Market coupling is one of the key-policies for achieving the EU single electricity market. The EU Commission praises the price-lowering effects of market integration in the first draft of the Internal Market Communication of August 30th: “wholesale electricity prices in the EU have risen much less thanks to competition facilitated by increasing cross-border trading and market integration”.
And common sense would indeed suggest that in competitive markets the average price of two market zones will be equal or lower when they are coupled than when they are separate. In fact, coupling should lead to lower average prices for typical electricity markets (increasing marginal cost on the supply side and price-inelastic demand). The intuitive reason is that the most expensive MWh in the expensive country might be replaced by switching on one additional MWh in the cheaper country. As marginal cost is increasing the switched-off MWh will be disproportionally more expensive than the switched on MWh. In our example (see Figure 1) in the first market 10 MWh with marginal cost of 158-176 EUR are switched on while 10 MWh with marginal cost of 176-239 EUR are switched off.
Figure 1: Market coupling with well behaving cost functions under perfect competition
Market 1: Marginal Cost = Volume(1.1) Demand = 100 | Market 2: Marginal Cost = Volume(1.4) Demand = 50 |
In a Cournot competition setting this effect is amplified by the increase in the number of players in the joint market. This increased competition in the joint setting will drive down prices compared to the separate market setting. In our example (see Figure 2) coupling two monopolistic markets (with similar cost curves) to one duopolistic market drives down the average price by 7 percent.
Figure 2: Market coupling with well behaving cost functions under imperfect competition
Market 1: Cost = ½*Volume2 Demand = 300-Volume | Market 2: Cost = ½*Volume2 Demand = 300-2xVolume |
The above illustrated supply curve smoothing and competition enhancing effect of market coupling would suggest, everything else being equal, that consumers in two markets in total face lower electricity bills when the markets are coupled, then when they are not.
We now move on to test this important proposition empirically using European spot electricity prices. The hypothesis is that at a given total demand in the two markets, average prices are typically lower when the markets are coupled than if they are decoupled. In Figure 3 the relation between load-weighted average prices (in €/MWh) in the Franco-German market and total volume (in GWh) in the Franco-German market is depicted (green dots). Thereby, the upper subplot only plots the instances were electricity prices in both countries are close to equal (price difference less than 5%) while the lower subplot restricts the sample to hours where the prices were differing by more than 5%. To make the subplots comparable the polynomial approximation to the volume-price relation has been depicted in both cases as a line.
Figure 3: Weighted average Franco-German price for coupled and decoupled hours in 2010
In Figure 4 the prices in both cases are compared. Based on the flattening supply curve and the competition effect one would expect that the red line that represents the average price when markets are decoupled is systematically higher than the blue line that represents the average price when markets are coupled (the lines are polynomial approximations to the price/volume curve). The surprising result is that price equalization (coupling) does only lead to a slightly lower average price at very high levels of electricity demand for the depicted Franco-German case.
Figure 4: Comparison of the weighted average Franco-German price for coupled and decoupled hours in 2010
This finding can be reproduced for many years and country combinations (see Figure 5). The only exception from this rule was the Germany-Nordic market combination in 2008 that featured significantly lower prices when the markets were coupled, than when they were not. Results are also not sensitive to the difference threshold. The figures for the 1% and 10% threshold for all 12 cases are suggesting no systematically higher price in case of decoupling like the here reported 5% threshold results.
Figure 5: Comparison of the weighted average prices for coupled and decoupled hours in various European markets
The presented findings are still very preliminary. If they can be confirmed more generally they would point to the conclusion that market coupling is much less beneficial for final customers than one might have expected. In this case, identifying the reasons why market coupling does not drive down prices is required to be able to reap the full benefits of market integration for consumers.
One possible explanation could be the sometimes unusual shapes of electricity cost curves (compared to microeconomic textbook cost curves). When moving to a cost function with a long flat left that become very steep at the right end the above outlined proposition that market coupling leads to lower average prices (due to the flattening supply curve and the competition effect) does not need to hold anymore. As illustrated in the below example (see Figure 6) the price might converge to the higher price when the more sizable low price zone is forced to accept higher prices due to coupling. More research on under which conditions to expect rising/falling prices from market coupling is necessary. This possibly needs to involve supply function equilibria to address the effects of oligopolistic competition in market integration.
Figure 6: Market coupling with non-linear cost functions under perfect competition
Just to be clear, the presented findings do not challenge the view that coupling markets is welfare enhancing. In all presented simulations, total welfare after coupling is larger or equal than in the uncoupled case. The results do also not allow deductions on additional benefits of market coupling such as potentially reduced volatility. So the sole finding I would like to put up for discussion is that market coupling might not have reduced average wholesale prices and hence did not increase consumer rents. I would encourage all readers to enlighten me on what might have caused this finding that was completely contrary to my prior expectation.
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------
PS
Figure 7
Figure 8
Comments
Subscribe to posts' commentsProteos 25th September 2012
What you find might also be the consequence of surprise productions. Let me explain: the grid has an operator that makes forecasts, and tells producers who should be ready to produce. The surprise producers can produce more power at a price of 0, except they don\'t warn anybody and the grid operator is forced to take their production. If the surprise is big enough, the link between the 2 markets is saturated whatever the forecast, and the 2 markets have difference prices. The market with the big surprise gets a much lower price: the surprise production can even burn through the baseload production (nuclear/lignite) which tend to be a must run techno at these timescales. The other market does not see such a dramatic decrease, since the link saturates. A big condition that seems to impose is that surprises are asymmetrical: there are bigger surprise production than surprise gaps. Or there might be enough reserves so that the price increase is only modest in these conditions. Now, since some time now, renewable productions are subsidized, especially in Germany. The forecasts for these productions is still imperfect: although the mean error is rather small, there are still times where the error is big in relative and absolute terms. The renewables thus fit well the description of a surprise producer. Wind also has some nasty properties: the mean production is relatively low (capacity factor of 18% in Germany…) but there are bursts of high production. Thus more surprises of production than gaps. Renewables also tend to lower the capacity factor of conventional plants, so there are lots of reserves in general, to be had at a reasonable price. What do you think of this idea?
Reply to Proteos
Petteri Haveri 27th September 2012
Interesting findings that require further studying. Could one explanation be, that the transmission capacity wouldn\'t been used optimally? It wasn\'t until late 2010, when CWE market coupling started delivering and transmission was subject to explicit allocation. I would recall seeing pictures, where right after the coupling prices converged and on a seemingly lower level. The same happened earlier after TLC-coupling. Would it make sence to do similar comparison with Franco-German data, but between years, that is years before and after ITVC? Of course there have been some changes in generation capacity. The shapes of curves are also rather interesting. In some of those the average price actually decreases with the highest volumes (2008 curves). Can that have something to do with the averaging process, because intuitively it\'s a rather strange result.
Reply to Petteri Haveri
Georg Zachmann 28th September 2012
Response to Proteos, Thank you for this quite compelling hypothesis. So you are saying that markets decouple because we have sometimes very low prices in one market due to renewables that do not spill-over to the second market due to congestion (Probably the notion of surprise is not even needed). The potential price mitigating effect of market coupling would in this case be dominated by the fact that decoupling is often due to negative price shocks in one country (conditional-decoupling effect). It took me a bit to reply to this comment as I wanted to do some quantitative check’s. Here is my take: 1) There is certainly some truth to the argument. The question is whether the conditional-decoupling effect simply dominates a possible price mitigating effect in the empirical results or whether market coupling itself is not mitigating prices. 2) In my opinion the answer is not straightforward. The conditional-decoupling effect is very compelling for the German-French case in the late 2000s. We see indeed German exports being highly correlated to renewables feed-in and congestion occurring because of this. To check we ran the analysis considering German renewables feed-in as a reduction in German load. The outcome confirms Proteos hypothesis, as now instances of decoupling are featuring systematically higher prices (see figure 7). A second approach led to comparable conclusions. Only considering hours where German prices have been above 50€/MWh (hence obviously not mitigated by massive renewable feed-in) provide the same outcome – coupling coincides with lower prices (See figure 8). 3) It is still a bit puzzling that we see this effect at all borders. Renewables play a much less important role in France and Italy. Hence, I fail to see whether the conditional-decoupling effect might really explain the absence of lower prices in the French-Italian case. From this I think it is fair to conclude that estimating the (static) value of market coupling is quite a complex endeavor that merits further research.
Reply to Georg Zachmann
Georg Zachmann 3rd October 2012
@Tracy W: Thank you for this remark on terminology. Indeed, market coupling here is only defined with respect to price-equalization (“instances were electricity prices in both countries are close to equal - price difference less than 5%”). Interestingly, the picture does not change much with formal “market coupling” being introduced in 2011 at the Franco-German border (see http://www.flickr.com/photos/88023334@N06/8050821613/).
Reply to Georg Zachmann
Tracy W 2nd October 2012
I\'m puzzled by your definition of market coupling. Is it whenever there is no price difference between the countries? Because Germany and France were only market coupled (as in allocating the transmission capacity between them by algorithm) in November 2010, so I don\'t see how you can have drawn your graphs for Germany-France in 2008 and 2006 with that definition of market coupling. And France and Italy aren\'t market-coupled yet, to the best of my knowledge.
Reply to Tracy W
Georg Zachmann 3rd October 2012
@Petteri Haveri Thank you for your interesting comment. I have looked into the post-CWE data and the results do not change (http://www.flickr.com/photos/88023334@N06/8050821613/). The rare instances of decoupling are neither featuring significantly higher nor lower prices. It would be interesting to see the pictures on significant price effect of coupling you refer to. I was also puzzled by the shapes of the price-volume-curves and unless there is something wrong with my data this is an interesting phenomenon. It is definitely not solely due to the averaging process (http://www.flickr.com/photos/88023334@N06/8050827316).
Reply to Georg Zachmann
Proteos 30th September 2012
Georg, Thank you for testing my hypothesis. As for your point 3/, not all european markets are coupled together. AFAIK, Italy is not yet coupled to France, so the old system of auctions may still be in place. It is known to be inefficient. Italy is also a perenial importer of electricity, the biggest in the world: it may face congestion problems at its borders. The fact that the only necessary condition is to have congestion at the border is also worrying: high voltage lines have a cost, and not only in terms of euros. For exemple, a line between France and Spain was planned in 1994. It will only open in 2014, after massive protests and both countries have vowed to never again build a terrestrial HV line! One can also look to the german grid plan... And for the consumers, renewables are far from free. So I wonder what becomes of the consumer surplus in those conditions!
Reply to Proteos
Petteri 4th October 2012
I don\'t have those pictures, but would think if you have access to spot price data, it would be a rather straight forward to draw price series that represent before and after ITVC-prices. I did some google however, and found two presentations, one for ITVC (http://www.eurelectric.org/download/download.aspx?UserID=7312&DocumentFileID=72652) and for TLC (http://www.theapex.org/images/uploads/Bert-denOuden-Coupling-Development.pdf)
Reply to Petteri