual California hydroelectric generation from EIA. The prediction for 2017 uses actual generation Jan-May and Jun-Dec generation from 2016.

Interestingly, the negative prices in 2011 are very different from the pattern in 2017, and very much fit the pattern for hydro. In 2011, negative prices peaked in May and June, considerably later in the year than in 2017. Also, 60% of negative prices in 2011 occurred between 4am and 6am, wee morning hours when electricity demand is at its lowest level.

Another Example

Northwest power prices provide another good example of U.S. electricity prices responding to hydro. The year 2011 was a record year for hydro generation up and down the west coast, and in the Pacific Northwest prices were often negative. The Northwest doesn’t have an ISO so I didn’t summarize these prices in the graph above, but a couple of years ago EIA made this nice figure:


Source: EIA, Today in Energy, “Negative Prices in Wholesale Electricity Markets Indicate Supply Inflexibilities”.

Off-peak prices were frequently negative during Spring 2011.   This was particularly true at the Mid-Columbia Hub (“Mid C”), right at the Washington/Oregon border.  Washington and Oregon both have large amounts of hydro generation, so during rainy years supply can outstrip demand resulting in negative prices.

California, Oregon, and Washington lead all U.S. states in hydroelectric generation and it is no coincidence that negative prices have occurred most often in these states. Contrast this with Texas, for example. I was surprised in the figure at the top of the blog to see virtually no negative prices in ERCOT. Texas has twice as much installed wind capacity as any other state so you might have expected to see negative prices during high wind periods. This may well happen in specific transmission-constrained locations (e.g. West Texas), but negative prices throughout the state are rare.

Why Not Just “Spill” Water?

But why are hydro generators calling 1-800-JUNK? Why don’t hydro operators instead just leave the water in the reservoir? Or “spill” water without running the generators? During these negative price hours, hydro owners are paying money to produce electricity – why? Why, for example, on April 9th of this year was so much hydro operating in the middle of the day, even while wholesale prices were negative?

april9.pngSource: CAISO Daily Renewables Watch. This figure shows hourly generation in California on April 9, 2017.

Part of the answer is that many hydro facilities are “run of the river”. Don’t think Shasta Dam. Think small-scale facilities in which the natural flow of the river is used to generate electricity.  Negative prices weren’t even envisioned when these facilities were first built, so many lack the ability to “spill” water when necessary and presumably it is very expensive to retrofit these facilities.

But large facilities like Shasta Dam are more puzzling. While it is true that these facilities are limited in how much water they can store, it is not clear this should matter. Nor should it matter if there are minimum and maximum flow constraints. When prices are negative, large hydro facilities should always be able to “spill” water and this would seem to be preferable.

Source: California’s Shasta Dam, public domain image.

This leaves me wondering whether there is something else: operational, legal, or contractual keeping these facilities running even when prices are negative. In the Northwest, dam operators have limits on how much they can “spill” because it churns air into the water which is bad for fish. Perhaps there is some similar regulation for California dams.

So my colleague Catherine Wolfram was right when she wrote back in March about the sinking duck. Water does have a lot to do with negative prices. Solar alone would not have driven such levels of negative prices this year if it hadn’t happened on top of near-record levels of hydro generation. In some sense, hydro can be even harder to turn off than wind and solar.



Lucas Davis View All

Lucas Davis is an Associate Professor of Economic Analysis and Policy at the Haas School of Business at the University of California, Berkeley. His research focuses on energy and environmental markets, and in particular, on electricity and natural gas regulation, pricing in competitive and non-competitive markets, and the economic and business impacts of environmental policy.

36 thoughts on “Is Solar Really the Reason for Negative Electricity Prices? Leave a comment

  1. Reblogged this on w70609 and commented:
    Lucas – very interesting piece – thank you! Are not solar subsidies and mandates part of the problem. The former were almost 70 cents/kWh, and are still about 10 cents per kWh! Paul Brady, UCD

  2. This post leads to a bigger question: Are hourly electricity prices true market metrics? In California the CAISO market prices are driven by variable fuel costs, yet increasingly its “zero cost” renewables that are supplying power. What happens when the only natural gas left running is providing local reliability needs, (or even none at all)? Back in 1996 I pointed out that CT operating costs didn’t represent a cap on capacity prices, and that auctions in hourly markets were function of a multi-hour/day unit commitment process. Even today, the system modelers can’t correctly or accurately simulate either of these market indicators. We should start thinking of a different metric for electricity marginal costs that are technology independent. We need to recognize that in fact electricity is stored in many ways and that the operations in one moment are linked closely to operations in other moments instead of pretending that each hourly market is independent. We aren’t clearing annual wheat futures markets–these are 8760 hourly or 105,120 5-minute markets annually.

  3. Negative power prices on the spot market occurred a couple of times last winter in Germany and Denmark. However, reason for that was not solar but rather wind power. Policies in Germany guarantee renewable power suppliers feed-in tariffs and additionally there is a first priority dispatch system in place. Thus, there is no incentive for renewable power suppliers to steer their intermittent power generation pending on demand (since they get paid for each kWh of power generation a guaranteed fixed price). Thus, mild winter days coupled with strong winds caused an oversupply in the power market so that the spot prices for electricity became negative. Traditionally, the base load of power supply in Germany is provide by coal (anthracite and bituminous coal) and these power stations cannot be easily turned on or off demanding on market demand and intermittent power supply.
    Thus, with the increasing power generation through renewables, the regulators and policy makers in Germany should review existing subsidies, feed-in and first priority dispatch schemes regarding renewables. Maybe it’s about time to move towards a more de-regulated and thus more competitive renewable sector, less relying on regulations and policies. I think political governance was important to support the emerging renewable sector. However, maybe some policies are not up to date anymore and rather create artificial negative power prices.

  4. The CPUC required utilities to track negative “avoided costs” beginning in the early 1980s, and there wasn’t a single utility report of negative prices to my knowledge for over 30 years, regardless of hydro conditions. (I worked at SCE for many years prior to retiring a few years ago) While hydro conditions and water obligations to downstream users can be reasonably expected to exacerbate negative prices, I think it is pretty evident that the large amounts of solar and associated “take and pay” contractual requirements are the reason for negative prices. Simulation models being run by the utilities and CAISO a number of years ago forecast the kinds of negative prices we are observing now, even under assumptions of “average hydro” conditions.

    • Carl,
      The “restructuring period” began in spring 1998 with great optimism in part because the hourly prices in the new CAISO markets were suppressed by the extreme wet hydro conditions of that winter. (Then the second wettest winter on record.) Prices were negative in April and May of that year for certain hours in the day ahead market. 1999 was a more normal year, and then in 2000, streamflows were at 90% of normal (in the “below normal” range in DWR parlance). That return to relatively normal hydro flows unmasked the vulnerabilities in the restructuring market design.

      • Yes, there weren’t visible spot prices in the market in the 1980’s and 1990’s. However, utilities tracked what was called “system lambda”, which represented the marginal dispatch cost of the next resource available. This is the basis for utilities’ negative avoided cost reporting, and what I referred to in my comment that solar, not hydro, was the direct cause of negative prices in today’s market. There are subtle differences between the RTM and system lambda, but these differences generally lead the RTM to include recovery of additional costs not included in the engineering-based system lambda metric, so I think the evidence that system lambda was never negative in even periods of substantial hydro spill is a pretty valid indicator.

        • The utility SRAC calculations for the system lambda assume average, not wet, hydro conditions. So it has missed the years when we’ve had excess hydro production, like in 2017. Those are the conditions that we are discussing here, not average conditions. So how the markets performed in previous wet years is the relevant metric here.

  5. With respect to this comment, “Plainly, the actual costs of production of electricity are never negative,” it depends on how you define costs. There are situations where injecting more power into the grid at one location can increase cost elsewhere. Hence, it may be necessary to discourage generation/encourage consumption.
    “A theory of electricity pricing that accurately reflects the underlying physical and engineering properties of electricity has not yet been developed. This article attempts to provide such a theory. The results we derive are quite different from spatial pricing results for conventional commodities. ….It is theoretically possible to have negative prices at some points for short periods, i.e., pay users to consume.”

    Negative prices are a sign of something undesirable, namely incomplete markets. If enough USERS saw these negative prices, they would increase their demand, thus avoiding negative prices. Since users are not allowed to see them, they cannot stabilize prices.

  6. The 5,000 MW wedge of inflexible thermal and nuclear is a significant driver of the negative prices. The nuclear output is almost constant on the graphic in the original article for all 24 hours of the day. Utterly inflexible. The thermal wedge is about a 2500 MW minimum, even at 3 AM and at 1 PM, but then widens out in the evening peak periods.

    These are units that must run during low-price (and negative price) hours in order to be available during higher-price hours. Some of the thermal are steam units running at their minimum levels (on the order of 20% of maximum output), and you can see them in the graphic ramp up sharply after 4 PM. In time, they will be replaced with flex units that can shut down completely (or almost completely) during low-value hours, and that will reduce the negative price periods.

    As the nuclear units retire (Diablo soon; Palo Verde probably not so soon, as the CE reactors have proved very reliable, economical and durable), this inflexible supply will go away. That will make room for about 2,500 MW in California, and 7,000 MW in the west, for intermittent generation. BUT, that will come with a price. Either we will build flex generation (gas) to fill in the gaps, and have both the cost and emissions of that, or we will build storage to fill in the gaps, and have significant costs for the storage in addition to renewable costs.

    An interesting opportunity I hope someone will write about. Geothermal has somewhat disappeared from the radar screen in California. It’s a logical replacement for retiring nuclear (tends to run evenly night and day). But the costs are high. However, one of the byproducts is the mineral content of the brine. Among these minerals is a LOT of lithium in the Salton Sea projects. Lithium is valuable for batteries (and for grease and some other things). If global demand for lithium continues to increase to provide batteries for EVs, mobile electronics, and utility grids, the byproduct “tail” of geothermal power may become valuable enough to wag the dog.

  7. There is a lot of technical discussion of hydro here, but it strikes me that the real issue here is about casuality. If this year there are both a lot of inflexible generation (hydro subject to noneconomic constraints or delivery requirements) and uncurtailable solar PV (rooftop or PPAs without curtailment provisions), which is the proximate cause of curtailments? The hydro plants (and also, incidentally, energy efficiency investments) have been in place longer, the solar is more recent, so we think of the hydro as creating a propensity for negative pricing (made it more likely to occur) and the PV at particular times as the proximate cause. Is that by itself enough of an explanation?

    Let me suggest a similar case which has been in the news wrong. Most of the readers in this blog probably believe in man-made global warming, and that it will create a greater likelihood of extreme weather events. Hurricane Harvey is an extreme weather event. I would say that the specific event was not _caused_ by global warming but that it was made more likely. Saying global warming was not the cause does not exonerate it, or mean we shouldn’t think about ways to address or ameliorate it. And even if we say that solar power (both rooftop and PPA) is the proximate cause of negative pricing, we have still wound up with all ratepayers (beneficiaries of the hydro, if you will) paying for it.

  8. Talk about singing to the choir! If you told anyone with economics training who is not involved with the electricity industry about negative prices, they would say that it’s not a normal good that your talking about. Yet the cognoscenti still go on about electricity “markets”. Plainly, the actual costs of production of electricity are never negative. Systems that generate “price signals” that communicate this are clearly perverse. No further progress can be made with electricity policy while the discussion is framed by a mindset that applies (actual) market economics to electricity systems.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: