Week 07 · Complete notes
Oligopolistic markets
How does strategic interdependence change price, output, differentiation and capacity decisions?
This print edition combines all five readings. Interactive definitions become ordinary text when printed.
01
Few firms, big stakes
What is an oligopoly, why do some markets end up with only a few firms, and how do we measure that?
A wheat farmer takes the price as given, and a monopoly sets its price facing only its customers. Most of the firms students go on to work for are in neither position. A handful of rivals share the market, and each one's best move depends on what the others do.
This reading defines oligopoly, looks at the forces that leave a market with few firms and introduces the two measures of concentration that competition agencies use.
1.1
Between monopoly and competition
An is a market with a few firms, each large enough that its choices move the market price or its rivals' sales. Global aluminum, Australian supermarkets, mobile networks, domestic airlines and memory chips are all oligopolies.
Neither of our earlier models fits. A competitive firm in Week 2 took the price as given, because its output was too small to move it. A monopoly in Week 4 faced the whole demand curve on its own. An oligopolist faces a demand curve that depends on its rivals: if a rival cuts its price or expands its output, the oligopolist sells less at every price it might charge. So its best decision depends on what it expects the rivals to do, and theirs depends on what they expect from it. That's a game, and Week 6 gave us the tools to solve it.
The products can be identical, as aluminum, crude oil and steel nearly are, or differentiated, as phones, cars and soft drinks are. The firms can compete on output, on price or on something else entirely, such as advertising, capacity or product design. This week we start from the two simplest cases, competition on quantity and competition on price, and then see what differentiation and capacity limits change.
1.2
Why so few firms
Oligopolies aren't accidents. Several forces keep the number of firms small, and they often reinforce one another.
- Large capital requirements. A smelter, a chip fabrication plant or a national mobile network costs billions. Few investors will commit that much to a market where one more firm's output would push the price down for everybody.
- Control of essential inputs. Firms with exclusive access to a mine, a spectrum license or prime retail sites have a cost advantage that a newcomer can't match.
- Legal restrictions and patents. Licenses limit entry in many utilities, and patents protect a few producers from imitation for years.
- Economies of scale. When average cost keeps falling over a large range of output, a few large firms can serve the market more cheaply than many small ones could. A small entrant starts at a cost disadvantage.
- Mergers. Firms combine to gain market power, share resources or reach a larger scale.
Memory chips show several of these at work. DRAM fabrication is expensive and has steep economies of scale, and prices swing with the business cycle. Around 2000, NEC and Hitachi merged their DRAM units into Elpida, and several other producers left the business after the dot-com crash. The 2008 crisis took out Qimonda, the memory spin-off of Infineon. Micron agreed to buy the bankrupt Elpida in 2012 and completed the purchase in 2013. Since then, three firms, Samsung, SK Hynix and Micron, have earned close to 90% of the world's DRAM revenue, according to the market researcher TrendForce. Their shares move from quarter to quarter, but the three stay on top.
1.3
Measuring concentration
How concentrated is a market? Two measures are in common use. Both start from market shares, the percentage of industry sales made by each firm.
The \(CR_n\) adds up the shares of the \(n\) largest firms, usually the top three or four. If the four largest firms hold 32%, 21%, 19% and 13%, then \(CR_4=32+21+19+13=85\%\). It's simple, but it ignores how the share is divided among the leaders and what happens outside them. A market where the top four hold 70%, 5%, 5% and 5% has the same \(CR_4\) as one where each holds 21.25%, but the first is much closer to a monopoly.
The uses every firm's share. Square each firm's share, in percent, and add the squares:
Three firms with 50%, 30% and 20% give \(50^2+30^2+20^2=2{,}500+900+400=3{,}800\). A monopoly gives \(100^2=10{,}000\). A market of 100 firms with 1% each gives 100. Squaring gives large firms more weight, so the two lopsided and even markets above get very different scores: \(70^2+3\times5^2=4{,}975\) against \(4\times21.25^2\approx1{,}806\), before counting the firms outside the top four.
A useful way to read the HHI: with \(n\) firms of equal size, it equals \(10{,}000/n\). An HHI of 2,500 is what four equal firms would give, and 5,000 is what two would give.
Competition agencies use the HHI to decide which mergers deserve a close look. The ACCC's merger assessment guidelines, published in 2025 for Australia's new merger regime, treat a market with an HHI above 2,000 as highly concentrated, and an increase of more than 100 points from a merger as significant. Five firms with 20% each give exactly 2,000. In the United States, the 2023 Merger Guidelines use 1,800 and 100. Both agencies say the HHI is one indicator among several, not a test that settles the question.
Keep in mind what the HHI measures. It describes the structure of a market, not how the firms behave in it. Two firms can compete so hard that the price falls to cost, as Reading 3 shows, and many firms can collude, as Reading 5 shows. A high HHI is a reason to look more closely, not proof of a problem.
1.4
Australian supermarkets
The ACCC's supermarkets inquiry, released in March 2025, estimated national supermarket sales shares for 2022–23. The report describes the industry as "an oligopoly".
| Supermarket | Share of sales |
|---|---|
| Woolworths | 38% |
| Coles | 29% |
| Aldi | 9% |
| Metcash (IGA and other independents) | 7% |
| Others | 17% |
Measure it
What is the three-firm concentration ratio?
\(CR_3=38+29+9=76\%\). Coles and Woolworths alone have 67%.
The four named firms contribute how much to the HHI?
\(38^2+29^2+9^2+7^2=1{,}444+841+81+49=2{,}415\). The 17% held by others adds at most \(17^2=289\) if it were one firm, and much less because it's spread across many. So the HHI is between about 2,400 and 2,700: above the ACCC's 2,000 line. It's close to what four equal firms would give.
Does an HHI above 2,000 show that Coles and Woolworths don't compete on price?
The HHI measures structure. The ACCC reached its conclusion that Coles and Woolworths have "limited incentive to compete vigorously with each other on price" from evidence on prices, margins and behavior, not from the HHI alone.
Concentration is only part of the story. The supermarkets also compete through differentiation: Aldi's small range of private-label products at low prices, the loyalty programs Everyday Rewards and Flybuys, store locations, online delivery and fresh food. Reading 4 shows why that kind of differentiation keeps prices above cost even with several rivals.
Summary. An oligopoly has a few firms whose choices move the price or their rivals' sales, so each firm's best decision depends on the others'. Large capital requirements, control of key inputs, licenses and patents, economies of scale and mergers keep the number of firms small. The concentration ratio adds the top firms' shares; the HHI adds the squares of all shares in percent, from near 0 to 10,000, and equals \(10{,}000/n\) for \(n\) equal firms. Both measure structure, not conduct.
02
Competing on quantity
How much should a firm produce when its rival's output moves the price it gets?
In some industries the main decision is how much to produce. Aluminum smelters, oil producers and steel mills decide output and capacity, and the market price is whatever clears the market for everything they bring.
This reading solves that game for two aluminum producers, first with three possible outputs and then with any output. The result sits between the monopoly and the competitive outcome, and it moves toward competition as firms are added.
2.1
AluStrong and MetaLite
AluStrong and MetaLite are the two producers of a high-grade aluminum used in cars and construction. Their products are identical, and their technology is well known: each has a cost of \(C(q)=20q\), where \(q\) is output in tonnes. Market demand is also well known. The inverse demand is
where \(q_a\) and \(q_m\) are the two firms' outputs. Each firm chooses its output without seeing the other's, and the price then adjusts so that buyers take everything produced. This is , often called Cournot competition after the economist who first analyzed it in 1838.
Start with three possible outputs: 30, 40 or 45 tonnes. Take one cell. If AluStrong produces 40 and MetaLite 30, the total is 70 tonnes and the price is \(140-70=70\). AluStrong's profit is \((70-20)\times40=2{,}000\) and MetaLite's is \((70-20)\times30=1{,}500\). Doing this for every cell gives the payoff matrix.
Before you solve it
MetaLite produces 30. What is AluStrong's best response?
Read down the first column: 1,800 at 30, 2,000 at 40 and 2,025 at 45. When the rival holds back, the best response is to expand. Against 40 or 45, though, AluStrong's best response is 40.
Underline each firm's best responses. Against 30, the best response is 45; against 40 it's 40; against 45 it's 40. Only one cell has both payoffs shaded.
The Nash equilibrium is 40 tonnes each, at a price of 60, with a profit of 1,600 each. Notice the cell (30, 30): each firm would earn 1,800 there, more than in equilibrium. We'll come back to it in Reading 5.
2.2
Marginal revenue that depends on the rival
Firms aren't limited to three outputs. To handle any output, use the rule from Week 4: a firm with market power produces where marginal revenue equals marginal cost. The difference is that AluStrong's marginal revenue now depends on MetaLite's output.
AluStrong's revenue is its output times the price, and the price depends on both outputs:
So AluStrong's marginal revenue is \(MR_a=140-2q_a-q_m\). Every tonne MetaLite produces lowers the price AluStrong gets on all its tonnes, and so lowers AluStrong's marginal revenue. Another way to see it: given MetaLite's output, AluStrong is a monopoly on the demand that's left over, \(p=(140-q_m)-q_a\). This residual demand has the same slope as market demand, but MetaLite's output shifts it in.
Set marginal revenue equal to marginal cost:
This is AluStrong's . For every output MetaLite might choose, it gives AluStrong's profit-maximizing output. If MetaLite produces nothing, AluStrong produces 60, the monopoly output. If MetaLite produces 40, AluStrong produces 40. If MetaLite produces 120, AluStrong produces nothing. The more the rival produces, the less AluStrong wants to produce. MetaLite faces the same problem, so its best response is \(q_m=\frac{120-q_a}{2}\).
2.3
The Cournot equilibrium
A Nash equilibrium is a pair of outputs in which each is a best response to the other: a point on both best-response functions at once. On a graph with AluStrong's output across and MetaLite's up, that's where the two lines cross.
Any other point fails. Pick MetaLite's output at 100: AluStrong's best reply is 10, but MetaLite's best reply to 10 is 55, not 100. Somebody would change their output, so it isn't an equilibrium.
To find the crossing, use the symmetry: the firms are identical, so in equilibrium they produce the same amount, \(q_a=q_m=q\). Substitute into either best response:
Each firm produces 40 tonnes, the total is 80 and the price is \(140-80=60\). Each firm earns \((60-20)\times40=1{,}600\). It's the same answer as the 3 × 3 matrix, because 40 happened to be on the list.
2.4
Between monopoly and competition
Compare the Cournot outcome with the two benchmarks from earlier weeks, using the same demand and cost.
| Market | Total output | Price | Industry profit |
|---|---|---|---|
| Monopoly | 60 | 80 | 3,600 |
| Cournot duopoly | 80 | 60 | 3,200 |
| Perfect competition | 120 | 20 | 0 |
The monopoly row comes from \(MR=140-2Q=20\), so \(Q=60\). The competitive row comes from \(p=MC=20\), so \(Q=120\).
Why do the two firms produce more than a monopoly would? When AluStrong adds a tonne, the price falls on all the aluminum sold. AluStrong counts the loss on its own 40 tonnes, but not the loss on MetaLite's 40. A monopoly owning both plants would count both, and it would produce less. Each Cournot firm ignores the harm its output does to the rival, so together they overproduce compared with the monopoly. That's good for buyers. It is also the reason the firms would like to agree to cut back, as Reading 5 shows.
2.5
More firms
What happens with \(n\) identical firms? Each firm's marginal revenue is \(140-2q_i-Q_{-i}\), where \(Q_{-i}\) is the output of all the others. In a symmetric equilibrium every firm produces the same \(q\), so \(Q_{-i}=(n-1)q\):
Total output is \(nq=\frac{120n}{n+1}\) and the price is \(p=20+\frac{120}{n+1}\).
| Firms | Each firm's output | Total output | Price | HHI |
|---|---|---|---|---|
| 1 | 60 | 60 | 80 | 10,000 |
| 2 | 40 | 80 | 60 | 5,000 |
| 3 | 30 | 90 | 50 | 3,333 |
| 5 | 20 | 100 | 40 | 2,000 |
| 11 | 10 | 110 | 30 | 909 |
One firm gives the monopoly. As firms are added, the markup over cost, \(\frac{120}{n+1}\), shrinks toward zero and the outcome approaches perfect competition. This is one reason competition agencies care about the number and size of firms: in a quantity-setting industry, fewer firms means a higher price. Cournot connects the concentration measures of Reading 1 to prices.
Summary. In quantity competition each firm sets \(MR=MC\), but its marginal revenue falls as the rival produces more. For AluStrong, \(MR_a=140-2q_a-q_m\), so the best response is \(q_a=\frac{120-q_m}{2}\). The Cournot equilibrium is where the best responses cross: 40 tonnes each at a price of 60. That lies between monopoly (60 tonnes at 80) and competition (120 at 20), because each firm ignores the price cut its output imposes on the rival. With \(n\) firms the price is \(20+\frac{120}{n+1}\), which falls toward cost as \(n\) grows.
03
Competing on price
What happens when firms selling the same product compete on price?
In Week 6 the two cafés could charge only $4 or $6, and both ended up at $4. Now let them charge any price. The result is stark: two firms are enough to drive the price all the way down to cost.
This reading derives that result, lists what it assumes and shows how a capacity limit undoes it.
3.1
The cafés, at any price
Return to the two cafés across the street. Their coffee is identical, each cup costs $2 to make, and customers buy from whichever café is cheaper. If the prices are equal, the customers split evenly. Each café can now charge any price it likes. This is , often called Bertrand competition after the economist who raised it in 1883.
Suppose both cafés charge the same price \(p\) above $2. Each serves half the customers and earns \(p-2\) on each cup. Now Café A cuts its price by one cent. Its margin falls by one cent, but it takes every customer, so it nearly doubles its sales. Undercutting pays. Café B reasons the same way. Any common price above $2 invites an undercut.
What about different prices, both above cost? Then the café with the higher price sells nothing, and it would do better by charging a cent less than its rival. Again someone wants to move.
What if both charge exactly $2? Each earns nothing on each cup. A café that raises its price loses every customer and still earns nothing. A café that cuts its price sells below cost and loses money. Nobody gains by moving.
The only Nash equilibrium is for both cafés to charge marginal cost, and both earn zero profit. This is the . It's striking because there are only two firms, the same number that gave a price of 60 in Cournot competition. Two firms selling identical products on price behave like a perfectly competitive market.
The Week 6 cafés got stuck at $4 for the same reason. With only two prices to choose from, $4 was as far down as undercutting could go.
3.2
What the result assumes
The Bertrand result is a clean benchmark, and it's useful mainly because it shows what has to be true for price competition to be this fierce.
- Identical products. Every customer switches for a one-cent difference. With brand loyalty, location or features, a price cut wins only some customers. Reading 4 treats this case.
- Equal costs. If Café A's cost is $2 and Café B's is $2.50, Café A can charge just under $2.50, take every customer and earn almost 50 cents a cup. The price ends at the higher-cost firm's cost, not the lower one's.
- Enough capacity. Undercutting pays because the undercutter can serve everyone. If it can't, the gain from cutting is much smaller. The next section treats this case.
- One-shot play. If the firms meet every day, a price cut today can be punished tomorrow. Reading 5 treats this case.
Notice that the result doesn't depend on entry. Zero profit in Week 3 came from new firms entering until profit disappeared. Here it comes from two firms undercutting each other, with no entry at all.
Real price wars do happen in markets close to these assumptions: petrol stations on the same corner, online sellers of the same branded product, airlines on the same route in the off-season. Just as often, firms in such markets work hard to avoid the assumptions, through loyalty schemes, service, location and limits on capacity.
3.3
Capacity limits
Go back to AluStrong and MetaLite and change two things. They now compete on price, not quantity, and each has a smelter that can produce at most 40 tonnes. Demand is still \(Q=140-p\), and the marginal cost is still 20.
Is a price of 60 for both an equilibrium? At 60, buyers want \(140-60=80\) tonnes, exactly what the two smelters can produce together. Each firm sells 40.
- Cut the price. AluStrong could charge 59 and attract every buyer, but it can supply only 40 tonnes, which it already sells at 60. The cut lowers its margin and adds no sales.
- Raise the price. If AluStrong charges more than 60, MetaLite still sells its 40 tonnes. AluStrong faces the demand that's left, \(q_a=100-p\). Its profit is \((p-20)(100-p)\), which is largest at \(p=60\). Raising the price loses more in sales than it gains in margin.
So both firms charging 60 is an equilibrium. That's the Cournot price, not the Bertrand one. A removes the reason to undercut, because a firm can't win customers it can't serve.
This points to a way of reconciling the two models. When firms first choose capacity and then compete on price, the capacities they choose play the role of the quantities in Cournot competition, and the outcome looks like Cournot. When capacity is easy to expand, as for software or online services, price competition is closer to Bertrand.
Summary. When firms with equal costs and enough capacity sell identical products on price, any price above marginal cost invites an undercut. The only equilibrium is \(p=MC\), with zero profit, and two firms are enough: the Bertrand result. It needs identical products, equal costs, enough capacity and one-shot play. A capacity limit removes the gain from undercutting: AluStrong and MetaLite with 40 tonnes each hold the Cournot price of 60 even when they compete on price.
04
Differentiated products
How does differentiation soften price competition, and when should we expect Cournot or Bertrand?
Few products are truly identical. Customers have favorite brands, favorite cafés and favorite phones, and many won't switch for a small price difference. Firms spend heavily to make that so.
This reading gives the cafés some regulars, finds the new equilibrium and compares how the two models of competition fit different industries.
4.1
Loyal customers and price-sensitive ones
In the smartphone market, some customers want an iPhone for its design, its integration with other Apple products and its ecosystem. Others prefer Samsung's Galaxy phones for their flexibility and customization. Both groups would switch for a large enough price difference, but not for a small one. A third group sees the two phones as close substitutes and buys whichever is cheaper.
changes the arithmetic of a price cut. A firm that cuts its price wins some of the price-sensitive customers, but few of the rival's loyal ones, and it gives a discount to all of its own customers, including those who'd have paid more. A price rise loses some price-sensitive customers but keeps most of the loyal ones. The best price balances the two, and it settles above marginal cost.
4.2
The cafés with regulars
Give the two cafés some regulars. Café A's daily sales are now
and Café B's are the mirror image. If Café A raises its price by $1, it sells 20 fewer cups a day. If Café B raises its price by $1, Café A sells 10 more. Coffee at the two cafés is now a pair of substitutes in the sense of Week 2, but not perfect ones. The marginal cost is still $2 a cup.
Café A's profit is \((p_A-2)(140-20p_A+10p_B)\). For a given \(p_B\), raising \(p_A\) by a dollar adds a dollar of margin on every cup sold but loses 20 cups at the current margin. The best price makes these two effects equal:
This is Café A's best-response function. If Café B charges $2, Café A's best reply is $5. If Café B charges $6, it's $6. If Café B charges $10, it's $7. In equilibrium both cafés charge the same price, so \(p=\frac{18+p}{4}\), which gives \(4p=18+p\) and \(p=6\).
Both cafés charge $6, sell \(140-120+60=80\) cups a day and earn \((6-2)\times80=320\) a day. Differentiation has lifted the price from $2 to $6.
4.3
Best responses that slope up
Compare the two graphs this week. In the aluminum game, AluStrong's best response slopes down: the more MetaLite produces, the less AluStrong wants to produce. In the café game, Café A's best response slopes up: the lower Café B's price, the lower Café A's best price.
So the two models predict different reactions to an aggressive rival. If a rival expands output, a quantity-setter cuts back and makes room. If a rival cuts its price, a price-setter cuts too, though by less: a quarter as much for the cafés. Economists call quantities in Cournot competition strategic substitutes and differentiated prices strategic complements. The names matter less than the direction of the reaction, which a manager needs to anticipate.
4.4
The price ladder
What if the two cafés set one price together, as if a single owner ran both? With a common price \(p\), each café sells \(140-20p+10p=140-10p\) cups. Joint profit per café is \((p-2)(140-10p)\):
Each café sells 60 cups at $8 and earns \(6\times60=360\) a day.
| Situation | Price | Cups per café | Daily profit per café |
|---|---|---|---|
| Identical coffee | $2 | Half the market | 0 |
| Regulars, prices set separately | $6 | 80 | 320 |
| Regulars, one price set together | $8 | 60 | 360 |
The Week 6 cafés couldn't hold $6, because their coffee was identical and undercutting took every customer. With regulars, $6 is the equilibrium. And $8 would be better still, if only each café didn't want to undercut it: Café A's best reply to $8 is \(\frac{18+8}{4}=6.50\). That's the problem a cartel has to solve, and it's the subject of Reading 5.
4.5
Cournot or Bertrand?
Which model should a manager use? Neither is right or wrong in general. Each fits industries where its assumption about the hard-to-change decision is realistic.
| Quantity competition (Cournot) | Price competition (Bertrand) | |
|---|---|---|
| Firms choose | Output or capacity; the market sets the price | Prices; they serve whoever buys |
| Fits when | Capacity is costly and slow to change, so it's chosen first | Capacity is flexible and prices are easy to change |
| Two firms, identical products | Price between monopoly and cost | Price equal to cost |
| Rival turns aggressive | Cut back your output | Cut your price too, by less |
| Examples | Aluminum, oil, steel, memory chips, grain | Software, online retail, insurance quotes, consulting bids |
Many industries combine the two. Airlines choose fleets and schedules months ahead and then set fares every day. As the capacity section of Reading 3 showed, when capacity is chosen first and prices second, the outcome looks like Cournot. Prices then fall hard only when there's spare capacity, which is when airline price wars tend to break out.
Which model fits?
For each market, choose the model that best describes how the firms compete.
A few oil producers decide how many barrels to pump each month. The price is set on world markets.
Producers choose output, and the market price clears. That's Cournot. It's also why OPEC sets production quotas, not prices.
Two firms sell almost identical tax-preparation software online. Serving one more customer costs almost nothing.
Capacity is unlimited and the products are close substitutes, so price is the decision and undercutting wins customers. Expect fierce price competition, and expect the firms to differentiate, for example with free basic versions and paid extras.
Two cement plants serve a region. Each plant takes years to build and runs close to full capacity.
Capacity is chosen first and is costly to change. Even if the plants quote prices, the capacity limit removes most of the gain from undercutting, and the outcome looks like Cournot.
Summary. Differentiation means a price cut wins only some of a rival's customers and gives a discount to your own, so prices stay above cost. For the cafés with regulars, the best response is \(p_A=\frac{18+p_B}{4}\), and the equilibrium is $6, against $2 with identical coffee and $8 if they set one price together. Best responses slope down for quantities and up for differentiated prices. Use Cournot when capacity is chosen first and hard to change, and Bertrand when capacity is flexible and price is the main decision.
05
Cartels and collusion
Why do firms want to collude, why do cartels break down, and when do they hold?
In every game this week, the firms would earn more if they could agree to compete less. AluStrong and MetaLite would earn 1,800 each by producing 30 tonnes instead of 1,600 at 40. The cafés with regulars would earn 360 a day at $8 instead of 320 at $6.
This reading looks at why such agreements are hard to keep, uses the repeated games of Week 6 to find when they hold, and asks what makes a market more prone to collusion. Then it turns to the law.
5.1
What a cartel wants
A is a group of firms that agree to restrict output or keep prices up, aiming for the profit a monopoly would earn and sharing it among themselves. OPEC is the best-known example: its members agree on production quotas to support the price of oil.
For AluStrong and MetaLite, a monopoly would produce 60 tonnes in total and earn 3,600. Split evenly, that's a quota of 30 tonnes each and a profit of 1,800 each, compared with 1,600 each in the Cournot equilibrium. The cartel gains by doing what Reading 2 said the Cournot firms fail to do: taking into account that each tonne one firm produces lowers the price the other gets.
5.2
The temptation to cheat
The trouble is that the quota isn't a best response to itself. If MetaLite produces its 30 tonnes, AluStrong's best response is \(\frac{120-30}{2}=45\). By producing 45, AluStrong earns 2,025 instead of 1,800. MetaLite faces the same temptation.
Look only at the quota of 30 and the Cournot output of 40:
This is a prisoner's dilemma. Producing 40 is a dominant strategy for each firm, and both end at 1,600 when both could have 1,800. The cartel agreement doesn't change the payoffs, so on its own it doesn't change the outcome. In a one-shot game the cartel breaks down. The same is true for the cafés at $8, and for every cartel: the price or quota that maximizes joint profit is one at which each member gains by cheating.
5.3
When the cartel holds
Cartels don't meet once. AluStrong and MetaLite choose their output every month, and Week 6 showed that repetition can sustain cooperation. Suppose each firm follows grim trigger: produce the quota of 30 as long as both firms always have; after any month in which someone produced more, produce 40 forever.
If AluStrong keeps to the quota, it earns 1,800 every month, worth \(\frac{1800}{1-\delta}\) today.
If AluStrong cheats by producing 40 this month, it earns 2,000 now. From next month, both produce 40, and AluStrong earns 1,600 a month after that.
The cartel holds if \(\delta\geq\frac{1}{2}\). This is the Week 6 formula \(\delta\geq\frac{T-R}{T-P}\) with \(T=2000\), \(R=1800\) and \(P=1600\). The punishment is credible, because producing 40 each is a Nash equilibrium of the one-month game: once the cartel has collapsed, neither firm can do better than 40.
A cheater could do even better than 40. Its best response to the rival's 30 is 45, worth 2,025. With that larger gain from cheating, the condition becomes \(\frac{1800}{1-\delta}\geq2025+\frac{1600\delta}{1-\delta}\), which gives \(\delta\geq\frac{225}{425}=\frac{9}{17}\approx0.53\). A bigger temptation needs more patience.
Price cartels work the same way. Two sellers of an identical product who agree on the monopoly price each earn half the monopoly profit. A cheater who undercuts by a cent takes the whole monopoly profit for one period, and then Bertrand competition leaves both with nothing. Cooperation needs \(\frac{1}{2}\cdot\frac{1}{1-\delta}\geq1\), so \(\delta\geq\frac{1}{2}\). With \(n\) firms each earns a share of \(\frac{1}{n}\), and the condition becomes \(\delta\geq1-\frac{1}{n}\): \(\frac{4}{5}\) for five firms and \(\frac{9}{10}\) for ten. The practice set works through this case.
5.4
What makes cartels fragile
The condition \(\delta\geq\frac{T-R}{T-P}\) explains most of what is known about when cartels survive.
- More members. Each member's share of the cartel profit shrinks, while the gain from cheating, taking the whole market for a while, doesn't. That's why the condition for \(n\) price-setters, \(\delta\geq1-\frac{1}{n}\), rises with \(n\).
- Hidden cheating. If a secret discount to a large customer goes unnoticed for months, the punishment arrives late. A late punishment is worth less today, which works like a lower \(\delta\). Cartels therefore try to make prices and outputs visible, through published price lists, shared sales data or a trade association.
- Infrequent or irregular sales. When a large contract comes up once a year, the gain from undercutting on it is large and the next chance to punish is far away.
- Shifting demand and costs. When conditions change, members disagree about the right price or quota, and it becomes hard to tell cheating from bad luck.
- Differences among members. A low-cost firm or one with spare capacity gains more from cheating and loses less from a price war. A member that badly needs cash now, which is a low \(\delta\), is tempted too. OPEC members facing a budget crisis have often produced above their quotas.
- Entry. High cartel prices attract new firms, which aren't party to the agreement. That's where Week 8 begins.
Firms can also hold prices up without any agreement. If every firm expects that a price cut will be matched at once, cutting doesn't pay, and prices stay high. This is . The Hats assignment from Week 6 showed one way it happens: a promise to match a rival's price, which sounds competitive, can remove the gain from undercutting.
5.5
Cartels and the law
In Australia, cartel conduct is prohibited by the Competition and Consumer Act 2010. Agreeing with a competitor to fix prices, restrict output, share markets or rig bids is both a civil contravention and a criminal offense, and individuals involved can be jailed. The ACCC investigates and brings civil cases; serious criminal cases go to the Commonwealth Director of Public Prosecutions.
Two cases show how cartels work and how they're caught.
- BlueScope Steel. In 2023 the Federal Court ordered BlueScope to pay $57.5 million, a record penalty for cartel conduct in Australia, in a case the ACCC brought in 2019. Between 2013 and 2014, BlueScope and a former general manager tried to get Australian steel distributors and an overseas manufacturer to agree on higher prices for flat steel. The agreements didn't have to succeed: trying to induce them was enough. The penalty was upheld on appeal in 2025.
- The European car-battery recyclers. In 2017 the European Commission fined three recyclers, Eco-Bat, Recylex and Campine, about €68 million. From 2009 to 2012 they had agreed to lower the prices they paid for scrap car batteries, a cartel of buyers rather than sellers. A fourth member, Johnson Controls, paid nothing, because it was the first to report the cartel.
The second case shows a powerful tool against cartels: leniency. The first member to confess and cooperate escapes the penalty, and later ones get smaller discounts. Leniency adds a new temptation to cheat, this time on the cartel itself, and it makes each member worry that another will confess first. It turns the cartel's repeated game back into a prisoner's dilemma.
Tacit collusion is harder for the law to reach. Firms that simply watch each other and match price changes haven't made an agreement. Competition law responds mainly through merger review: it's easier to stop a market from becoming concentrated enough for tacit collusion than to prove collusion afterward.
5.6
Looking ahead to Week 8
This week the number of firms was fixed. In Week 8 it isn't: an incumbent earning oligopoly or monopoly profits has to worry about firms that might enter. The tools from Weeks 6 and 7 come together there. Capacity becomes a way to commit, a threat of a price war against an entrant has to be credible, and a small entrant can sometimes do best by staying small.
Summary. Firms in an oligopoly earn more if they restrict output or raise prices together, but the joint optimum is never a best response to itself: AluStrong's best reply to the quota of 30 is 45. A one-shot cartel is a prisoner's dilemma. In a repeated game, grim trigger sustains the quota when \(\delta\geq\frac{T-R}{T-P}\): \(\frac{1}{2}\) if cheating means producing 40, and \(\frac{9}{17}\) if the cheater produces 45. Cartels are harder to hold with more members, hidden cheating, infrequent sales, shifting demand, different members and entry. Cartels are illegal, and leniency for the first to confess makes them harder to hold. Tacit collusion can do similar harm without an agreement.