The Economics of Regulation, Geopolitics and the Labour Market

Branch: Economics & Corporate Completions (Finance-Plus Ring) · Node EC1.05 · ~18 focused hours · Standard gate ≥ 85%

Two ready-mix concrete plants in the same city announce a merger. On the share table the regulator is holding, the combined firm takes 42% of the catchment, the concentration index jumps 720 points, and in most jurisdictions a jump that size, on a base that high, carries a presumption of harm the parties then have to rebut. Then somebody asks whether site-mixed concrete, poured on the building site out of a rented drum, belongs in the same market. Say yes and the index falls to 1,345, the jump falls to 353, and the identical transaction clears without a single condition attached.

Nothing about the plants changed. A definition changed.

Regulation is the part of economics where a definition is worth more than a plant, and where the people who write the definitions are lobbied by the people the definitions bind. An analyst who cannot follow that machinery ends up doing what most of the street does: putting "regulatory risk" in a bullet at the bottom of a page with no number beside it, and then being surprised twice a decade.

The work ahead is the opposite of that bullet. Every rule gets a mechanism, every mechanism gets an arithmetic, and every arithmetic ends up on a line of somebody's model.

Three subjects sit here that are usually taught apart and belong together. Regulation proper exists because markets fail in five specific ways and governments answer with instruments that have their own failure modes. Geopolitics is regulation written by states against each other rather than against their own firms, and it arrives at a company through the same channels: a cost line, a customer, a licence, a port. The labour market is the largest market most economies contain, and the one where the headline number lies most often.

Here is the labour number that should bother you. A country's unemployment rate rises from 4.09% to 4.96% in a single month, and the correct reading is that the economy improved. Not one person in the sample had lost a job. What happened is that 1.55 million people who had stopped looking for work started looking again, which moved them out of the "not in the labour force" box and into the "unemployed" box, because a person who is not searching is not counted as unemployed in any headline rate. Participation rose 0.58 points on the same data. Read the unemployment rate alone and you get the story exactly backwards.

Geopolitics gets the same treatment. A framework that ends at "tensions are elevated" is worth nothing to a model. A framework that ends at "34% of revenue crosses a border where a levy of 4.2% is 85% likely, and 60% of that levy is passable to customers, so the expected drag is ₹23.30 crore against an EBIT of ₹672 crore" is worth a row in a valuation. Both sentences describe the same world.

The oligopoly work closes a gap. You already solve Cournot, Bertrand and the collusion condition. What has been missing is the pair of models that explain why an oligopoly price can sit unmoved through a seven-fold swing in marginal cost, and what happens when one firm moves first and can prove it.

The second produces a result worth carrying into the antitrust half: a market can become measurably more concentrated and simultaneously cheaper for its customers, and the concentration index will report the first fact while missing the second entirely.

One standing warning. Regulatory and geopolitical material is where a confident model is most dangerous, because the outputs are dates, thresholds and statutory names, and a wrong threshold reads exactly like a right one. A notification threshold, a tariff line, a concentration band and a labour-force definition all change by circular, and all of them have changed recently.

Treat an unflagged threshold from any source, human or machine, as unverified until you have opened the primary text yourself.

Learning objectives

By the end you can:

1. State the economic case for a regulation by naming which of the five failure modes it answers (externality, public good, informational friction, weak competition, a distributional objective), and say what a market would do in its absence.
2. Map a regulator onto its type and its instrument: statutory law, administrative regulation or judicial precedent; a government department, an independent commission or a self-regulating organisation, with the accountability difference each implies.
3. Diagnose regulatory capture from its three preconditions rather than from suspicion, and state the mechanism through which a concentrated interest beats a diffuse one even when the diffuse side is larger.
4. Run a merger review numerically: compute concentration before and after, verify the change by an independent route, run a SSNIP test from a margin and an elasticity, compute a diversion-based unilateral-effects screen, and reach a clearance, remedy or block verdict.
5. Distinguish a structural from a behavioural remedy by computing what each does to the concentration measure, to the market-definition test and to the expected harm, and state which one needs a monitor and why.
6. Price a natural monopoly two ways: build a revenue requirement under rate-of-return regulation, show why it rewards capital spending over operating savings, then value the same cost saving under a price cap and quantify the difference.
7. Convert a regulation into a model line: run a cost shock at three pass-through assumptions, weight the scenarios, and express the residual uncertainty as both a scenario value and a discount-rate equivalent, saying which of the two is defensible.
8. Classify a geopolitical risk as event, exogenous or thematic; assign velocity, likelihood and impact; and build a firm-level exposure map from revenue, cost, licence and asset lines rather than from headlines.
9. Compute the labour-market panel end to end: labour force, participation rate, unemployment rate, employment-population ratio, a broad underemployment measure and the effect of discouraged workers re-entering, then apply the frictional, structural and cyclical taxonomy and Okun's relationship in both directions.
10. Solve the two oligopoly completions: find a kinked demand curve's marginal-revenue discontinuity and the exact range of marginal cost over which price does not move; and solve a Stackelberg leader-follower equilibrium against a stated demand system, comparing price, quantity, profit, consumer surplus and concentration with the Cournot and Bertrand outcomes.

Prerequisites & connections

Builds on. M0.02 named the referees and where each one's writ actually runs; the machinery of why a referee exists picks up from there. M4.03 supplies industry structure and the informal buyer-power question that becomes the SSNIP test here. EC1.01 is the direct parent: surplus, elasticity, concentration measures, critical loss, Cournot, Bertrand and the grim-trigger collusion condition are all its property and are used here without re-derivation. ES1.02 owns the securities-market regulators in operating detail, so the financial-regulation section below points at it rather than rebuilding it. M7.04 supplies the fiscal instruments a state uses on itself, and M7.05 the cycle whose position decides how much of an unemployment rate is cyclical. Ordinary algebra and a derivative read as a slope are the only mathematics required.

The deliberate non-overlaps run as follows. EC1.01 owns market structures, elasticity, consumer and producer surplus, deadweight loss, the concentration measures themselves, and the Cournot, Bertrand and game-theory core; the kinked-demand and Stackelberg completions live here by design and are the only oligopoly material added. ES1.02 owns the SEC, SEBI and FCA regimes, the Marketing Rule, the Investment Advisers and Research Analyst regulations, and GIPS; nothing here re-states a securities rulebook. M7.05 owns business-cycle dating, the indicator toolkit and the credit cycle; the labour work below supplies the measurement definitions that cycle analysis assumes and never defines. M7.07 owns India's labour narrative, informality and the female-participation debate as a country diagnosis; the arithmetic that makes those claims checkable is what appears here. M4.03 owns Porter and industry definition. M3.03 owns the cost of capital that the last step of every risk-pricing exercise below feeds into.

Feeds forward. Every risk register you write from here is a set of rows in a model rather than a page of adjectives, and the mini-project builds the first one. The valuation phases inherit the discount-rate-versus-scenario discipline directly. Credit work inherits the regulatory-risk half, because a licence condition is a covenant nobody negotiated. The macro dashboard work inherits the labour panel, since a participation rate is the single most misread number on it.

Core teaching content

Who regulates, and the three kinds of law that bind

A firm is bound by three different things that all get called "the law", and they behave differently.

Statutory law is passed by a legislature. It is slow to make and slow to change, it sets objectives and delegates the detail, and it is where the big discontinuities live. Administrative regulation is issued by an agency under a delegation from a statute: fast, technical, voluminous, revisable by the agency alone, and where almost all of the operating burden on a real company sits. Judicial law is made by courts interpreting the other two, unpredictable in timing and able to invalidate a rule a company has already spent money complying with.

The bodies applying those three come in three kinds.

A government department or ministry is directly accountable to the elected executive, which makes it responsive to political priority and vulnerable to political cycle. An independent regulator is created by statute with its own board, its own funding and a fixed-term leadership designed to be hard to remove, which buys predictability at the cost of democratic responsiveness. Securities regulators, central banks, competition authorities and utility commissions are usually built this way. A self-regulating organisation is an industry body with delegated authority over its own members, which buys technical competence and speed at the obvious cost, since the members are also the governed.

An exchange policing its listed companies, an accounting body setting audit standards and a payments network writing scheme rules are all in that last category, where independence is a claim to be tested rather than a fact.

Two structural facts matter more than the taxonomy. Regulators are not the only enforcers: private litigation, insurance underwriting, lender covenants and index-provider eligibility rules all discipline firms with no state involvement at all, and sometimes faster. And jurisdiction is contested, since a single transaction can require clearance from three competition authorities, a sectoral regulator and a foreign-investment screen, any one of which can stop it.

The binding constraint is the slowest and strictest of them, never the domestic one.

Regulation of commerce and of labour

Commerce regulation is the least glamorous and the most likely to appear in a cost line. It covers company law, contract enforcement and bankruptcy, product safety and standards, competition law, consumer and data protection, environmental compliance, and the whole apparatus of customs, tariffs, quotas and export controls that governs a good crossing a border. For a manufacturer the customs half of that list is not background.

It is a direct input to landed cost, and a change in it moves gross margin within one quarter.

Labour regulation binds the largest single cost line in most service businesses. The instruments are a minimum wage, hours and overtime rules, safety standards, hiring and firing procedure, collective-bargaining rights, social-insurance contributions, and the definition of who counts as an employee at all. That last one is the live question in most jurisdictions: a platform whose workers are contractors has a materially different cost structure from one whose workers are employees, and the difference is a legal test rather than an economic one.

The economics of a binding minimum wage is the most commonly mis-taught result in the subject. In a competitive labour market a wage floor above the market-clearing wage reduces employment, by the standard supply-and-demand result. In a monopsony, where one buyer dominates the hiring side, the employer already holds wages and employment below the competitive level, so a floor set between the monopsony wage and the competitive wage raises both.

Which case applies is an empirical question about the local labour market rather than a matter of ideology, and the literature is genuinely contested. Ask how concentrated the local employer side is before assuming a sign.

Financial-market regulation is the third branch, and it is deliberately not rebuilt here. The disclosure regime, the fiduciary standard, the registration thresholds and the conduct rules are covered in operating detail elsewhere, and duplicating them would produce two versions to keep current instead of one. What belongs here is the economic reading of why that regime exists at all: an informational-friction regime almost end to end, layered over a systemic-externality regime for the institutions whose failure damages others.

Assembling one register

Everything above lands in the same place: a table with a company's name on it.

The register has one row per risk and six columns, and the discipline is that no row may be left qualitative.

Two rules make the register usable rather than decorative. The first is that a risk with no model line does not belong on the register; it belongs in a note, because a risk that cannot move a number cannot be priced and pretending otherwise crowds out risks that can. The second is that the register is scored later. Write the probabilities down, revisit them in a year, and count how often the 20% risks happened.

That count is the only feedback loop this discipline has, and without it a register is a genre exercise.

Common mistakes & how experts think differently

1. Treating the concentration index as the verdict rather than an input. The tempting logic is that a high index means pricing power and a low one means competition. The Stackelberg arithmetic above breaks it in one line: the more concentrated market had the lower price. An expert computes the index because a regulator will.

Then comes the separate question the index cannot answer: how elastic is the demand this firm actually faces, once every substitute and every potential entrant is counted?

2. Skipping market definition because the share table was already in the pack. A share table arrives pre-computed and looks like data. It is not data; it is the output of a definition somebody chose. The identical ready-mix merger clears or fails depending on whether site-mixed concrete is in the market, and no amount of care applied afterwards recovers from getting it wrong.

An expert always asks what the denominator is a denominator of, then runs the critical-loss test rather than accepting the boundary.

3. Believing a behavioural remedy has done something structural. A price commitment is a promise, and a promise is worth the probability it is enforced multiplied by what it would have prevented. The index does not move, the market definition does not move, and the harm falls only in expectation. Experts price behavioural remedies as options with a counterparty and a monitoring cost, and prefer a divestiture whenever one is available at similar efficiency loss.

4. Putting an idiosyncratic regulatory risk into the discount rate. It is one keystroke and it feels conservative. It is neither correct nor conservative: it penalises every year of the forecast, including the years after the rule lapses, and it prices a diversifiable risk as though it were systematic. The tariff example needs 866.58 basis points of premium to reproduce a permanent shock, and 122.22 basis points to reproduce a three-year one, which is the same fact told twice. An expert puts the effect in the cash flows, with the duration stated, and quotes the discount-rate equivalent only so the reader can judge the size.

5. Reading a falling unemployment rate as good news without checking the denominator. The rate has a moving denominator, and people leaving the labour force reduce it exactly as effectively as people finding jobs. Look at the participation rate and the employment-population ratio in the same glance.

An expert quotes all three or none, and treats a fall in unemployment alongside a fall in participation as a warning rather than a result.

6. Applying Okun's coefficient as though it were a constant of nature. It is an empirical regularity estimated on one country's data over one institutional era, and moving it from 1.6 to 2.4 on the same gap changes the output estimate by hundreds of billions of dollars. Experts state the coefficient, the country and the period it came from, then run the range rather than the point.

7. Treating a sticky price as evidence of a stable business. Inside the kinked demand curve's rigidity range, the price is frozen and the margin is not. Contribution can fall by 44% with the price unchanged to the rupee. An expert watching an oligopoly through a cost shock looks at gross margin and unit contribution, not at the price list.

8. Reading a Stackelberg leader's output as aggression rather than commitment. The leader's advantage comes entirely from being unable to change its mind. A capacity announcement that is reversible is not a commitment and buys nothing; a plant under construction with take-or-pay supply contracts is a commitment and changes the equilibrium. Experts ask what makes the announcement irreversible before pricing any first-mover advantage.

9. Diagnosing regulatory capture from outrage rather than from structure. Capture is a prediction that follows from asymmetric stakes, information and attention, and it is testable: count the consultation responses by category, count the technical staff funded independently of the industry, check whether the rule has a sunset. An expert names the mechanism and the evidence, because "the regulator is captured" without a mechanism is unfalsifiable and therefore useless in a model.

10. Using a geopolitical risk index as a firm-level input. A text-based index measures how much a topic is being written about, which is not how exposed one company is. A firm with no cross-border revenue is unaffected by a spike; a firm with a single-source input is exposed at an index reading of zero. Experts use the index to date and size a regime, and the filing to size the company.

11. Quoting a threshold from memory. Notification thresholds, concentration bands, tariff lines and labour-force definitions all change by circular and have all changed recently. A confidently quoted stale threshold is the most common error in this entire subject, and it survives review because it sounds exactly like a current one.

Experts cite the instrument, the date they checked it, and where it lives.

Worked examples

Every figure below was recomputed in Python before it was written down, with the cross-checks stated in each solution. Companies are synthetic and labelled so. Curves, share tables and panels are illustrative, internally consistent, and shaped like the real thing without being any particular real thing.

Worked example 1: A merger review, end to end (India, ₹)

Nilgiri Ready-Mix (synthetic) proposes to acquire Chenab Mix (synthetic). Both supply ready-mix concrete into one metropolitan catchment, which absorbs 4.2 million cubic metres a year at a delivered price of ₹5,600 per cubic metre. Volume shares are Nilgiri 30%, Palghat Concrete 22%, Vindhya Bulk Concrete 18%, Chenab 12%, and four fringe producers at 4.5% each. The industry contribution margin is 28%. Market-level demand elasticity for ready-mix inside the catchment is estimated at −2.0; if site-mixed concrete is included in the market, the nearest substitute sits inside the boundary rather than outside it and the relevant elasticity falls to −1.2. Reach a verdict.

Solution. Step 1, check the table before using it. The eight shares sum to 100.0. A table that does not sum is the commonest source of a wrong index, and the check costs nothing.

Step 2, concentration before and after. Squaring and summing the shares gives HHI = 1,933.0 and CR4 = 82.0. After the merger the combined firm holds 30 + 12 = 42.0%, and the index becomes HHI = 2,653.0 with CR4 = 86.5.

Step 3, verify the change two ways. By subtraction, 2,653.0 − 1,933.0 = 720.0. By the closed form, 2 × s1 × s2 = 2 × 30 × 12 = 720.0. The two agree exactly, which confirms the share table and the arithmetic together.

Read the index a second way: 10,000 ÷ HHI is the number of equal-sized firms that would produce it. The market goes from 5.173306 equivalent firms to 3.769318, so the transaction removes about 1.4 equal competitors' worth of dispersion.

Step 4, test the market definition before trusting any of that. With a tested price rise of 5% and a contribution margin of 28%, the critical loss is CL = 5 / (5 + 28) = 15.151515%. At the candidate market's elasticity of −2.0, the actual loss is 2.0 × 5 = 10.0000%, which is smaller, so a hypothetical monopolist over ready-mix alone could profitably raise price and the narrow market stands. The elasticity at which the test exactly flips is 3.030303, so the verdict is not close but it is not safe either.

Step 5, price the alternative definition, because the parties will argue for it. Suppose site-mixed concrete belongs in the market, ready-mix is 70% of all concrete poured in the catchment, and the site-mix half is supplied by twenty contractors at 1.5% each. Every ready-mix share scales by 0.70, and the index falls to 992.17 before the merger and 1,344.97 after, a change of 352.80 again confirmed by 2 × 21.0 × 8.4. At the wide market's own elasticity a 5% rise loses 1.2 × 5 = 6.0000% of sales, inside the same 15.151515% critical loss, so the wide market passes the test as well. The SSNIP picks the smallest set of products that passes, which is why the narrow definition is the one the authority should be defending. On the narrow definition the transaction sits above every structural screen in use; on the wide one it sits below the unconcentrated threshold entirely. One definitional judgment, two opposite answers, the same two plants.

Step 6, screen for unilateral effects. Assume diversion proportional to the shares of the remaining sellers, and equal prices across a homogeneous product. Of the customers who would leave Nilgiri after a price rise, 12 / (100 − 30) = 17.142857% go to Chenab, so Nilgiri's upward pricing pressure is 0.171429 × 0.28 = 4.80%. Running it the other way, 30 / (100 − 12) = 34.090909% of Chenab's lost customers go to Nilgiri, giving Chenab a pressure index of 9.545455%. Weighted by the two firms' shares, the merged entity's index is 6.155844%. Under linear demand the illustrative price rise is roughly half the index, so about 3.077922%.

Step 7, size the harm the remedy has to fix. The catchment is worth 4.2m × ₹5,600 = ₹2,352.0 crore a year, of which the merged firm holds 42%, or ₹987.84 crore. A 3.077922% price rise on that base transfers ₹30.404945 crore a year from buyers. Over five years at 11% the annuity factor is 3.695897, so the present value of the transfer is ₹112.373547 crore.

Step 8, the behavioural remedy, priced honestly. The parties offer a five-year delivered-price cap escalating only with inflation. Assume, as the authority must, that the cap is enforced with probability 0.70. Expected harm prevented is 112.373547 × 0.70 = ₹78.661483 crore; expected residual harm is ₹33.712064 crore. Now the uncomfortable part. The concentration index after the remedy is still 2,653.0, the change is still 720.0, and the SSNIP answer is still that the narrow market stands. A behavioural remedy moves neither, because both are structural and definitional instruments and a promise is neither.

Step 9, the structural remedy, for contrast. Divest batching plants representing 5 share points to a capable new entrant. The merged firm falls to 37.0%, the index falls to 2,283.0, and the change against the pre-merger position falls to 350.0. The divestiture removes 370.0 index points, needs no monitor, and is verifiable at closing. If the authority wanted the change brought down to 100, the required divestiture solves (42 − d)² + d² − 1,044 = 100, giving d = 9.554477 share points and a merged share of 32.445523%. Note the floor: even divesting all 12 of Chenab's points to a new entrant only returns the index to 1,933.0, its pre-merger level, because the market was already concentrated before anybody proposed anything.

Read it. The decision that mattered was made in Step 4, and everything after it was arithmetic.

Regulators and defence economists fight over market definition rather than over index thresholds precisely because the definition is worth more. The second lesson is the behavioural one: a remedy that leaves every structural measure unchanged is worth its enforcement probability and nothing else, so a monitor with no budget converts a remedy into a press release.

Worked example 2: A tariff shock in a US model (US, $)

Allegheny Fastener Corp (synthetic) makes industrial fasteners. Revenue is $1,800.0m, cost of goods sold is $1,260.0m and fully variable, selling and administrative expense is $360.0m and fixed, interest expense is $24.0m, the tax rate is 23.0%, and 60.0m shares are outstanding. Imported inputs are 45% of cost of goods sold. A 25-percentage-point tariff is imposed on those inputs. Demand elasticity is −0.9. Price the shock at pass-through of 0%, 50% and 100%, then convert the residual uncertainty into a valuation adjustment.

Solution. Step 1, the base year. EBIT is 1,800.0 − 1,260.0 − 360.0 = $180.0m, a 10.0% margin. Pre-tax income is $156.0m, net income $120.12m, EPS $2.002000.

Step 2, the raw cost of the tariff. Imported inputs are 1,260.0 × 0.45 = $567.0m; domestic inputs are $693.0m. The tariff adds 567.0 × 0.25 = $141.75m, so cost of goods sold at unchanged volume becomes $1,401.75m. Unit cost rises 11.25%, which is 7.875% of revenue. Check: 693.0 + 708.75 = 1,401.75, and 1,401.75 − 1,260.0 = 141.75.

Step 3, run the three pass-through cases. Passing through a fraction θ raises the price index by θ × 7.875% and, at an elasticity of 0.9, cuts the volume index by 0.9 × θ × 7.875%. Revenue is 1,800.0 × price × volume; cost of goods sold scales with volume alone.

At θ = 0 the EBIT fall is exactly the tariff, 180.0 − 38.25 = 141.75, which is the arithmetic check on the whole table. EPS falls 90.8654%, 56.0895% and 24.5337% across the three cases.

Step 4, weight the scenarios. Judgement, stated openly: 25% weight on no pass-through, 45% on half, 30% on full. Weights sum to 1.00. Probability-weighted EBIT is $93.705922m and probability-weighted EPS is $0.894559.

Step 5, the wrong way to price it, done anyway so the size is visible. Value the base EPS as a growing perpetuity at a 9.0% required return and 2.0% growth: 2.002000 / 0.07 = $28.600000 a share. The weighted EPS gives $12.779419, a 55.32% fall. The discount rate reproducing that value from the base EPS is 0.02 + 2.002000 / 12.779419 = 17.665814%, a premium of 866.5814 basis points. Nobody would write that premium down, because it silently assumes the tariff is permanent.

Step 6, the right way. Assume the tariff binds for three years and then lapses. The present value of three years of the weighted EPS at 9.0% is 0.894559 × 2.531295 = $2.264393, and the present value of reverting to the base valuation after three years is 28.600000 / 1.09³ = $22.084448. Total value is $24.348841, a fall of 14.864193%. The equivalent discount rate is now 10.222157%, a premium of 122.2157 basis points.

Read it. Two numbers describe the same shock: 866.58 basis points and 122.22 basis points.

Only the duration assumption differs, and the discount-rate route hides it while the scenario route forces you to state it. The pass-through row is where the real analysis lives, since the spread between θ = 0 and θ = 1 is 24.53% versus 90.87% of EPS, and every hour you spend on the pass-through question is worth more than every hour spent arguing about the discount rate.

Worked example 3: Two labour-force panels (US and India)

Two panels, both illustrative and shaped like the surveys they imitate. Compute the standard measures, then move people between the boxes and watch the headline reverse.

The first panel is American, in thousands of persons aged 16 and over. Civilian non-institutional population 268,400; employed 161,900; unemployed 6,900; not in the labour force 99,600. Marginally attached 1,550, of whom 520 are discouraged. Part-time for economic reasons 4,300.

Solution. Step 1, check the partition. 161,900 + 6,900 + 99,600 = 268,400, so the three boxes account for everybody exactly once.

Step 2, the four measures. Labour force = 168,800. Participation rate = 168,800 / 268,400 = 62.891207%. Unemployment rate = 6,900 / 168,800 = 4.087678%. Employment-population ratio = 161,900 / 268,400 = 60.320417%.

Step 3, verify by identity. 62.891207% × (1 − 0.04087678) = 60.320417%, matching the direct computation. Any panel failing this has a definitional problem.

Step 4, the broad measure. (6,900 + 1,550 + 4,300) / (168,800 + 1,550) = 12,750 / 170,350 = 7.484591%, which is 3.396913 points above the headline. The marginally attached enter both numerator and denominator; the involuntary part-timers enter the numerator only, since they are already counted in the labour force as employed.

Step 5, move the boxes. Suppose all 1,550 marginally attached workers start searching. Nobody lost a job and nobody gained one. The labour force rises to 170,350, the unemployed count rises to 8,450, the unemployment rate rises to 4.960376%, and participation rises to 63.468703%. The headline worsened by 0.872698 points on unambiguously good news.

The second panel is Indian, in millions of persons aged 15 and over. Population 1,040.0; labour force 452.0; employed 437.0. Female population 508.0 with a participation rate of 25.0%. Of the employed, 61.0 million report wanting more work than they get.

Solution. Step 1, the measures. Unemployed = 15.0m. Participation = 452.0 / 1,040.0 = 43.461538%. Unemployment rate = 15.0 / 452.0 = 3.318584%. Employment-population ratio = 42.019231%, which again reconciles to participation times one minus the unemployment rate. Underemployment among the employed is 61.0 / 437.0 = 13.958810%.

Step 2, split participation by sex. Female labour force = 508.0 × 0.25 = 127.0m, so the male labour force is 325.0m on a male population of 532.0m, a male participation rate of 61.090226%. The aggregate rate of 43.46% is the average of 25.0% and 61.09%, weighted by population.

Step 3, the counterfactual that reverses the headline. Suppose female participation rises to 40.0%. That is 76.2m new entrants. If employment grows enough to absorb half of them, the labour force becomes 528.2m (participation 50.788462%), employment becomes 475.1m, and the unemployed count becomes 53.1m. The unemployment rate is now 10.053010%, three times the starting figure, while the employment-population ratio has risen 3.663462 points to 45.682692%.

Read it. A 3.32% unemployment rate on 43.46% participation and 13.96% underemployment is a labour market with a serious problem and a flattering headline. The counterfactual is the reason: the single most valuable thing that could happen to this economy would triple its unemployment rate on the way through.

Quote the participation rate and the employment-population ratio beside every unemployment rate, or quote none of them.

Worked example 4: Okun's relationship, run both ways (US, $)

An economy has a natural rate of unemployment of 4.2%, an observed unemployment rate of 6.1%, potential output of $23,400bn, trend growth of 2.0%, and an estimated Okun coefficient of c = 2.0 in gap form and β = 0.5 in difference form.

Solution. Step 1, the gap. The unemployment gap is 6.1 − 4.2 = 1.9 percentage points, all of it cyclical by construction. The output gap is −2.0 × 1.9 = −3.80%.

Step 2, in currency. Actual output is 23,400 × (1 − 0.038) = $22,510.8bn, so lost output is $889.2bn a year.

Step 3, the parameter risk, computed rather than asserted. At c = 1.6 the gap is −3.04% and the loss is $711.36bn. At c = 2.4 the gap is −4.56% and the loss is $1,067.04bn. The spread across a plausible range of one parameter is $355.68bn, which is 40% of the central estimate.

Step 4, the difference form, forwards. If real output grows 4.4% next year against a 2.0% trend, unemployment changes by −0.5 × (4.4 − 2.0) = −1.20 points, taking the rate to 4.90%.

Step 5, the difference form, backwards. To close the whole 1.9-point gap in one year requires 2.0 + 1.9 / 0.5 = 5.80% growth. Spread over two years it requires 3.90% a year. Check: at 5.80% growth the implied change is −0.5 × (5.80 − 2.0) = −1.90 points, exactly closing the gap.

Read it. Okun's relationship bridges an output forecast and a labour forecast, and its main use is consistency checking rather than prediction. A forecast pairing 2.5% growth with a 1.5-point fall in unemployment is internally inconsistent at any plausible coefficient, and spotting that costs one line of arithmetic.

The relationship weakens sharply where a demand shock is absorbed as underemployment rather than as unemployment, which is exactly the situation the second panel above describes.

Worked example 5: A kinked demand curve, solved (India, ₹)

Tungabhadra Cables (synthetic) is one of three producers of low-tension power cable in a regional market. The prevailing price is ₹400 per metre and the firm sells 60 thousand metres a month. Raise price alone and rivals hold, so the demand it faces has slope b = 2.0. Cut price and rivals match, so the slope is b = 6.0. Find the marginal-revenue discontinuity, then the range of marginal cost over which the price does not move, then what happens outside that range.

Solution. Step 1, write both segments through the kink. Upper segment (price rise, rivals hold): P = 520 − 2q, since 400 + 2 × 60 = 520. Lower segment (price cut, rivals match): P = 760 − 6q, since 400 + 6 × 60 = 760. Both pass through (60, 400), which is the check.

Step 2, elasticity at the kink, to confirm the asymmetry. On the upper segment, E = (−1/2)(400/60) = −3.333333; on the lower, E = (−1/6)(400/60) = −1.111111. The ratio is exactly 3.0, which equals the ratio of the two slopes, as it must. Demand is three times more elastic for a price rise than for a price cut.

Step 3, the marginal-revenue discontinuity. Marginal revenue on the upper segment is 520 − 4q, which at q = 60 equals ₹280. On the lower segment it is 760 − 12q, which at q = 60 equals ₹40. Marginal revenue therefore falls vertically from ₹280 to ₹40 at the kink, a gap of ₹240 per metre.

Step 4, the rigidity range. Any marginal cost inside [₹40, ₹280] leaves the optimum at q = 60 and P = ₹400. Marginal cost can rise seven-fold across that range with the price not moving one rupee.

Step 5, what moves inside the range, which is the part people miss. At MC = ₹150, monthly contribution is (400 − 150) × 60 = 15,000 thousand rupees, or ₹1.500000 crore. At MC = ₹260 it is (400 − 260) × 60 = 8,400 thousand rupees, or ₹0.840000 crore. Contribution fell ₹0.660000 crore a month, or 44%, with the price list unchanged and volume unchanged.

Step 6, step outside the range in both directions. At MC = ₹300, above the gap, the firm optimises on the upper segment: 520 − 4q = 300 gives q = 55.0 and P = ₹410.00, a rise of ₹10 and a volume loss of 5 thousand metres. At MC = ₹30, below the gap, it optimises on the lower segment: 760 − 12q = 30 gives q = 60.833333 and P = ₹395.00, a cut of ₹5.

Read it. Price rigidity is a statement about marginal cost sitting inside a gap, not about a business being stable. Earnings moved 44% inside the range.

Two limits stay attached to this model wherever you use it. It explains why an existing price is sticky and cannot say how the price got there, and George Stigler's 1947 test across seven American oligopolies found rivals did not respond with the asymmetry it assumes.

Worked example 6: Stackelberg, Cournot, Bertrand and a dominant firm on one demand system (India, ₹)

Take the same cement demand system solved earlier for Cournot: inverse demand P = 600 − Q, with Q in thousand tonnes a month and P in rupees per tonne, and constant marginal cost of ₹300 per tonne. Solve the sequential game, compare all four market structures on price, quantity, profit and surplus, and then solve a dominant firm facing a competitive fringe on the same demand.

Solution. Step 1, the follower's reaction function. Given the leader's q_L, the follower maximises (600 − q_L − q_F − 300) q_F, so q_F = (300 − q_L)/2. Identical to the Cournot reaction function, because the follower is in a Cournot position once the leader has moved.

Step 2, the leader substitutes and maximises. With q_F = (300 − q_L)/2, market price becomes P = 600 − q_L − (300 − q_L)/2 = 450 − q_L/2. Leader profit is (450 − q_L/2 − 300) q_L = (150 − q_L/2) q_L, maximised at q_L = 150.0. The follower then produces (300 − 150)/2 = 75.0.

Step 3, the outcome. Total quantity 225.0, price ₹375.00. Leader profit (375 − 300) × 150 = 11,250 thousand rupees = ₹1.125000 crore; follower profit 75 × 75 = 5,625 thousand rupees = ₹0.562500 crore.

Step 4, check against the closed forms. With a − c = 300 and b = 1: π_L = (a−c)²/(8b) = 90,000/8 = 11,250 and π_F = (a−c)²/(16b) = 5,625. Both match. The Cournot benchmark (a−c)²/(9b) = 10,000 thousand rupees, or ₹1.000000 crore each, matches the figure already established for this market. The leader's profit is also exactly half the monopoly profit of ₹2.25 crore, which is a property of the linear symmetric case rather than an accident of these numbers.

Step 5, the full comparison.

Relative to Cournot, the leader gains 12.50% and the follower loses 43.75%; the leader earns exactly 2.0 times the follower.

Step 6, the commitment test. The leader's Cournot best response to a follower producing 75 would be (300 − 75)/2 = 112.5, not 150. The leader is producing 37.5 thousand tonnes more than it would choose if it could reconsider, and that output is credible only because it is irreversible: capacity built, contracts signed, the kiln lit. Remove the irreversibility and both the equilibrium and the advantage collapse back to Cournot.

Step 7, concentration against welfare. Output shares under Cournot are 50/50, giving an index of 5,000.0. Under Stackelberg they are 66.6667/33.3333, giving 5,555.555556. Concentration rose 555.555556 points, price fell ₹25.00, and total surplus rose ₹0.21875 crore a month. Monopoly scores 10,000 with the worst surplus; a Bertrand duopoly scores 5,000 with the best. The index and the welfare ranking are not the same ordering.

Step 8, a dominant firm with a fringe on the same demand. Let a competitive fringe supply S_f = −180 + 0.6P for prices above its shut-down point, and let the dominant firm have the same ₹300 marginal cost. Residual demand is Q_d = (600 − P) − (−180 + 0.6P) = 780 − 1.6P, so P = 487.5 − 0.625 Q_d. Marginal revenue is 487.5 − 1.25 Q_d; setting it equal to 300 gives Q_d = 150.0 and P = ₹393.75. Fringe supply at that price is 56.25, total quantity is 206.25, and market demand at ₹393.75 is also 206.25, so the market clears. Dominant-firm profit is (393.75 − 300) × 150 = 14,062.5 thousand rupees = ₹1.406250 crore, on a share of 72.727273%, with a Lerner index of 0.238095. Fringe producer surplus is ₹0.263672 crore, computed both as a triangle and by integrating under the price line above the fringe supply curve, which agree exactly.

Read it. A 72.7% share and a Lerner index of 0.238 sit together comfortably, because it is the fringe's supply elasticity rather than the dominant firm's share that disciplines the price. Change the fringe slope and the share stays put while the markup moves.

Any merger screen that stops at the share table will read this market wrong in both directions.

Worked example 7: Rate-of-return regulation against a price cap (US, $)

Cumberland Valley Water Company (synthetic) is a regulated water utility. Its rate base is $1,200.0m, its allowed return is 7.20%, split half equity at a 9.60% allowed return on equity and half debt, operating expenses are $210.0m, depreciation is $60.0m, the corporate tax rate is 21.0%, and it delivers 62,000 million gallons a year. Build the revenue requirement and the tariff, then run one cost-saving project through two regulatory regimes.

Solution. Step 1, check the capital structure. An allowed return of 7.20% with half the capital at 9.60% implies a debt cost of (7.20 − 0.5 × 9.60) / 0.5 = 4.80%. Confirm: 0.5 × 9.60 + 0.5 × 4.80 = 7.20.

Step 2, build the revenue requirement, line by line. The allowed return is 0.072 × 1,200.0 = $86.40m. The equity portion of that return is 0.5 × 1,200.0 × 0.096 = $57.60m, which has to be grossed up for tax at 57.60 × 0.21 / (1 − 0.21) = $15.311392m. The revenue requirement is 210.0 + 60.0 + 15.311392 + 86.40 = $371.711392m.

Step 3, the tariff. Dividing by 62 million thousand-gallon units gives $5.995345 per thousand gallons.

Step 4, the Averch-Johnson incentive, made numerical. Management can install leak detection and metering for $40.0m of capital with a 20-year life, saving $12.0m a year of operating cost. Under rate-of-return regulation, the rate base rises to $1,240.0m, operating expense falls to $198.0m, depreciation rises to $62.0m, and the revenue requirement becomes $365.101772m, a tariff of $5.888738 and a customer saving of $6.609620m a year. The utility's allowed profit rises by 0.072 × 40.0 = $2.880000m a year.

Step 5, the same saving with no capital in it. Now suppose the identical $12.0m of annual saving is available from a purely operational change costing nothing. Under rate-of-return regulation the utility's allowed profit changes by exactly zero, because allowed profit is the return on the rate base and the rate base did not move. The whole saving is handed to customers at the next review. A regime that pays $2.88m for a capital-intensive fix and $0.00m for a free one has told management precisely what to do.

Step 6, the same saving under a five-year price cap. Fix the tariff for five years and the utility keeps the saving for the control period. At a 6.0% discount rate the five-year annuity factor is 4.212364, so the operational-only saving is worth 12.0 × 4.212364 = $50.548365m to the firm, against zero under rate-of-return regulation. The capital project clears too: 50.548365 − 40.0 = $10.548365m of value. At the next review the regulator resets the tariff to the achieved cost base, a one-off reduction of 12.0 / 371.711392 = 3.228311% of the revenue requirement.

Read it. The gap between $0.00m and $50.55m for the identical saving is the whole argument for incentive regulation, and it is arithmetic rather than ideology.

The counter-arguments are equally concrete. With revenue fixed, deferred maintenance is the fastest route to a margin, so a cap without enforceable service standards degrades the asset. And if the regulator resets the efficiency factor to claw back everything the firm outperformed, the firm learns in period one and stops trying in period two.

Worked example 8: A priced regulatory and geopolitical register (India, ₹)

Barmer Speciality Chemicals Ltd (synthetic) has revenue of ₹4,800 crore, an EBIT margin of 14.0%, a contribution margin of 22.0%, a tax rate of 25.17%, a cost of capital of 12.0% and long-run growth of 5.0%. Exports to the European Union are 34% of revenue. One customer whose end market falls under an export-control regime is 9% of revenue. One intermediate is single-sourced from one country. All manufacturing sits on one site. Build the register and price it.

Solution. Step 1, the base. EBIT is 4,800 × 0.14 = ₹672.00 crore. NOPAT is 672.00 × (1 − 0.2517) = ₹502.8576 crore, and the value of the firm at 12.0% and 5.0% growth is 502.8576 / 0.07 = ₹7,183.68 crore.

Step 2, price each row on stated assumptions.

Row 1, a carbon border levy on the EU-exported volume. Thematic, velocity in years, likelihood 0.85. EU revenue is ₹1,632.00 crore; a levy modelled at 4.2% of it is ₹68.544 crore, of which 60% is passed to customers, leaving an EBIT hit of ₹27.4176 crore. Expected: ₹23.30496 crore.

Row 2, the export-control customer is listed and the revenue is lost. Event, velocity in days, likelihood 0.12. Revenue at risk is ₹432.00 crore; at a 22% contribution margin the EBIT hit is ₹95.04 crore. Expected: ₹11.40480 crore.

Row 3, the single-sourced intermediate is cut off for 45 days. Event, velocity in days, likelihood 0.08. Lost contribution is 4,800 × 45/365 × 0.22 = ₹130.191781 crore. Expected: ₹10.415342 crore.

Row 4, the single site is out for 30 days from a natural hazard. Exogenous, velocity in days, likelihood 0.05. Lost contribution is 4,800 × 30/365 × 0.22 = ₹86.794521 crore. Expected: ₹4.339726 crore.

Step 3, read the two totals separately. The expected annual drag is ₹49.464828 crore, which is 7.360838% of EBIT. If all four fired in the same year the drag would be ₹339.443901 crore, or 50.512485% of EBIT. The first number belongs in the base case; the second belongs in a stress test, and confusing them is a common way to double-count.

Step 4, put it in the model. Reduce EBIT by the expected drag: NOPAT becomes ₹465.843069 crore and value becomes ₹6,654.900983 crore, a fall of 7.360838%, identical to the EBIT percentage because the valuation is linear in NOPAT here. The discount rate that would reproduce that value from the unadjusted NOPAT is 12.556200%, a premium of 55.6200 basis points.

Read it. Quote the 55.62 basis points so a reader can judge the size, then put the effect in the cash flows, because three of these four risks are idiosyncratic and a discount-rate premium prices risk as though it were systematic.

Velocity is the column that changes behaviour rather than valuation. Rows 2, 3 and 4 have velocities measured in days, so there is no time to act after the trigger and the only responses are ex ante: dual sourcing, insurance, position size. Row 1 has a velocity in years, so it is a re-underwriting problem rather than an emergency, and it also carries the largest expected number on the page.

Verify block

Every figure printed above, restated at full precision, so any of it can be reproduced or falsified.

Practice set

Work each problem before reading its solution. A calculator is assumed and answers within about ±2% score as correct unless the problem says otherwise. Every entity is synthetic and every panel illustrative but internally consistent.

P1 (guided). Concentration, computed and checked. A US market has seven firms with shares 38, 24, 14, 9, 6, 5 and 4. The largest proposes to acquire the fourth largest. Compute the index before and after, the change by two routes, and the four-firm ratio before and after.

Solution. Shares sum to 100.0, which is the first check. Before: 38² + 24² + 14² + 9² + 6² + 5² + 4² = 2,374.0, with CR4 = 85.0. After, the merged firm holds 47: 47² + 24² + 14² + 6² + 5² + 4² = 3,058.0, with CR4 = 91.0. The change is 3,058.0 − 2,374.0 = 684.0 by subtraction and 2 × 38 × 9 = 684.0 by the closed form. Agreement confirms both the table and the arithmetic.

P2 (guided). Critical loss and the break-even elasticity. A candidate market has a contribution margin of 55%. (a) Compute the critical loss for a 5% price rise and for a 10% one. (b) Compute the break-even elasticity in each case. (c) At an estimated market elasticity of −1.1, does the candidate market survive a 5% SSNIP?

Solution. (a) CL = t/(t+m). At t = 5: 5/60 = 8.333333%. At t = 10: 10/65 = 15.384615%. (b) The break-even elasticity is the critical loss divided by t: 1.666667 at 5% and 1.538462 at 10%. (c) The actual loss is 1.1 × 5 = 5.5%, which is below the 8.333333% critical loss, so the price rise is profitable and the candidate market stands. Note the direction: a higher margin makes the critical loss smaller, so high-margin businesses need very little customer loss to make a price rise unprofitable, which is why margin is an input to market definition rather than a distraction from it.

P3 (guided). A unilateral-effects screen. Firm 1 and firm 2 propose to merge. The diversion ratio from firm 1 to firm 2 is 0.22, firm 2's contribution margin is 34%, and firm 2's price is 1.15 times firm 1's. Compute firm 1's upward pricing pressure index, and recompute it if the two prices were equal.

Solution. GUPPI = D × m2 × (P2/P1) = 0.22 × 0.34 × 1.15 = 8.602%. At equal prices, 0.22 × 0.34 = 7.480%. Both sit above the 5% level often used as a warning threshold. The price ratio matters because the index converts a lost sale on firm 1's price base into a recaptured margin on firm 2's, so a partner selling a dearer product creates more pricing pressure for the same diversion.

P4. A labour panel from scratch. A country's population aged 15 and over is 96.4 million. Employment is 55.2 million and unemployment is 3.1 million. Compute the labour force, the participation rate, the unemployment rate and the employment-population ratio, then verify by identity.

Solution. Labour force = 55.2 + 3.1 = 58.3m. Participation = 58.3/96.4 = 60.477178%. Unemployment rate = 3.1/58.3 = 5.317324%. Employment-population ratio = 55.2/96.4 = 57.261411%. Identity check: 60.477178% × (1 − 0.05317324) = 57.261411%, matching to the last displayed digit.

P5. A broad measure. On the same panel, 2.4 million people are marginally attached and 6.8 million are working part-time for economic reasons. Compute the broad rate and the gap to the headline.

Solution. (3.1 + 2.4 + 6.8) / (58.3 + 2.4) = 12.3/60.7 = 20.263591%, against a headline of 5.317324%, a gap of 14.946267 points. The marginally attached appear in both numerator and denominator; the involuntary part-timers appear only in the numerator, because they are already inside the labour force as employed.

P6. The denominator moves. On the original panel, 1.8 million unemployed people give up searching and leave the labour force. Compute the new unemployment rate and participation rate, and state what a reader of the headline alone would conclude.

Solution. Labour force falls to 56.5m and the unemployed count to 1.3m. The unemployment rate falls to 1.3/56.5 = 2.300885%, down 3.016439 points, and participation falls to 56.5/96.4 = 58.609959%. A reader of the headline alone would conclude the labour market improved dramatically, when 1.8 million people stopped looking and nobody found work. Employment is unchanged at 55.2m, which is the tell.

P7. Okun in both directions. An economy has an Okun coefficient of 1.8 in gap form, a natural rate of 4.6%, and an observed rate of 7.4%. Separately, its difference-form coefficient is 0.4 and trend growth is 2.5%. (a) Compute the output gap. (b) Compute the growth rate needed to cut unemployment by 1.5 points in a year.

Solution. (a) −1.8 × (7.4 − 4.6) = −5.04%. (b) 2.5 + 1.5/0.4 = 6.25%. The second answer is a useful reality test: an economy needing 6.25% growth to cut unemployment 1.5 points is being told, in arithmetic, that its stated unemployment target is not reachable in one year at any plausible growth rate.

P8. Classify five situations. Assign each to frictional, structural or cyclical unemployment and name the policy instrument that addresses it. (a) A software engineer quits in March and starts a new job in June. (b) A port city's container-handling jobs are automated and the displaced crane operators are 50 years old. (c) Construction employment falls 18% in a recession and recovers fully two years later. (d) Graduates take four months to find a first job in a normal year. (e) A textile cluster's mills close permanently as production moves offshore.

Solution. (a) Frictional; the instrument is information and matching, and this is a healthy market. (b) Structural; the instrument is retraining and relocation, and the age of the displaced workers is why it is expensive. (c) Cyclical; the instrument is demand management, and the full recovery is the diagnostic. (d) Frictional; the qualifier "in a normal year" places it, since the same delay in a downturn would be partly cyclical. (e) Structural; permanence is the marker, and stimulating demand will not reopen the mills. The natural rate is the sum of (a), (d) and (e), and only (c) responds to a central bank.

P9. Find the kink. A firm sells 40 thousand units a month at ₹250. Above the current price its demand slope is 1.5; below it, 5.0. Write both segments, compute marginal revenue on each at the kink, and state the rigidity range.

Solution. Upper segment: P = 250 + 1.5 × 40 − 1.5q = 310 − 1.5q. Lower segment: P = 250 + 5.0 × 40 − 5.0q = 450 − 5.0q. Marginal revenue is 310 − 3.0q = ₹190 just below the kink and 450 − 10.0q = ₹50 just above it, so the discontinuity is ₹140 and the rigidity range for marginal cost is [₹50, ₹190].

P10. Step outside the range. On the same demand system, marginal cost rises to ₹220. Compute the new quantity and price.

Solution. ₹220 is above the top of the range, so the firm optimises on the upper segment: 310 − 3.0q = 220 gives q = 30.0 and P = 310 − 1.5 × 30 = ₹265.00, a rise of ₹15 and a volume loss of 10 thousand units. Check that the answer lies on the correct branch: q = 30 is below the kink quantity of 40, as a price rise requires.

P11. Stackelberg against Cournot. Inverse demand is P = 240 − 2Q and marginal cost is 40 for both firms. Compute the Cournot equilibrium and the Stackelberg equilibrium, then verify both against the closed forms.

Solution. With a − c = 200 and b = 2: Cournot quantity per firm is 200/(3 × 2) = 33.333333, total 66.666667, price 106.666667, profit each (106.666667 − 40) × 33.333333 = 2,222.222222. Stackelberg: leader 200/(2 × 2) = 50.0, follower (200 − 2 × 50)/(2 × 2) = 25.0, total 75.0, price 90.0, leader profit 2,500.0, follower profit 1,250.0. Closed-form checks: (a−c)²/(9b) = 2,222.222222, (a−c)²/(8b) = 2,500.0, (a−c)²/(16b) = 1,250.0, all three matching. The leader gains 12.5% against Cournot and the follower loses 43.75%, the same percentages as the cement system, because those ratios depend only on the linear symmetric structure.

P12. A dominant firm with a fringe. On the same demand P = 240 − 2Q, a competitive fringe supplies S_f = −10 + 0.5P. The dominant firm's marginal cost is 40. Find the price, the dominant firm's output, fringe output and the dominant firm's share.

Solution. Market demand in quantity form is Q = 120 − 0.5P. Residual demand is Q_d = 120 − 0.5P − (−10 + 0.5P) = 130 − P, so P = 130 − Q_d. Marginal revenue is 130 − 2Q_d; setting it equal to 40 gives Q_d = 45.0 and P = 85.0. Fringe supply is −10 + 0.5 × 85 = 32.5, total quantity 77.5, and market demand at 85 is 120 − 42.5 = 77.5, so the market clears. Dominant-firm profit is (85 − 40) × 45 = 2,025.0, its share is 45/77.5 = 58.064516%, and fringe producer surplus is 1,056.25.

P13. Build a revenue requirement. A regulated distributor has a rate base of $640m, operating expenses of $118m, depreciation of $32m and an allowed return of 6.8%. Ignore the tax allowance. Compute the revenue requirement, then compute the change if the firm adds $50m to the rate base.

Solution. Allowed return is 0.068 × 640 = $43.52m, so the revenue requirement is 118 + 32 + 43.52 = $193.52m. Adding $50m of rate base raises the allowed return by 0.068 × 50 = $3.40m, taking the requirement to $196.92m. The utility's allowed profit rose by exactly the return on the increment, which is the Averch-Johnson incentive in one line: the firm is paid for the asset, not for the service.

P14. Value a saving under a cap. A firm under a five-year price cap finds an operating saving of $9m a year that requires no capital. At a 7.0% discount rate, what is the saving worth to the firm over the control period, and what would it be worth under rate-of-return regulation?

Solution. The five-year annuity factor at 7.0% is 4.100197, so the saving is worth 9.0 × 4.100197 = $36.901777m under the cap. Under rate-of-return regulation it is worth $0.00, because allowed profit is a return on the rate base and the rate base did not change; the saving passes to customers at the next review. The gap between the two figures is the incentive the cap creates and the rate-of-return regime destroys.

P15. One pass-through case. Using the Allegheny Fastener structure from Worked Example 2, compute EBIT and EPS at a pass-through of 0.75.

Solution. The price index is 1 + 0.75 × 0.07875 = 1.059063 and the volume index is 1 − 0.9 × 0.75 × 0.07875 = 0.946844. Revenue is 1,800 × 1.0590625 × 0.94684375 = 1,804.980076 and cost of goods sold is 1,401.75 × 0.94684375 = 1,327.238227, so EBIT is $117.741850m and EPS is $1.203020, an EPS fall of 39.909071% against the base. The answer sits between the θ = 0.5 and θ = 1.0 rows of the worked example, as it must.

P16. Price a three-row register. A company has EBIT of ₹310 crore, a tax rate of 25.17%, a cost of capital of 11.5% and growth of 4.5%. Its register has three rows: a 30% chance of a ₹42 crore hit, a 10% chance of a ₹180 crore hit, and a 55% chance of a ₹16 crore hit. Compute the expected drag, the base and adjusted values, and the discount-rate equivalent.

Solution. Expected drag is 0.30 × 42 + 0.10 × 180 + 0.55 × 16 = ₹39.40 crore. Base NOPAT is 310 × 0.7483 = ₹231.973 crore, so base value is 231.973 / 0.07 = ₹3,313.90 crore. Adjusted NOPAT is (310 − 39.40) × 0.7483 = ₹202.489980 crore, giving ₹2,892.714 crore. The equivalent discount rate is 0.045 + 231.973/2,892.714 = 12.519217%, a premium of 101.9217 basis points. Quote the premium for scale and put the ₹39.40 crore in the cash flows, because two of the three rows are firm-specific and diversifiable.

Applied mini-project

Build a regulatory and geopolitical risk register for one company you already cover, and make every row move a number.

Pick a company whose filings you have read and whose model you have built or can rebuild. Something you already cover beats something novel, because the point of the exercise is the pricing rather than the research.

Part 1, the exposure map (2 tests). From the filing alone, build four tables: revenue by geography and by end-market; cost by origin, naming every disclosed single-source or sole-supplier dependency; every licence, approval or registration with its issuing authority and expiry; and every material asset with its physical location. Test 1: every percentage traces to a page number in the filing, and the geography split sums to 100% or you state what the residual is. Test 2: you produce a separate, explicit list of the four items above that the filing does not disclose. That second list is the real output of the part, because an undisclosed dependency is an unpriced one.

Part 2, the register (3 tests). Ten rows, each with the six columns the teaching section specifies: risk, class, velocity, likelihood, model line, size. Mix the classes; a register with ten event risks and no thematic ones has not been thought about. Test 3: every likelihood is a number between 0 and 1, not a word. Test 4: every size is a currency amount with its pass-through, duration and recovery assumptions stated on the row. Test 5: every model line names a line that exists in your model, and somebody else could find it.

Part 3, the pricing (2 tests). Compute the expected annual drag as the probability-weighted sum, and separately the all-fire stress case. Reduce the base case by the expected drag and revalue, then compute the discount-rate premium that would reproduce the same value from the unadjusted cash flows. Test 6: the premium you quote actually reproduces your adjusted value when fed back through the valuation. Test 7: you state in one sentence per row whether that row's risk is idiosyncratic or systematic, and therefore whether it belongs in the cash flows or the discount rate. A register where every row is systematic has not been examined.

Part 4, one counterfactual. Change one structural fact about the company: a second source for the single-sourced input, a second manufacturing site, or a geography reduced to half its share. Re-run the register and report what the expected drag becomes. This is the number that tells management what the mitigation is worth, and it is the one thing on the deliverable a chief financial officer would read twice.

Deliverable. The register on one page, plus a one-page memo. The memo carries the base and adjusted values, the expected drag as a percentage of EBIT, the stress case, and the counterfactual result. It also carries the discount-rate equivalent with one sentence on why you did not use it, and a dated note of every threshold or policy figure you relied on with the primary source you checked it against.

Rubric (20 points; pass at 16 with no zero on any row).

Any part whose test fails scores zero on that row regardless of how the page looks, and the reason is the same reason the whole register exists: a risk you cannot check is a risk you cannot improve at.

Reading & resources

Use every item below as an anchor to verify against, never as a source to lift a number from without tracing it yourself. Policy thresholds in all of them move; check the current text.

Regulation, theory and practice:

• George Stigler, The Theory of Economic Regulation (1971), the paper that put capture on an economic footing rather than a moral one. [Paid] [Advanced]
• Harvey Averch and Leland Johnson, Behavior of the Firm Under Regulatory Constraint (1962), the original statement of the capital bias in rate-of-return regulation. [Paid] [Advanced]
• Jean-Jacques Laffont and Jean Tirole, A Theory of Incentives in Procurement and Regulation (1993), the modern treatment of incentive regulation, price caps and the ratchet. [Paid] [Advanced]
• W. Kip Viscusi, Joseph Harrington and David Sappington, Economics of Regulation and Antitrust, the standard textbook and the best single ramp into the first half here. [Paid] [Intermediate]
• OECD regulatory policy material on regulatory impact assessment and administrative burden, free and full of worked national examples. [Free] [Intermediate]
• The US Office of Management and Budget's circular on regulatory analysis, the primary text behind American federal cost-benefit practice. [Free] [Intermediate]

Antitrust and merger review, primary sources:

• The US Department of Justice and Federal Trade Commission Merger Guidelines, which define the concentration screens and the unilateral-effects framework. The thresholds have been revised more than once; read the current edition rather than a summary. [Free] [Intermediate]
• The European Commission's guidelines on the assessment of horizontal mergers, which is where the diversion-ratio and upward-pricing-pressure reasoning is set out most clearly for a non-specialist. [Free] [Intermediate]
• The Competition Act 2002 as amended, and the Competition Commission of India's combination regulations, for the Indian notification thresholds, the deal-value test and the appreciable-adverse-effect factors. Those figures have changed recently; take them from the gazette text. [Free] [Intermediate]

Oligopoly completions:

• Paul Sweezy, Demand Under Conditions of Oligopoly (1939), and Robert Hall and Charles Hitch's paper of the same year, the two independent origins of the kinked demand curve. [Paid] [Advanced]
• George Stigler, The Kinked Oligopoly Demand Curve and Rigid Prices (1947), the empirical test that found against it, and the reason the model is taught with a caveat. [Paid] [Advanced]
• Heinrich von Stackelberg, Market Structure and Equilibrium (1934, English translation 2011), the source of the leader-follower solve. [Paid] [Advanced]
• Jean Tirole, The Theory of Industrial Organization (1988), the graduate reference for Stackelberg, dominant-firm pricing and merger analysis in one place. [Paid] [Advanced]

Labour markets, definitions and data:

• The US Bureau of Labor Statistics Handbook of Methods on labour force statistics, which defines employed, unemployed, marginally attached and discouraged precisely, and documents the U-1 through U-6 family. [Free] [Beginner]
• India's Periodic Labour Force Survey annual report and methodology annexe, for the usual-status and current-weekly-status definitions and why they answer differently on the same population. [Free] [Intermediate]
• The International Labour Organization's statistical standards, in particular the resolution defining work, employment and labour underutilisation, against which every national definition is measured. [Free] [Intermediate]
• Arthur Okun, Potential GNP: Its Measurement and Significance (1962), four pages, free, and worth reading in the original because the original is far more tentative than the textbooks citing it. [Free] [Intermediate]

Geopolitics:

• The Geopolitical Risk index of Dario Caldara and Matteo Iacoviello, with its long history and full documentation, free from the authors' own page. [Free] [Intermediate]
• The Economic Policy Uncertainty index of Scott Baker, Nicholas Bloom and Steven Davis, free, with country-level series. [Free] [Intermediate]
• The sanctions and export-control lists themselves, from the issuing authority rather than a vendor's summary, because the summary is a day behind and a designation takes effect on publication. [Free] [Intermediate]

One exercise beats any of the reading above. Take a merger order from any competition authority, find its share table, and recompute the concentration index and its change yourself before you read the authority's own figures. Then find the paragraph where the market was defined and ask what the index would have been on the definition the parties argued for instead. An hour of that teaches more about regulatory risk than a term of reading about it.

Teach it back & journal

Feynman prompt. Find somebody who reads company filings and explain, without using the word "regulation" more than twice, why the same ready-mix merger clears on one market definition and fails on another. Draw them the two share tables. Make them compute the change in the concentration index themselves, both by subtraction and by doubling the product of the two shares, and let them find the agreement rather than being told about it. Then ask them what they would pay for a five-year price commitment, and keep asking until they say the words "times the probability it is enforced". If they get there on their own, the behavioural-remedy idea is installed and will not come out.

For the second half, sit with the same person and a labour panel. Give them the three boxes, the four ratios and the identity connecting three of them, then move 1,550 thousand people from the third box into the second and ask them whether the economy got better or worse. Most people answer worse, because the unemployment rate went up. Let them sit with the fact that nobody lost a job. The moment they reach for the participation rate without being prompted is the moment the lesson has landed, and it is worth waiting for in silence.

Journal prompt. For four weeks, stop every time you meet the phrase "regulatory risk" or "geopolitical risk" in a broker note, a pitch, a board pack or an annual report. Before reading further, write down four things: the channel it would arrive through, its class as event, exogenous or thematic, its velocity, and the number in a model it would move. At the end of the month, answer in writing. Count how many of them named a channel, how many gave a probability rather than an adjective, and how many named a line rather than a mood. Where none of the four was answerable from the document, say what the writer actually meant and what you would have written instead.

Then write one page on the harder question. The register discipline asks you to put a number on something you cannot know, and the honest objection to it is that a made-up probability is worse than no probability at all, because it launders a guess into a calculation. Argue the objection properly, at its strongest, and then answer it. Your answer has to deal with the fact that the alternative is not "no number": the alternative is an implicit number buried in a valuation nobody can see or check.

Which of your own recent estimates would you be least willing to have scored a year from now, and what does your reluctance tell you about it?

This module's flashcards and mastery quiz are wired into the app: see the node's Quiz and Reviews.