Skip to main content

Electricity Market Rules Did Not Provide A Worthy Opponent For JPMorgan's Brainpower

  • Author:
  • Updated:

The Federal Energy Regulatory Commission fined JPMorgan $410mm today and you can see why JPMorgan would be okay with that. The details are in this marvelously complicated FERC order and settlement agreement,1 but the outlines of the story are simple. FERC built a terrible box, and the box had some buttons that were labeled "push here for money," and JPMorgan pushed them and got money. You can understand the category mistake very easily:

  • FERC thought the box was for generating electricity at market prices but with a robust backup system to ensure reliable supply, and
  • JPMorgan thought the box was for dispensing money.

It's a perfectly understandable mistake to make if you have spent your career building and operating boxes that dispense money, as JPMorgan global commodities head Blythe Masters has. What else could the box be for?

I suppose we should talk about how the box worked, because this is that sort of blog. The settlement agreement details eight strategies, cleverly labeled Strategy A, Strategy B, Strategy C, Strategy D, Strategy E, Strategy F, Strategy G, and Strategy H. Strategy A is a doozy and we should talk about it in the footnotes,2 but really Strategy B, in addition to being much simpler, is a better encapsulation of what is going on here. Here is all of it:

42. Strategy B (January 2011). On January 1, 2011, JPMVEC [JPMorgan Ventures Energy Corporation] took over responsibility for four additional plants. JPMVEC developed a new bidding strategy for two of the plants (AL3&4): for every third hour, submitting an energy self-schedule while submitting prices of between $73/MWh and $98/MWh in the two hours in between. (A self-schedule means that the generator is a price taker, that is, the unit will accept any price set by the market, making a self-schedule more attractive than any price a generator can submit to CAISO [California Independent System Operator, the relevant grid operator].)

43. CAISO gave Day-Ahead awards for the hours in which JPMVEC had submitted self-schedules. JPMVEC’s bids in the intervening hours would typically be out of the money. But CAISO’s software needed to take into account the physical limitations of plants, including the fact that just as cars can only accelerate or brake at a certain rate, a power plant can only increase or decrease its output (i.e., “ramp up” or “ramp down”) at a certain speed, called its “ramp rate.” The CAISO system gave the units large Day Ahead awards every third hour due to the self-schedules and also gave JPMVEC Day Ahead awards in the intervening hours to respect the units’ ramp rates that JPMVEC had registered with CAISO. JPMVEC was paid as bid (via BCR) at the $73/MWh to $98/MWh prices it submitted to CAISO for the intervening hours.

BCR is "bid cost recovery," the system under which CAISO pays operators non-market bonuses to basically make sure they have their plants on when needed. The strategy is simple enough:

  • The market clearing price for electricity will be, say, $30/MWh.
  • Your inefficient old plants cost about $40/MWh to produce electricity.
  • So you offer to be a price taker every third hour, selling your electricity at $30/MWh.
  • The other two hours you offer $80/MWh, which is way way way above market.
  • The system always lifts the offers of price-takers, so you're contracted to supply electricity every third hour.
  • But the system, in its infinite wisdom, understands that you can't just turn a power plant on and off like that.
  • So it tells you to produce energy, at the prices you offered - effectively $80, $80, $30 - for each of those three hours.
  • The system, in its frankly circumscribed wisdom, doesn't understand (1) that you are fucking with it or (2) that your average price for those three hours is actually $63.
  • So you make a profit of $23/MWh instead of the loss of $10/MWh that you'd have at market rates.

I mean! The only proper response to that is, the first time you do it, CAISO calls you up and tells you to knock it off, and then changes its rules to prevent this sort of obvious gaming. JPMorgan didn't write in big block letters at the top of its bid that day "HI GUYS WE FOUND A LOOPHOLE AND THIS IS IT" but surely its bid should have had the same effect. We don't, however, live in the first-best world where CAISO just called them and told them to knock it off, so instead JPMorgan are paying a big fine two and a half years later.3 And they look like villains, and CAISO look like dopes, the two great stock characters of the theater of financial scandal. How you feel about this is almost purely a matter of personal aesthetic preferences; if you're an electric ratepayer you probably feel Bad and if you're the inventor of synthetic CDOs you're like "well of course we did that, we would be nuts not to, it was right there in the rules, they were practically begging us to do it."

The other strategies are exactly like that in concept4: FERC and the ISOs built a dumb system that rewarded a modicum of perfectly transparent gamesmanship; JPMorgan gamed them in the obvious and perfectly transparent way; JPMorgan made money that it did not, you would have to say, "deserve" to make; and now FERC is mad. The record has a certain amount of JPMorgan maybe lying about its intent and feelings on the matter,5 but there was no lying, or need for lying, about what happened. This is not like Libor, where you had to affirmatively lie. Here JPMorgan just followed the rules, simply and literally, but they followed them to places they were not meant to go.

A generic and important problem with financial scams is one of allocative inefficiency: if you find a scammy way to package mortgages or manipulate interest rates or just Ponzi it up, then money is flowing to mortgages or interest-rate swaps or your Ponzi scheme that in a world of perfect information would be going to more productive uses. That problem is often sort of abstract - telling the average person "well Libor manipulation probably lowered your mortgage rate, but really that money shouldn't have been going to your mortgage," tends not to boil his blood - but you can always fall back on a rhetoric of theft: even if it's hard to see exactly who was hurt by Libor manipulation, the banks' undeserved profits on it is reason enough to get mad.

Here, sure, JPMorgan made undeserved profits, and those undeserved profits came directly from electricity customers, so you can get mad about the theft if you want. (The fact that the theft was done through utterly transparent manipulation of dumb ISO rules makes me fairly sympathetic, but probably doesn't do much for the irate ratepayers, who after all didn't make those rules.)

But it's also harder to miss the inefficiency. JPMorgan was selling electricity from '50s and '60s era steam boiler power plants that produce electricity too inefficiently to compete with modern plants. They ran them anyway, as a frankly incidental component of their box-button-pushing strategy - if they could have done the whole thing synthetically they probably would have. They were just exploiting mistakes in abstract rules by sending around spreadsheets; they didn't have any particular interest in doing anything in the real world. But as it is, the rules they exploited required them to turn on massive inefficient power plants every day so as to generate overpriced and overpolluting electricity. Rather than generate electricity more efficiently they just found more efficient ways to get overpaid for the electricity they generated. As a metaphor for how financial innovation substitutes for real innovation it's hard to beat.

FERC, JP Morgan Unit Agree to $410 Million in Penalties, Disgorgement to Ratepayers [FERC]
Order Approving Stipulation and Consent Agreement [FERC]

1.You can always count on an electric regulator for a fancy prose style:

32. JPMVEC’s buyback bids at 120% of Day Ahead prices were intended to be above Real Time LMPs, but only by a small amount. Although Real Time bids at higher prices would have made a dec-down more likely, higher bids could also have reduced JPMVEC’s BCR payments. In response to JPMVEC’s bids, CAISO often dec’ed JPMVEC’s units down to their Pmin in the Real Time Market, resulting in a MEAF of zero.
33. JPMVEC’s bids took into account that, under the BCR formula as then applied, a zero MEAF had two important impacts on payments from CAISO. First, a zero MEAF largely neutralized reductions to BCR that would otherwise have resulted from JPMVEC’s -$30 Day Ahead bids.

Just beautiful stuff.

2.The gist of Strategy A, as I understand it, is roughly that CAISO decided which plants should turn on each day based on the average price they bid for that day in the previous day's one-day-ahead market, but then actually paid the plants that were turned on based on a two-tier system where their base generation could be at a much higher price than their marginal generation (or the market price). So JPM would effectively bid $90/MWh for its first 91 MWs (its base load, the minimum to turn on the plants) and -$30 (negative $30) for its next 134 MWs, for an average bid of $18ish. They'd get awarded the full 225 MWs at the market price, say $30. Then they'd go into the real-time market the next day and buy back the marginal 134 MWs at slightly above market prices, leaving them with just the base 91 MWs. And through the magic of the pricing system it was able to sell those 91 MWs at the $90/MWh that it had bid. My effort to do the approximate fake math is here.

3.TBF Strategy B only lasted a month so maybe they did catch on quick. The rule change to fix it is obvious: you gotta evaluate the average effective price of the whole award before awarding a ramp-down, instead of just awarding the self-schedule and then the rampdown at whatever price the plant wants.

4.Strategy D is a particular favorite of mine. It's almost exactly the same as B except it uses the ramp-down trick over the hours before and after midnight: "Because CAISO's system evaluated only one day's bids at a time," and because it knew plants couldn't turn off on a dime, it'd give JPMorgan an award to generate electricity at market prices (based on a sure-to-be-accepted bid of -$30/MWh) from 11 to midnight, and then at $999/MWh from midnight to 2am. Which is not exactly a peak time! "CAISO paid JPMVEC for those awards at the units’ bid price of $999, even though market prices for the period between midnight and 2 a.m. were about $12/MWh." Hahahaha I mean how can you not admire that at least a little bit? "Boss, I have found a way to charge 80x the market price for electricity at 1am." Why invent cold fusion when you can do that? Also, like, imagine the people getting their electric bills and finding out that their 1am porn-surfing cost them $1,000.

5.E.g., and here just kinda let the jargon wash over you:

In phone calls with the CAISO Market Monitor (MMU) in March and April 2011, when asked about the reasons for the company’s bids to CAISO, the Principal Investments personnel who spoke on behalf of JPMVEC stated that their goal was to have the units picked up by CAISO and to respond to the Real Time market. The Principal Investments personnel did not mention BCR or the MEAF as relevant factors.