Demand/Supply Analysis

What is it?

This analysis quantifies future demand and supply curves in order to predict future prices

When is it useful?

This analysis is most useful when you need to forecast in a commodity industry. It does not give you a “magic number”, but it does provide deep insight into future pricing and capacity dynamics .

An Example?

This example is for intraday spot pricing for electricity. Imagine a pool, where individual power plants can bid their capacity at different prices; these bids are accepted in order from the cheapest to the highest until all the demand is served.

Hydro plants have very low marginal costs – mostly in maintenance and distribution. They will bid very low to stay operational.

“Baseload” demand is also served by the Nuclear power plants. Although the capital and operating costs of these plants is huge, the marginal cost of capacity is small – all that is used is the Uranium fuel. So these plants will stay running even at very low demand.

Coal power stations are next in marginal cost. Gas power generation will bid higher still. It has higher marginal cost that coal, because gas is a more expensive fuel. However, the fully loaded cost of generation will be less than coal, so most of the new capacity is likely to be added at this point in the supply curve.

Oil is an expensive fuel to use for power generation. In this analysis, Oil Power generating stations will only be switched on when the price of power in the pool reaches $60/MWh. They will be mothballed for much of the year, only activated to serve “peak” demand.

This analysis shows bio fuels as still not competitive with oil. This may change as technology improves and the subsidy regime changes.

On the demand side, we can see that most of the demand in inelastic. At prices above $80/MWh, there are “interruptable” industrial customers who can switch to alternatives. Peak electricity demand equals 540GW.

This analysis enables us to see the price-setting mechanism with prices stabilising at $60/MWh. The more efficient oil power stations are operating, they are the price setting capacity. If demand falls, these will be priced out and switch off, with the price dropping to $45/MWh. If demand rises, all the oil powered stations will light up until they are all operating, at which point the price will rise to $80/Mwh and future demand increases will be accommodated by switching off interruptable customers. Bio-fuel powered generation capacity will still lie dormant while there are still interruptable customers to switch off.

If new nuclear generation capacity was built, it would enter at the low cost end of the supply curve, shifting it to the right and depressing electricity prices at all price points

[Practical note: This analysis is an abstraction of the real price-setting mechanism. In practice, supply and demand are frequently tied up in long term supply agreements. However, this does not reduce the model’s validity, providing it accurately represents price-setting at the margin.]

How do you do the analysis?

The analysis requires a combination of research, model building and judgement.

STEP 1) Build the current supply curve

The first step is to build the current supply curve. In the short term, capacity is inelastic – no new capacity can be introduced. In order to flex production therefore, all companies can do is to decide whether to run or not. A key assumption is made: Plants will price down to their marginal cash cost. Forget the fixed costs and capital invested – the business owner who is a price taker only has one decision to make – do I run or not? They are better off running if price is above marginal cost. This assumption is valid in fragmented, highly competitive markets – it is not true under oligopoly or markets with cartels operating.

In order to construct the current supply curve, you need to construct a database of all current capacity, together with the marginal cost of running that capacity. This will include raw material cost, logistics and distribution cost, variable labour, variable tariffs, commission/royalty payments. It is impossible to calculate this for every plant – to fill in the gaps, assumptions are required – for example, similar size plants with similar technology and management can be assumed to have similar operating costs. Industry association/benchmarking data can help.

STEP  2) Forecast future supply curve

One starting point is to map out future capacity expansions that are already committed or highly likely to occur. It is important to include the construction leadtimes , which are different for every industry. What is the typical length of time between the decision to increase capacity to when it is fully operational? Steadily more uncertain expansions should be added to the model. For future expansions, two different costs are needed – the marginal cost (as above) and the fully loaded cost. The fully loaded cost should included all fixed costs and cost of capital – it should represent the future price expectation at which the capacity has a Net Present Value of Zero. The underlying assumption is that at prices above this, the capacity will be built, below this, it will not be built.

You can layer on more complexity once you have the basics covered. For example, capacity extensions to an existing facility are often lower cost that a greenfield site. There may be breakthroughs in technology that make future plants more efficient than today’s plants, reducing their marginal cost. Exisiting plants can reduce costs through continuous improvement.

These two costs are used dynamically:

  • Once the capacity has been committed, include it in the supply curve at marginal cost.
  • Before the capacity is committed, include it in future supply curves at fully loaded cost
STEP 3) Forecast future demand curve
To forecast demand, we should split demand into different end use segments. Then extrapolate demand in each of these end-use segments based on historical patterns and assumptions about the future environment.
Many end-use segments will be insensitive to price in the short term. Use of gas for residential heating will not change year-to-year based on price. These demands can be aggregated together and shown as a vertical line at the right volume on our supply/demand plot.
If the end-use segment is price sensitive in the short term, we will need to determine the price at which buyers will stop demanding our product. This may depend on other commodity prices – for example, dual feedstock petrochemical plants can switch between using oil or Natural Gas dependent on price differentials. Insert this volume as a block at the switching price level.
STEP 4) Create dynamic scenarios
Modelling pricing/capacity is a dynamic activity. Early years pricing will impact capacity decisions, which will impact future pricing levels. Therefore the one year snapshot on the graph above is not sufficient to explore the dynamics.
Use the model to calculate the impact of different competitive capacity decisions. Then apply game theory to these payoffs to determine likely competitive dynamics and reactions.
Create 3-4 integrated consistent multi-year scenarios, modelling capacity additions in early years stepping into the supply curve after the construction leadtime.
STEP 5) Drive conclusions
  • What is the range of future pricing forecasts?
  • When are the key capacity decisions? When should we add capacity?
  • What are the Key events to monitor?
  • What determines future industry profitability? For example, if all capacity capacity has similar marginal costs the industry has a “flat cost curve”. This will result in low industry profitability as prices are driven close to marginal cost. It can be worth keeping some inefficient capacity around in order to act as a price “umbrella” for the rest. This strategy will be prone to a prisoners dilemma problem – competitors can benefit by adding efficient capacity at the expense of the company limiting its own capacity to maintain the price umbrella.

I want to know more

How can you adapt this concept?