The Quiet Logic of Monetary Policy

Monetary policy has a public relations problem. It is, at once, the most powerful lever a modern state routinely pulls and the one most people understand least. Interest rates go up, mortgages get expensive, someone on television looks grave — and that's the whole story most of us carry. The actual mechanism is more interesting, and more uncertain, than the headlines suggest.

The textbook channel

The standard story runs through the interest rate channel. A central bank sets a short-term policy rate; that rate propagates, imperfectly, along the yield curve; borrowing gets cheaper or dearer; spending adjusts; output and inflation follow. Formally, a stripped-down New Keynesian framework gives us three equations:

where $\pi_t$ is inflation, $\tilde{y}_t$ is the output gap, $i_t$ is the nominal policy rate, and $r_t^n$ is the natural real rate. The first is the Phillips curve; the second is the IS curve with forward-looking expectations; the third is a Taylor-type rule.

The intuition matters more than the algebra. Expectations are doing most of the work. Households and firms are not reacting to today's rate in isolation; they are reacting to what they believe rates and income will be tomorrow. This is why central bankers spend so much energy on communication — a credible forward path is, in the model, almost a substitute for action.

A central bank doesn't so much set the price of money as set the expectation of the price of money.

Where the model leaks

The textbook is a map, not the territory. Three leaks show up again and again:

1. The zero lower bound. When policy rates hit zero, the conventional channel stops and the bank must lean on forward guidance and asset purchases — tools whose transmission is contested.
2. Financial fragmentation. The same policy rate produces very different lending conditions in different parts of an economy or currency union. One rate, many economies.
3. The natural rate is unobserved. Every term in the Taylor rule is measurable except $r_t^n$, which must be estimated. Get it wrong and the whole prescription rotates.

A small simulation

To make the dynamics tangible, here's a tiny simulation of the three-equation model under a demand shock, solved with the method of undetermined coefficients in Python:

import numpy as np
 
def simulate(beta=0.99, kappa=0.16, sigma=1.0,
             phi_pi=1.5, phi_y=0.125,
             shock=1.0, periods=20):
    """Impulse response of a 3-equation New Keynesian model to a demand shock."""
    pi, y, i = np.zeros(periods), np.zeros(periods), np.zeros(periods)
    # Shock hits the IS curve in period 0 only
    for t in range(periods - 1):
        # Simple backward solving: expectations approximated by next period
        e_pi = pi[t + 1] if t + 1 < periods else 0.0
        e_y  = y[t + 1]  if t + 1 < periods else 0.0
        demand = shock if t == 0 else 0.0
        pi[t + 1] = beta * e_pi + kappa * y[t]            # Phillips curve
        y[t + 1]  = e_y - sigma * (i[t] - e_pi) + demand  # IS curve
        i[t]      = phi_pi * pi[t] + phi_y * y[t]         # Taylor rule
    return pi, y, i
 
pi, y, i = simulate()
print(f"Peak inflation: {pi.max():.3f}")
print(f"Peak output gap: {y.max():.3f}")

Run it and you'll see the familiar hump-shaped responses: inflation peaks a period or two after the output gap, the policy rate rises in sympathy, and both decay back to trend. That decay — the persistence — is entirely a function of $\beta$ and $\kappa$. Tweak them and you've changed the character of the business cycle in your toy world.

The honest takeaway

The model is useful precisely because it tells you what you are assuming. When inflation behaves nothing like the Phillips curve predicts — as it has, repeatedly, in the last few years — the fault is not in the world for being unruly but in our quiet assumption that expectations form the way the equations require.^[There is a deeper point here about the microfoundations of expectation formation that deserves its own essay.]

Monetary policy works. But it works the way a slow rudder works on a large ship: continuously, with lags, and always in conversation with the weather. Anyone who tells you it is either all-powerful or entirely powerless is selling you something.