Opening: Kevin Warsh on Inflation as a Choice

Context and Relevance

Professor Tantri opens Class 11 by playing a clip from an interview with Kevin Warsh, former Federal Reserve Board Governor and widely expected to be the next Federal Reserve Chair. The interview is from the Hoover Institution’s Uncommon Knowledge series.

Key Argument: Inflation Is a Choice

Warsh channels Milton Friedman’s core insight: inflation is always and everywhere a monetary phenomenon. The Federal Reserve was given sole responsibility for price stability by Congress, precisely so that no one else could be blamed.

“Inflation is a choice. The central bank can hit any inflation level that it wants. We might not like how they do it, but the idea that they should be blaming someone else strikes me as quite antithetical to good economic history.” — Kevin Warsh

The Distinction: Price Level Changes vs. Inflation

Warsh draws a crucial distinction that many commentators miss. A one-time price change in a specific good (due to supply disruptions, geopolitics, or pandemics) is not inflation. Inflation occurs when that one-time price change becomes self-fulfilling: higher prices beget higher prices, and the price level becomes unpredictable for households and businesses alike.

One-time price changes: Supply shocks like Putin’s invasion of Ukraine, pandemic-related supply chain disruptions, or crude oil spikes. These are relative price adjustments in a market economy.
True inflation: When a one-time price shock becomes embedded in expectations and spirals. This is the central bank’s responsibility to prevent.

Professor Tantri’s Commentary

Professor Tantri emphasizes that Warsh’s views are directly relevant because he is likely to shape future Federal Reserve policy. Understanding his intellectual framework helps anticipate where monetary policy is headed and its downstream effects on global markets and the Indian economy.

The Production Function

Setup: Y = AN

The medium-term model begins with a simple production function that strips away mathematical complexity to focus on core insights.

Y = A × N

Y: Total output (GDP)
A: Productivity parameter (output per worker)
N: Number of workers

For simplicity, Professor Tantri assumes A = 1, so Y = N. This means each worker produces exactly one unit of output. Two workers produce two units of GDP. This assumption is not realistic (diminishing returns, Cobb-Douglas functions, etc. could be used), but it keeps the math clean without changing the directional insights.

“I can make it A-squared and bring all kinds of stuff. But let’s keep it very simple. Y = AN. We are just trying to derive simple insights.”

The Wage Setting Equation

Derivation

The wage setting equation captures how workers negotiate wages. A worker’s wage demand depends on three factors:

W = Pᵉ × F(u, z)

W: Nominal wage
Pᵉ: Expected price level
u: Unemployment rate (negative effect on wages)
z: Catch-all institutional variable (positive effect on wages)

Understanding Each Component

Expected Price (Pᵉ): Positive Effect on Wages

Wages are always set in advance. Workers do not know the actual future price level when they negotiate, so they base their wage demands on what they expect prices to be. Higher expected prices naturally lead to higher wage demands, because workers want to maintain their purchasing power.

Unemployment Rate (u): Negative Effect on Wages

Higher unemployment depresses wage demands through two channels. First, there are many people willing to do the same job, increasing competition for positions. Second, if a worker demands too high a wage and loses their job, the high unemployment rate means it will take much longer to find another position. Both effects push wage demands down.

The z Variable: Positive Effect on Wages

This is the institutional framework within which labor markets operate. Think of it as the regulatory and structural environment: minimum wage laws, labor protections, firing restrictions, union strength, and other pro-worker regulations. Higher z pushes wage demands up.

The Expected Real Wage

Dividing both sides of the wage setting equation by expected price:

W / Pᵉ = F(u, z) = Expected Real Wage

This tells us what workers expect to receive in terms of actual purchasing power. Since each worker produces one unit of output (from Y = N with A = 1), the wage divided by expected price represents the expected real wage: how much goods and services the worker expects their wage to buy.

“So the wage I’m giving you is a real wage. Expected real wage. Absolutely correct answer.”

Important caveat: This expected real wage is what workers demand. It is not necessarily what will prevail in the market. Equilibrium is determined when both sides (workers and employers) agree. As Professor Tantri emphasizes: “The guy on the other side is hallucinating. It’s okay. Does not mean you have to give it.”

The Price Setting Equation

Derivation

The price setting equation comes from the employer’s perspective. Employers set prices based on the wages they pay plus a markup:

P = W × (1 + m)

P: Actual price level (not expected, since firms can adjust prices in near-real-time)
W: Wage paid to workers
m: Markup over costs

Why P and Not Pᵉ?

A key assumption: prices are far more flexible than wages. Wages, once fixed, take time to revise (you don’t renegotiate wages every 15 days). Product prices, however, can be revised frequently, some even on an auction basis in real time. This asymmetry is why the price setting equation uses actual P rather than expected Pᵉ.

The Markup (m)

The markup represents the firm’s profit margin over labor costs. It exists because markets are not perfectly competitive. In a theoretically perfect competitive market, m = 0. But perfect competition exists only on paper.

“Markup is a kind of a bad word. It’s possible only because the market is not competitive. In a perfectly competitive market, this m guy will be zero. But as the market becomes more and more competitive, the markup tends to zero. That’s a better way of learning about markup.”

The Actual Real Wage

Rearranging the price setting equation:

W / P = 1 / (1 + m) = Actual Real Wage

Numerical Example

Suppose a worker produces 1 kg of rice (since Y = N and A = 1). The price of rice is 40 rupees, but the worker’s wage is only 30 rupees.

Real wage (W/P): 30/40 = 0.75 kg of rice
This equals 1/(1+m): 0.75, which implies m = 33%
The employer pays the worker 75% of their output and retains 25% as markup/profit.

Natural Rate of Unemployment and Potential GDP

Definition

The natural rate of unemployment is the unemployment rate at which the expected price level equals the actual price level (Pᵉ = P). At this point, no one is surprised: workers get the real wage they expected, and employers get the markup they planned for.

Natural Rate of Unemployment (Uₙ): The unemployment rate where Pᵉ = P

Potential GDP (Yₙ): The GDP level that prevails at Uₙ

Graphical Derivation

Plot unemployment rate on the horizontal axis and real wage on the vertical axis:

Price setting line (horizontal): W/P = 1/(1+m). This is flat because the employer’s real wage offer does not depend on the unemployment rate. The employer simply says: here is what I can pay given my markup and market competition.
Wage setting curve (downward-sloping): W/Pᵉ = F(u, z). Workers demand higher real wages when unemployment is low (more bargaining power) and lower real wages when unemployment is high.

The intersection of these two lines determines the natural rate of unemployment. The market clears through unemployment: if workers demand wages above what employers will pay, some workers remain unemployed until the wage demands come down to the equilibrium level.

“The way the market will clear is some of the people who are asking this much wage will not get a job. It will keep them down. Like any other price clearing mechanism, it will settle.”

The z Variable: Institutional Framework and Real-World Applications

What z Represents

The z variable captures the entire institutional framework within which labor markets operate. It includes labor laws, minimum wage regulations, firing restrictions, union strength, and all pro-worker institutional features. Higher z shifts the wage setting curve upward: workers demand higher wages at every unemployment rate.

Critical Subtlety

Higher z pushes wage demands up, but this does not necessarily mean equilibrium wages will be higher. Because employers’ markup hasn’t changed and their technology hasn’t improved, the market’s way of clearing is through higher unemployment. The wage setting curve shifts up, but the price setting line stays flat. The new intersection occurs at a higher unemployment rate with the same real wage.

“Don’t misunderstand that France’s wage is higher than US. This is the role of z itself. Equilibrium may not necessarily lead to higher wages. z on its own pushes wages up. Does not necessarily mean equilibrium wage up.”

Case Study: Maharashtra vs. West Bengal

Professor Tantri references a well-cited paper by Amartya Lahiri (who taught at ISB) comparing Maharashtra and West Bengal. West Bengal was once far more economically advanced, but diverged sharply. The key difference: the z variable. West Bengal’s pro-labor institutional framework pushed wage demands higher, leading to higher equilibrium unemployment and slower economic development. This is one of the most robust findings in economics.

“This is one of certain results in economics which hold up as good as… very close to a natural science. You can predict reasonably well. Tomorrow, let’s say India makes wages $40 an hour. You can almost predict it will lead to higher unemployment. And eight out of ten times, it will.”

India’s Recent Labor Law Changes

Professor Tantri offers a nuanced analysis of India’s recent labor law amendments, cautioning against simplistic interpretations.

The Pro-Employer Claim

Threshold change: The threshold for retrenchment approvals moved from 100 to 300 workers. This is genuinely pro-employer, making it easier to adjust workforce size.

The Hidden Pro-Worker Element

PF base redefinition: The definition of “basic salary” for provident fund calculations has changed. Nearly the entire salary now counts as basic, dramatically increasing PF outflows for employers.
Infosys example: Infosys had to make an extra provision of ₹1,000 crores in one quarter.
IT sector impact: The IT industry alone provisioned approximately ₹45,000 crores due to this change.

Classical economists would argue that if costs increase, employers simply reduce salaries. But in the short run, you cannot reduce salaries. If PF is now a function of a bigger base, it’s a real cost increase that employers must absorb. The net effect of these reforms on the z variable is ambiguous: it could be pro-labor or pro-worker, and the political credit implications differ accordingly.

“If it turns out to be pro-worker, at least get political credit for it. If it turns out to be pro-worker, then economically, it’s gonna be bad. But you will also not get political credit because in the polity, you go to Economic Times and claim it’s a reform. It may not be a reform.”

AI and the Macroeconomy: An Open-Ended Discussion

AI’s Effect on the Model

A student raises whether AI affects both unemployment and the z variable. Professor Tantri engages in an extended discussion about AI’s macroeconomic implications, acknowledging that conventional economic frameworks may be inadequate for analyzing AI’s impact.

Possibility 1: Markup Effects

For AI creators: If AI development requires massive CapEx that not everyone has access to, the big creators of AI will see their markups increase (m rises). Less competition at the frontier means more pricing power.
For AI users (services companies): Markup decreases as AI commoditizes services. The price setting line shifts up (W/P = 1/(1+m) increases as m falls), implying higher real wages at the same unemployment rate.

Possibility 2: Demand-Side Uncertainty

Even if markups fall for users, AI-driven layoffs could shift the wage setting curve itself. If workers are replaced en masse, the demand for labor changes in unpredictable ways. Unlike conventional technology, AI threatens to replace all jobs, not just some categories.

“No technology was like this. History is a very bad comparison for AI because there was no technology which was threatening to replace all jobs of everyone.”

The Larry Summers Argument: 25% GDP Growth?

Professor Tantri references Larry Summers’ now-famous argument. Before the Industrial Revolution, GDP grew at 0.2% annually. The Industrial Revolution made it 2% (a 100x increase). If AI can compress 100 years of research into 5 years, the US could theoretically grow at 25% per year. The question is distribution: who captures this growth?

The Counter-Argument: AI Making Humans Dumber

Professor Tantri shares research he is developing with Aditi Kolekar, arguing that AI’s long-term effects may be negative for growth through an unexpected channel.

The learning problem: AI currently learns from what humans have already produced. It is not generating fundamentally new knowledge.
The deskilling effect: Evidence suggests AI is making people less inclined to learn foundational skills. Why learn calculus if you can press a button? Why learn physics or modeling?
The pipeline problem: Brilliant scientists (Newton, Einstein, Hawking) emerge from systems where millions of people attempt and fail. You cannot ex-ante identify who will make breakthroughs. If AI reduces the number of people attempting deep learning and original research, the pipeline of potential geniuses dries up.
The long-run risk: AI may boost growth for 30-40 years by accelerating existing knowledge. But over 200 years, it could leave us worse off if it cannot generate fundamentally new knowledge on its own and has simultaneously destroyed the human capacity to do so.

“If you don’t have a system where lakhs and lakhs of people try, you will not get these brilliant people. For brilliant people to come out, you need the system where millions of people make an attempt. That is for sure.”

The $1 Trillion Hole

Wearing a finance hat, Professor Tantri notes that projected AI investment (trillions of dollars of CapEx) currently exceeds projected revenue by approximately $1 trillion. If this gap becomes apparent, it could trigger significant market corrections. His best guess: a phase of negativity and bubble-bursting, followed eventually by AI delivering real value, similar to the dotcom cycle.

“If you know that you put in 5 trillion and the whole revenue is gonna be 4 trillion, then you want to get out before the rest of the world realizes.”

Current Evidence

As of the lecture date, AI has not yet appeared in aggregate productivity statistics. This echoes Robert Solow’s famous 1987 observation about computers: “You can see the computer age everywhere but in the productivity statistics.” By the late 1990s, the internet did show up as roughly a 1 percentage point productivity gain. AI may follow a similar pattern.

Looking Forward: The Phillips Curve

Professor Tantri concludes by previewing the next session’s topic: the formal derivation of the Phillips curve from the wage setting and price setting equations developed in this lecture.

“Phillips curve is one of those discoveries in economics which is as applied as gravity. Everybody uses it. We’ll formally derive the Phillips curve with the toy model. And that is where the use of toy model is.”

The Phillips curve will connect the natural rate framework to inflation dynamics, revisit real interest rates, and bring the analysis full circle back to monetary policy, completing the medium-term economics toolkit.

Key Takeaways

Inflation is a central bank choice. One-time price changes (from supply shocks, geopolitics, or pandemics) are not inflation. Inflation occurs when those changes become self-fulfilling in expectations. The Federal Reserve can hit any inflation target it wants.
The Wage Setting Equation W = Pᵉ × F(u, z) captures how workers negotiate: higher expected prices raise wage demands, higher unemployment lowers them, and stronger labor protections (z) raise them.
The Price Setting Equation P = W × (1 + m) captures employer behavior: prices reflect wages plus a markup determined by market competition.
The natural rate of unemployment is where expected and actual prices converge (Pᵉ = P). At this point, the labor market is in equilibrium and no one is surprised. GDP at this unemployment rate is potential GDP.
Increasing z (pro-labor institutions) raises unemployment, not wages. When labor laws strengthen, workers demand more, but employers don’t change their markup. The market clears through higher unemployment. Maharashtra vs. West Bengal illustrates this powerfully.
India’s labor reforms are more ambiguous than reported. The retrenchment threshold change (100→300) is pro-employer, but the PF base redefinition significantly increases employer costs. The net effect on z is unclear.
AI’s macroeconomic impact is genuinely uncertain. It could reduce markups, boost productivity, and create unprecedented growth. Or it could destroy the human learning pipeline needed for future breakthroughs, leaving us worse off in the very long run.
Simple toy models yield powerful insights. The Y = N production function, combined with wage and price setting equations, builds a complete framework for understanding inflation, unemployment, and monetary policy in the medium term.