In 2008 the Fed cut its rate toward zero, and companies paid more to borrow

7.45% The average yield on solid (BAA-rated) U.S. corporate bonds in 2008, up from 6.48% in 2006

Over the same two years the 10-year Treasury yield fell from 4.79% to 3.67%. The central bank pushed one interest rate down, and the rate that firms actually pay went up. How can that happen, and what does it do to output and inflation?

Open with the puzzle. The numbers are the notebook’s Fig. 1 key numbers: annual averages of Moody’s BAA yield and the 10-year Treasury yield, 2006 and 2008.

Chapters 11 to 13 assumed that the Fed sets the real interest rate at which firms and households borrow. This chapter drops that assumption. Between the Fed’s rate and the rate on a loan sits a premium for the risk of not being repaid, and in a financial crisis it jumps.

Chapter 10 told the story of the crisis with the finance toolkit: balance sheets, leverage, runs. This chapter puts the crisis inside the short-run model and asks what policy can do when the interest rate reaches zero.

Ask the room: if you ran a firm in late 2008 and wanted to build a new plant, which rate would you care about, the Fed’s overnight rate or the rate on your own bonds?

Today: three questions

• What does the crisis look like in the data? A risk premium that tripled, a policy rate stuck at zero, and bubbles before the great crashes.
• What does a financial friction do to the short-run model? It acts like a fall in demand; at the zero lower bound, deflation can turn a slump into a spiral.
• How did policy respond, and what came after? Money growth, quantitative easing, forward guidance; a lower path for potential output and productivity.

These are the three acts and the three branches of the mind map on the title slide.

The headline: a jump in the risk premium works like a collapse in demand, and the zero lower bound stops the Fed from offsetting it. Deflation can then turn a recession into a spiral, so central banks and governments need tools beyond the interest rate.

The model is the one of Chapters 11 to 13 (the IS curve, the MP curve, the Phillips curve and the AS/AD diagram) with one new parameter, the financial friction f̄.

The crisis in the data

Act I

When the Fed cuts, does borrowing get cheaper?

From 2007 to the end of 2008 the Fed cut its rate to almost zero. What happened to the rate on 10-year Treasury bonds, and to the rate on corporate bonds?

Predict the direction of each yield, and whether the gap between them grew or shrank.

Ask the room first. Most students expect both yields to fall, because “the Fed lowers interest rates”. Ask for a reason for each answer before showing the chart.

Hint if needed (only out loud): who is the safest borrower in the country, and where do scared investors put their money?

The government borrowed more cheaply, companies more dearly

Data: Moody’s BAA corporate bond yield (BAA); Federal Reserve Board, 10-year Treasury constant maturity (GS10), via FRED; the spread is our calculation. Chart inspired by Fig. 14.1 in Jones (2024), Macroeconomics, 6th ed.

How to read it: monthly yields in percent. The top line is the yield on BAA-rated corporate bonds (investment grade, but not risk-free), the middle line the 10-year Treasury yield, and the bottom line their difference, the spread, in percentage points. The hover box shows all three for one month. Click a line in the legend to hide it.

Do live:

1. Hover over 2006, then over December 2008: the spread jumps to about 6 points.
2. Hide the spread and look at the two yields in 2007–2008: they move in opposite directions.
3. Point at the smaller spikes: 2002 after the dot-com crash, early 2016 with the oil-price collapse, spring 2020 with Covid-19.

The euro-area version of this chart is in the book chapter: government bond spreads over Germany, which exploded in 2010–2012 (Greece 27.4 points in February 2012). Chapter 20 returns to that crisis.

The spread is the financial friction, and in 2008 it tripled

• From 1996 to 2007 the spread averaged 2.16 points. In December 2008 it reached 6.01, about three times its usual size.
• From 2006 to 2008 the Treasury yield fell from 4.79% to 3.67%, while the BAA yield rose from 6.48% to 7.45%.
• Investors fled to safety: they bought Treasuries and sold the bonds of firms that might not repay.

The Fed’s cuts reached the government, not the companies that invest.

Numbers: the notebook’s Fig. 1 key numbers (annual averages for 2006 and 2008; monthly peak of the spread).

The spread is our measure of the risk premium f̄ in the next act: the extra interest a lender asks a risky borrower to pay. In calm times it is small and stable; in a crisis it jumps because lenders fear default and because they want safe, liquid assets.

Contrast with Chapter 10, which used the TED spread, a 3-month rate between banks. This one is a 10-year rate for non-financial companies, closer to the rate that matters for building a factory.

Ask: why does the Treasury yield fall in a panic? (Demand for the safest asset rises, so its price rises and its yield falls.)

A zero policy rate does not mean cheap credit

“The Fed’s rate was zero, so money was cheap for everyone.”

Firms pay the policy rate plus a risk premium. In 2008 the premium rose by more than the Treasury yield fell, so the rate on corporate bonds went up, from 6.48% to 7.45%.

A common mistake in the news and among students: equating the central bank’s rate with “the” interest rate.

The policy rate is an overnight rate between banks. Households and firms borrow for years, from lenders who can lose money. Their rate is the policy rate, plus a term premium for lending long, plus a risk premium for default. In a financial crisis the last term can swamp the first.

For the policy students: when a minister says “rates are at record lows”, ask “which rates, and for whom?”

Loose or tight?

After 2008 the Fed’s rate sat at zero for seven years. By the standard of a Taylor rule, was that policy too loose, too tight, or about right?

Remind the room of the Taylor rule from Chapter 13: the rate rises with inflation above target and with output above potential. Ask students to guess what the rule recommended in 2009, with output far below potential and inflation low.

Before 2008 the Fed sat below its rule; after 2008 it was stuck at zero

Data: Federal Reserve Board (FEDFUNDS); BEA (PCEPILFE, GDPC1); CBO (GDPPOT); NBER (USREC), via FRED; the rule is our calculation. Chart inspired by Fig. 14.4 in Jones (2024), Macroeconomics, 6th ed.

How to read it: quarterly, in percent. One line is the actual fed funds rate, the other the rate a Taylor rule recommends: i = r̄ + π + 0.5(π − 2) + 0.5Ỹ, with core PCE inflation over four quarters and short-run output measured against the CBO’s potential GDP. With r̄ = 2 this is i = 1% + 1.5π + 0.5Ỹ. Shaded bands are NBER recessions. The slider sets r̄, the long-run real rate in the rule; the title reports the average gap between the rule and the actual rate in 2002–2006.

Do live:

1. Start at r̄ = 2 (the traditional value): from 2002 to 2006 the rule sits above the actual rate by 1.4 points on average.
2. Move r̄ to 1, the value many economists now prefer: the rule drops by one point in every quarter, and the 2002–2006 gap falls to 0.4 points.
3. Point at 2021–2022: with core inflation near or above 5 percent, even the lower rule called for rates far above the actual ones.

The inputs of the rule (a deep slump after 2008, low inflation until 2021) are drawn in the book chapter; data revisions have raised short-run output by 1 to 2 points since 2013.

By the rule, the Fed looked loose both before 2006 and after 2011

• In 2009 the rule itself was near or below zero (between −0.48% and 0.95% across the four quarters). From 2011 to 2015 it called for at least 1.05%, while the actual rate never rose above 0.19%.
• But firms did not borrow at the fed funds rate: with the spread of the last chart, a zero rate was not generous.
• From 2002 to 2006 the actual rate was 1.4 points below the rule, a possible fuel for the housing boom.

A rule is a benchmark, not a verdict: it ignores the risk premium, and it moves one for one with \(\bar{r}\).

Numbers: the notebook’s Taylor-rule key numbers and the Fig. 4 slider titles. The highest rule rate in the sample, 9.84 percent in January 2022, came when the fed funds rate was 0.12 percent: the inflation surge of Chapter 10 caught the Fed far behind its rule.

Beyond the U.S. (Fig. 4c in the book chapter): in 2022 all 14 central banks were far below their rules, by 6.8 points in the U.S., 6.5 in the euro area and 13.8 in the U.K.; in 2002–2006 the euro area was also below its rule, by 1.4 points a year on average. “Too low for too long” was not only an American story.

The debate on 2002–2006 returns in Act III (should central banks lean against bubbles?).

Why inflation did not keep falling after 2008, although output was far below potential, is the Devil’s advocate at the end.

Bubbles in the stock market

Stock prices are high today relative to company profits. How can we tell when prices have run too far ahead of what firms earn?

Ask: what is a share of a company worth? Let students arrive at “the profits it will pay its owners, now and in the future”. Then ask how we would compare prices across a century in which the economy grew many times over.

Prices ran far ahead of profits before the crashes of 1929 and 2000

Data: Robert J. Shiller, cyclically adjusted price-earnings ratio (CAPE), shillerdata.com. Chart inspired by Fig. 14.7 in Jones (2024), Macroeconomics, 6th ed.

How to read it: Shiller’s price-earnings ratio, the real S&P stock price index divided by the average real earnings of the previous ten years (smoothing out the swings of profits over the cycle). Monthly since 1881. The dashed line is the average from the year on the slider to today.

Do live:

1. Point at the two labelled peaks: September 1929 (32.6) and December 1999 (44.2). Both were followed by long falls.
2. Move the slider from 1881 to 2001: the average rises from 17.8 to 27.6. Today’s ratio, 40.7, looks extreme against the long average and less so against the recent one.

Why only the U.S.: Shiller’s long series is free only for the U.S. market.

A high ratio has always come back down, but “normal” is hard to pin down

• From September 1929 to June 1932 the ratio fell from 32.6 to 5.6: −83%.
• From December 1999 to March 2009 it fell from 44.2 to 13.3: −70%.
• In October 2026 it stands at 40.7. Against the average since 1881 (17.8) that is a bubble; against the average since 2001 (27.6), less clearly.

A bubble is easy to see afterwards and hard to prove in real time.

Numbers: the notebook’s “two stock market crashes” key numbers and the Fig. 7 slider titles.

Shiller used this ratio to warn of a bubble in the late 1990s, before the dot-com crash. But the ratio has stayed above its long-run average for most of the time since 1990. Either markets have been overvalued for thirty years, or something changed: lower real interest rates, a different mix of firms, different accounting of earnings. The next slide shows why a small change in the discount rate moves the “right” ratio a lot.

Home prices tell a similar story; Chapter 10 drew them (real prices fell 35.7 percent from 2005 to 2012). The book chapter adds 58 economies: Ireland −56.7 percent, Spain −44.5 percent.

Stock prices are very sensitive to the interest rate

\[\frac{P}{E} = \frac{1}{R-g}\]

• Price \(P\) = present value of profits \(E\) growing at \(g\), discounted at \(R\).
• With \(R = 7\%\) and \(g = 2\%\): \(P/E = 20\). One point more on \(R\): \(P/E = 16.7\), a fall of 17%.
• When prices collapse, household wealth and spending fall: a lower \(\bar{a}\).

Asset prices link the financial system to demand: a premium that jumps also cuts wealth.

The derivation (in the book chapter’s Model overview): the price is a geometric series E/(1+R) + E(1+g)/(1+R)² + …, with ratio (1+g)/(1+R) < 1, which sums to E/(R − g). The notebook’s Check 6 adds 2,000 years of discounted profits and gets 20.00 both ways; with R = 8 percent, 16.67, a change of −17 percent.

The log derivative is −1/(R − g) = −20: one point of R lowers ln P by about 0.2 at the baseline. The exact fall for a full point is 17 percent because the slope flattens as R − g grows.

A ratio far above what plausible values of R − g can explain is the textbook’s sign of a bubble. The policy question, whether the central bank should try to deflate one, is in Act III.

Friction, deflation and the zero lower bound

Act II

Firms borrow at the Fed’s rate plus a risk premium

\[R_t = R^{ff}_t + \bar{f}\]

\[\widetilde{Y}\!_t = \bar{a} - \bar{b}\,(R_t - \bar{r}) \quad\Longrightarrow\quad \widetilde{Y}\!_t = \underset{\text{demand shock}}{(\bar{a} - \bar{b}\bar{f})} - \bar{b}\,(R^{ff}_t - \bar{r})\]

Measured against the rate the Fed controls, a higher premium looks exactly like a fall in \(\bar{a}\).

The new assumption: the Fed sets the real fed funds rate R^ff, and lenders add a premium f̄ ≥ 0 for the risk of not being repaid. In calm times f̄ is close to zero (the textbook’s normalization); Fig. 1 measured it as the BAA spread.

Substitute the friction into the IS curve of Chapter 11 and split the bracket: Ỹ = ā − b̄(R^ff + f̄ − r̄) = (ā − b̄f̄) − b̄(R^ff − r̄).

Intuition: a firm does not care why its loan is expensive. A higher premium cuts investment just as a loss of confidence would. One point of premium costs b̄ points of output, exactly like a fall of b̄ in ā.

Baseline values used in every diagram (the book chapter’s illustrative set): r̄ = 2 percent, b̄ = 1, π̄ = 2 percent, m̄ = 1, ν̄ = 0.5. The 2008 shocks: the housing bust lowers demand by ā′ = −2, and the crisis adds f̄ = 4 points.

Experiment: a housing bust, a Fed cut, and then a financial crisis

The housing bust lowers demand by \(\bar{a}' = -2\). The Fed cuts \(R^{ff}\) from 2% to 0%, just enough to offset it. Then the crisis adds a premium \(\bar{f} = 4\) while the Fed stays at 0%. Where does output end: at potential, at \(-2\%\) (the Fed doing nothing), or worse?

Ask for a number. Students who use the equation of the last slide should get Ỹ = −2 − 1 × (0 + 4 − 2) = −4 percent.

A frequent wrong answer: “the Fed’s cut offsets the bust, so output is back at potential”. Ask: which rate do firms pay after the crisis?

A risk premium can undo the Fed’s cut, and more

Model figure: the IS/MP model with a financial friction, \(R = R^{ff} + \bar{f}\); baseline \(\bar{r} = 2\), \(\bar{b} = 1\), \(\bar{a}' = -2\). Adapted from Fig. 14.2 in Jones (2024), Macroeconomics, 6th ed.

How to read it: two IS/MP diagrams, output Ỹ across and the real interest rate R up. Panel (a): the housing bust shifts IS back to IS′; with no Fed action the economy moves from A to B; the Fed cuts to MP′ and reaches potential at C. Panel (b) adds the crisis: firms pay R = R^ff + f̄ (MP″), and the economy lands at D. The slider sets the friction f̄; the title computes B, the rate firms pay, and D.

Do live:

1. Start at f̄ = 4: R = 4 percent, D at −4 percent.
2. Move to 0: D and C coincide, the cut fully offsets the bust.
3. Move to 2: D sits exactly at B, as if the Fed had done nothing.
4. Move to 6: D at −6 percent.

Ask: does the size of the housing shock matter for D, once the Fed has cut to 0? (No: the cut removed it; D = −b̄f̄.)

The answer: output falls to −4%, twice the recession with no Fed at all

• With \(\bar{f} = 0\) the cut to 0% brings the economy back to potential at \(C\).
• With \(\bar{f} = 2\), the premium cancels the cut: \(D\) lands on \(B\), at −2%.
• With \(\bar{f} = 4\), firms pay 4%, two points above \(\bar{r}\), and output falls to −4%.

Two shocks conspired: households spent less, and credit cost more.

Numbers: the notebook’s “quick check of the 2008 story”: B −2.0 percent, C 0.0 percent, D −4.0 percent.

After the cut that offsets the bust, what is left is −b̄f̄: every point of premium costs one point of output, whatever the size of the housing shock.

The textbook’s panel (b) only says “a deep recession at D”; the baseline numbers make it concrete.

To close the gap the Fed needed a real rate of −4%

\[R^{ff,*} = \bar{r} + \frac{\bar{a}}{\bar{b}} - \bar{f} = 2 + \frac{-2}{1} - 4 = -4\%\]

• Set \(\widetilde{Y}\!_t = 0\) in the IS curve with the friction and solve for \(R^{ff}\).
• The Fed must offset the demand shock and cut one extra point for every point of premium.
• With 2% inflation, a real rate of −4% means a nominal rate of −2%: below zero.

Here the short-run model meets a floor. The Fed cannot set a negative nominal rate.

Derivation: 0 = ā − b̄f̄ − b̄(R^ff − r̄), so R^ff − r̄ = ā/b̄ − f̄. The notebook’s zero-lower-bound key numbers: the real fed funds rate needed is −4.0 percent, a nominal rate of −2.0 percent; the best the Fed can do is R^ff = −2 percent, so firms pay R = 2 percent and output is −2 percent, point B.

This is why the Fed turned to other tools (Act III). The floor itself is three slides ahead.

The friction enters the AD curve as one more demand shock

\[R_t - \bar{r} = \bar{f} + \bar{m}\,(\pi_t - \bar{\pi})\]

\[\widetilde{Y}\!_t = \underset{\text{AD shock}}{(\bar{a} - \bar{b}\bar{f})} - \bar{b}\bar{m}\,(\pi_t - \bar{\pi}) \qquad\qquad \pi_t = \pi_{t-1} + \bar{\nu}\,\widetilde{Y}\!_t\]

The AD curve of Chapter 13 with \(\bar{a}\) replaced by \(\bar{a} - \bar{b}\bar{f}\); the AS curve does not move at first.

Add the monetary policy rule of Chapter 13, R^ff − r̄ = m̄(π − π̄), to the friction equation: firms pay the premium on top of whatever the rule sets. Substitute into the IS curve: the new AD curve. It slopes down because higher inflation makes the Fed raise rates.

The AS curve is the Phillips curve with adaptive expectations, π^e = π_{t−1}. The economy sits where AD and AS cross; solving the two gives this year’s output: Ỹ_t = [ā − b̄f̄ − b̄m̄(π_{t−1} − π̄)] / (1 + b̄m̄ν̄).

With b̄ = m̄ = 1 and ν̄ = 0.5 the denominator is 1.5.

Experiment: the 2008 shocks in the AS/AD diagram

The housing bust and the premium together give an AD shock of \(\bar{a} - \bar{b}\bar{f} = -6\). Inflation starts at the 2% target, and the Fed follows its rule. Predict output and inflation in the first year, and where inflation ends in the long run.

Students who use the solution on the last slide’s notes get Ỹ₁ = (−6 − 0)/1.5 = −4 percent and π₁ = 2 + 0.5 × (−4) = 0 percent.

For the long run, ask: in the long run inflation stops changing. What must output be then? (Zero.) Then read inflation off the AD curve at Ỹ = 0.

A deep recession, then inflation slides below zero

Model figure: the AS/AD model with a financial friction; \(\bar{a} - \bar{b}\bar{f} = -6\), \(\bar{b} = \bar{m} = 1\), \(\bar{\nu} = 0.5\). Adapted from Fig. 14.3 in Jones (2024), Macroeconomics, 6th ed.

How to read it: output Ỹ across, inflation π up. AD is the old aggregate demand curve, AD′ the new one after the shocks. AS is this period’s aggregate supply curve, which shifts down each period because firms expect last period’s inflation. A is the start, B the first period, C the long run. D is where the economy would jump if the shocks suddenly ended: today’s AS curve meets the old AD curve. The slider sets the period t; the title reports output and inflation in that period.

Do live:

1. Period 1: Ỹ = −4 percent, π = 0 percent, at B.
2. Period 2: −2.7 percent and −1.3 percent. Period 4: −1.2 and −2.8.
3. Period 12: output back at potential, inflation −4 percent, at C.

Point at how far D drifts from A as AS moves down: ending the crisis late would bring an inflationary boom on the way back.

The recession fades; the deflation stays

• Year 1: output −4%, inflation 0%. Without the rule, the AD shock would cost 6% of output: falling inflation makes the Fed cut.
• Then the AS curve shifts down each year, and the economy slides along AD′.
• In the long run output is back at potential, but inflation is −4%: deflation.

Starting from only 2% inflation, a large enough shock ends in falling prices.

Numbers: the notebook’s Check 2: year 1 −4.00 percent and 0.000 percent; year 2 −2.67 and −1.333; year 3 −1.78 and −2.222; year 8 −0.23 and −3.766. Long-run inflation, equation (6): −4.00 percent.

The textbook’s figure ends at −2 percent inflation; with our baseline (m̄ = 1) it is −4. The book chapter’s “Your turn” asks which m̄ gives −2 (answer: m̄ = 1.5).

Deflation ends when the rule has cut the borrowing rate enough to offset both shocks. That only works if the rule can deliver the negative real rate it asks for, the hidden assumption of the AS/AD diagram.

Each year the economy closes one third of the remaining gap

\[\pi_C = \bar{\pi} + \frac{\bar{a} - \bar{b}\bar{f}}{\bar{b}\bar{m}} = 2 + \frac{-6}{1} = -4\%\]

\[\pi_t - \pi_C = \left(\frac{1}{1 + \bar{b}\bar{m}\bar{\nu}}\right)^{t} (\pi_0 - \pi_C) = \left(\frac{2}{3}\right)^{t} \times 6\]

A more aggressive rule (a larger \(\bar{m}\)) makes the first recession milder and the convergence faster.

Long run: inflation stops changing only when Ỹ = 0 (the AS curve). Put Ỹ = 0 into the AD curve and solve for π.

Speed: write x_t = π_t − π_C. Using the long-run equation, the solution for output becomes Ỹ_t = −b̄m̄x_{t−1}/(1 + b̄m̄ν̄), and the AS curve then gives x_t = x_{t−1}/(1 + b̄m̄ν̄). With 1 + b̄m̄ν̄ = 1.5, two thirds of the gap remain each year: π₁ = −4 + 6 × 2/3 = 0, π₂ = −4 + 6 × 4/9 ≈ −1.3. The notebook’s Check 2 confirms the closed form year by year to three decimals.

This is a stable difference equation: the factor is below one. Keep it in mind for the spiral, where the factor is above one.

A higher premium lowers output for a while and inflation for good

A one-point rise in …	AD curve	Output, year 1	Inflation, year 1	Long-run \(\pi_C\)	Long-run output
demand shock \(\bar{a}\)	right by 1	+0.67	+0.33	+1	0
risk premium \(\bar{f}\)	left by \(\bar{b} = 1\)	−0.67	−0.33	−1	0

Each point of premium leaves inflation \(1/\bar{m}\) points lower in the long run.

The table is the notebook’s comparative statics, from the derivatives of the period-1 solution and of π_C: first-year output ∓1/1.5 = ∓0.67, first-year inflation ∓ν̄/1.5 = ∓0.33, long-run inflation ±1/(b̄m̄) or −1/m̄. The notebook’s Check 3 raises the premium by one point and gets −0.667, −0.333 and −1.000.

The friction moves only the AD curve. A recession that comes from a premium is temporary; the disinflation it causes is lasting unless policy or the target changes.

Ask: with inflation starting at 2 percent and m̄ = 1, how large must the AD shock be for long-run inflation to turn negative? (Worse than −2.)

At a zero nominal rate, the lowest real rate is minus inflation

\[R_t = i_t - \pi_t, \qquad i_t \geq 0 \quad\Longrightarrow\quad R_t \geq \bar{f} - \pi_t\]

\[{\widetilde{Y}}^{\max}_t = \bar{a} - \bar{b}\bar{f} + \bar{b}\,(\bar{r} + \pi_t)\]

The zero lower bound caps the output the Fed can reach, and the cap falls with inflation.

The Fisher equation of Chapter 12, rearranged for the real rate: R = i − π. (The textbook’s display of the rearranged equation prints a plus sign; it is a typo, and the text around it uses the minus.)

The zero lower bound: a nominal rate cannot go much below zero, because savers can always hold cash, which pays exactly zero. Several central banks set slightly negative rates in the late 2010s, by fractions of a point, so the floor is a little below zero, but it exists.

At the bound the real fed funds rate is −π, and firms pay f̄ − π. Put that into the IS curve: the best output the Fed can reach. The bound binds when the rate the Fed needs is below the floor, which is the same as Ỹmax < 0.

A liquidity trap: at a zero rate, money and short-term government bonds both pay zero, so swapping one for the other (an ordinary open-market operation) changes nothing.

Deflation lowers the best output the Fed can reach

Inflation \(\pi\)	Lowest real fed funds rate	Rate firms pay (\(\bar{f} = 4\))	Best output \({\widetilde{Y}}^{\max}\)
+2%	−2%	2%	−2%
0%	0%	4%	−4%
−2%	+2%	6%	−6%

In 2008, with 2% inflation, even a zero rate left output at \(-2\%\); with 2% deflation it would leave \(-6\%\).

Numbers: the notebook’s Check 4, which computes the best output both from the formula and step by step (zero nominal rate → Fisher equation → friction → IS curve): −2.0, −4.0 and −6.0 percent.

The same check shows the policy rule of our AS/AD baseline, i = 2π, asking for i = +4, 0 and −4 percent at these inflation rates: as soon as inflation turns negative, the rule asks for a negative nominal rate. In the path of the last experiment that happens from period 2 on, so the AS/AD figure quietly assumed something the Fed could not do. That is why the textbook switches back to the IS/MP diagram for deflation.

Each point of deflation costs b̄ = 1 point of reachable output.

Falling prices are not good news when rates are stuck

“Deflation is good for consumers: everything gets cheaper.”

With the nominal rate stuck at zero, each point of deflation adds a point to the real interest rate. Debts must be repaid in dollars that are worth more, and investment falls: at 2% deflation, the best the Fed can reach is −6% output, not −2%.

Deflation that comes from faster productivity growth (cheaper computers) is not the problem. The danger is a general fall in prices in a slump, when the nominal rate cannot fall with it.

The textbook’s example: repaying a 100-dollar loan in a year with dollars that buy more than when you borrowed. A borrower’s real burden rises; borrowers spend less, and lenders do not spend the gain one for one.

Ask the policy students: why do central banks aim for 2 percent inflation rather than zero? (A buffer above the floor, so the real rate can go negative in a slump; and measured inflation overstates true inflation a little.)

A frozen nominal rate turns a stable model into a spiral

\[\widetilde{Y}\!_t = (1 + \bar{b}\bar{\nu})\,\widetilde{Y}\!_{t-1}\]

When inflation falls …	the real rate	each year the gap is multiplied by
the Fed follows its rule (\(\bar{m} = 1\))	falls	\(1/(1 + \bar{b}\bar{m}\bar{\nu}) = 0.67\)
the nominal rate is frozen	rises	\(1 + \bar{b}\bar{\nu} =\) 1.5

A slump of −6% becomes −9%, then −13.5%: each year half as deep again.

Derivation: with the nominal rate stuck at ī and adaptive expectations, firms pay R_t = ī − π_{t−1} + f̄. The IS curve becomes Ỹ_t = ā − b̄(ī + f̄ − r̄) + b̄π_{t−1}. Write it one year earlier and subtract: every barred term cancels, and Ỹ_t − Ỹ_{t−1} = b̄(π_{t−1} − π_{t−2}) = b̄ν̄Ỹ_{t−1} by the Phillips curve.

Numbers: the notebook’s Check 5 freezes the nominal rate at the 1929 level (5 percent) and starts from the crash: −6.00, −9.00, −13.50, −20.25 percent, a ratio of 1.50 each year.

There is no resting point except Ỹ = 0, and the economy moves away from it. A frozen nominal rate behaves like a policy rule with a negative m̄. This is the deflationary spiral; the next figure applies it to the 1930s.

Experiment: the Great Depression, one shock at a time

Late 1920s: output at potential, inflation about zero. First the Fed raises the interest rate by 3 points. Then the stock market crash lowers demand by \(\bar{a}' = -3\). Then prices start to fall, and the Fed keeps the nominal rate where it is. Predict output after each of the three steps.

The textbook’s worked exercise. With b̄ = 1, each of the first two steps costs 3 points of output: students should get −3 and −6. The third step needs the Phillips curve: at −6 percent, inflation falls by 0.5 × 6 = 3 points, so the real rate rises by 3 points if the nominal rate is fixed.

Ask: why would a central bank raise rates in 1928? (It worried about a stock market bubble; Act I showed the 1929 peak of the price-earnings ratio.)

Each step of the 1930s cost another 3 points of output

Model figure: the IS/MP model with the Fisher equation, \(R = i - \pi\); a 3-point rate hike, \(\bar{a}' = -3\), \(\bar{\nu} = 0.5\). Adapted from Fig. 14.12 in Jones (2024), Macroeconomics, 6th ed.

How to read it: output Ỹ across, the real interest rate R up. IS slopes down; each MP line is the real rate the economy faces. The slider walks through the steps: 0 = A, 1 = B, 2 = C, 3 = D. The title reports output, the real rate and inflation at each point.

Do live:

1. Step 0, A: the late 1920s, Ỹ = 0, R = 2 percent.
2. Step 1, B: the Fed raises rates, MP → MP′, R = 5 percent, Ỹ = −3 percent.
3. Step 2, C: the crash shifts IS back to IS′, Ỹ = −6 percent at the same R.
4. Step 3, D: deflation of 3 percent with a fixed nominal rate raises R to 8 percent (MP″), Ỹ = −9 percent.

The answer: −3%, −6% and −9%, and the last step feeds itself

• \(B\): the rate hike raises \(R\) from 2% to 5%: output −3%.
• \(C\): the crash shifts the IS curve back: output −6%, at the same real rate.
• \(D\): prices fall 3%; with \(i\) fixed, \(R = i - \pi\) rises to 8%: output −9%.

At \(D\) the slump is deeper, so prices fall faster, so the real rate rises again: the spiral.

Numbers: the notebook’s depression steps (and the slider titles): A (0, 2, 0), B (−3, 5, 0), C (−6, 5, 0), D (−9, 8, −3) for (Ỹ, R, π).

Nothing in the model stops the spiral unless policy changes. In the book chapter’s notebook, setting ν̄ = 0 (prices that never react) puts D on top of C: the spiral needs prices that fall in a slump.

History: the economy turned around after 1933, when the Fed eased and the U.S. left the gold standard. Chapter 20 returns to the gold standard.

Deflation in the 1930s

Every point \(B\), \(C\) and \(D\) lies below potential. With \(\bar{\nu} = 0.5\), by how much does inflation fall each year at each point?

Students apply Δπ = ν̄Ỹ: −1.5, −3 and −4.5 points. Ask: inflation was about zero in the late 1920s; what does a fall of 3 points mean for the price level? (Prices fall: deflation.)

The deeper the slump, the faster prices fall

Model figure: the Phillips curve \(\Delta\pi = \bar{\nu}\widetilde{Y}\) with \(\bar{\nu} = 0.5\), at the points of the Great Depression exercise. Adapted from Fig. 14.13 in Jones (2024), Macroeconomics, 6th ed.

How to read it: output Ỹ across, the change in inflation Δπ up. The Phillips curve goes through the origin. The dots are the points A to D of the last figure; the slider walks through them. Below the dotted horizontal line, next year’s inflation is lower than this year’s.

Do live: step through 0 to 3 and read Δπ in the title: 0, −1.5, −3 and −4.5 points.

Recession lowers inflation, deflation deepens the recession

• The fall in inflation is −1.5 points a year at \(B\), −3 at \(C\) and −4.5 at \(D\).
• The fall at \(C\) is what raises the real rate from \(C\) to \(D\); the larger fall at \(D\) would raise it again.
• In the 1930s prices fell by about a quarter between 1929 and 1933.

A steeper Phillips curve makes deflation, and so the spiral, faster.

Numbers: the notebook’s d_pi column of the depression steps; “about a quarter” is the book chapter’s figure for the fall in prices from 1929 to 1933.

The two diagrams tell one story: the IS/MP diagram turns a real rate into output, and the Phillips curve turns output into the next change in inflation, which, through the Fisher equation with a fixed nominal rate, becomes the next real rate.

Friedman and Schwartz (1963) blamed the Depression on a Fed that let the money supply collapse. Act III checks whether the Fed repeated that mistake after 2008.

Every way out raises the output the Fed can reach

Policy	What it moves in the model	Effect
cut the fed funds rate	\(R^{ff}\)	works until the floor \(-\pi\)
rescue banks, raise capital	the premium \(\bar{f}\)	lower borrowing rates
quantitative easing	long-term and mortgage rates	borrowing rates beyond \(R^{ff}\)
forward guidance	expected future short rates	lower long rates at a zero short rate
fiscal stimulus	demand \(\bar{a}\)	raises output one for one at the bound
anchored inflation expectations	\(\pi\) in \({\widetilde{Y}}^{\max}\)	stops the real rate from rising

Act III asks which of these tools were used after 2008, and how well they worked.

The table is the book chapter’s summary of the policy block of the model. Each row raises Ỹmax = ā − b̄f̄ + b̄(r̄ + π) or lowers the rates firms actually pay.

Bank rescues and capital requirements were Chapter 10’s policy slides (the TARP capital injections lowered leverage and the premium). This deck focuses on the tools of the central bank at the floor and on fiscal stimulus.

Secular stagnation (Summers, 2014): the case in which the rate the Fed would need, R^ff*, stays below the floor for years, because saving is high and investment demand weak.

Policy and the aftermath

Act III

Did the Fed repeat the 1930s?

In the Great Depression the U.S. money supply shrank. After 2008, did money shrink, stall, or grow fast?

Predict the growth of currency, M1 and M2 at the end of 2008.

Remind the room: currency is notes and coins; M1 adds checking accounts; M2 adds savings accounts and small time deposits (Chapter 8). Ask for a direction and a rough size, and a reason.

Bernanke, then a Fed governor, told Friedman and Schwartz at a 2002 celebration that the Fed had caused the Depression and would not do it again. This chart is the test.

This time the money supply grew fast

Data: Federal Reserve Board, H.6 money stock: currency (CURRSL), M1SL, M2SL; NBER (USREC), via FRED; 12-month growth is our calculation. Chart inspired by Fig. 14.6 in Jones (2024), Macroeconomics, 6th ed.

How to read it: the growth of each money measure over the previous 12 months, in percent, monthly since 1990. Shaded bands are NBER recessions. Click a measure in the legend to hide it.

Do live:

1. Zoom on 2008–2009: M1 growth jumps from about 7 to about 17 percent between October and December 2008, currency from 4 to 7 percent.
2. Move to 2020–2021: M2 grows by up to 27 percent with the Covid-19 relief programs.
3. Point at the gap in M1: the annotation explains it.

Data hygiene: in May 2020 the Fed redefined M1 to include savings deposits, and M1 jumped by about 11 trillion dollars in one month. Growth rates that compare the new definition with the old, May 2020 to April 2021, are meaningless (291 to 357 percent), so the chart leaves them out.

Money did not collapse: it grew faster in the crisis

• Between October and December 2008, M1 growth rose from about 7% to about 17%; all three measures stayed high for most of 2009.
• In 2020–21 M2 grew by up to 27%; from late 2022 to early 2024 it fell, for the first time since 1959.
• Abroad, broad money stopped growing in 2009 in the euro area (−0.3%) and the U.K. (−0.1%).

The Fed passed Friedman and Schwartz’s test. Money growth alone did not end the slump.

Numbers: the notebook’s Fig. 6 “What to notice” and the world broad-money chart (Fig. 6b in the book chapter): of the 160 economies with data for 2009, 13 saw broad money shrink. China’s grew 28 percent with its stimulus.

A single year of fast money growth does not mean inflation is coming. At a zero interest rate, people and banks are happy to hold money that pays nothing, so velocity falls (Chapter 8). The quantity theory’s link from money to prices loosens.

Data hygiene from the world chart: Sierra Leone shows +131,119 percent in 2001; that is a break in how the stock was recorded, not money creation. Always check a surprising growth rate against inflation.

With its rate at zero, what else could the Fed do?

The fed funds rate cannot go below zero. Which other assets could the Fed buy, and how would that lower the rates firms pay?

Let students name candidates: long-term government bonds, mortgage-backed securities, loans to banks, commercial paper. Ask how buying a bond changes its yield (a higher price is a lower yield).

The Fed’s balance sheet grew about tenfold, in waves

Data: Federal Reserve Board, H.4.1: total assets (WALCL), Treasuries by maturity (e.g. TREAS1T5, TREAS10Y), MBS (WSHOMCB), FEDDT, WCPFF, via FRED; our five layers. Chart inspired by Fig. 14.9 in Jones (2024), Macroeconomics, 6th ed.

How to read it: the Fed’s assets in billions of dollars, weekly, stacked by type: Treasuries with more than a year to maturity (“long-term”), mortgage-backed securities, Treasuries due within a year, credit-market facilities, and lending to banks and other. The labels mark QE1, QE2, QE3 and the Covid-19 purchases. Hover to read each layer.

The textbook reprints a Cleveland Fed chart that is no longer published; we split Treasuries by remaining maturity instead, so our long-term layer is not zero before 2009.

Do live: point at late 2008 (the lending layers swell and disappear within two years), then the steps of QE1 to QE3, then the jump of 2020 and the slow run-off since 2022.

First the Fed lent; then it bought long-term assets

• Late 2008: emergency lending to banks and credit markets, repaid within about two years.
• QE1 to QE3: mortgage-backed securities and long-term Treasuries, to about 4.5 trillion dollars by 2015.
• Covid-19: more than 4 trillion more in two years, to a peak of 8,965 billion (10.3 times January 2007). In 2026, about 6.8 trillion.

Quantitative easing buys long-term assets to pull down rates the Fed does not set directly.

Numbers: the notebook’s Fig. 9 “What to notice” and the Table 14.1 key numbers (weekly peak 8,965 billion dollars, 10.3 times January 2007).

Beyond the U.S. (Fig. 9b in the book chapter, central bank assets as a share of GDP from the BIS): at their 2021 peaks the Eurosystem held 68 percent of GDP, the Bank of England 48 and the Bank of Japan 129, against 37 for the Fed. The Bank of Japan was already at 21 percent in 2007: its zero rates and asset purchases started a decade earlier. Switzerland’s 145 percent is mostly foreign currency, bought to hold down the franc, not QE.

The 2008 lending is the lender-of-last-resort role of Chapter 10; QE is the new tool.

Bank reserves, not new currency, paid for the expansion

Billions of dollars	May 2007	May 2015	Sep 2026
Assets: Treasury securities	790	2,460	4,556
mortgage-backed securities	0	1,731	1,910
everything else	118	333	329
Liabilities: currency	811	1,366	2,483
bank reserves	11	2,595	2,948
everything else	86	563	1,364
Total (both sides)	908	4,524	6,795

The balance sheet grew fivefold from 2007 to 2015; new currency paid for only 15.3% of it.

The notebook rebuilds the textbook’s Table 14.1 from monthly averages of the weekly H.4.1 data and matches it within 11 billion dollars in every cell; the last column is the latest month. Total 2007 → 2015: 908 → 4,524 billion, 4.98 times as large, plus 3,616 billion.

The identity: Treasuries + MBS + other assets = currency + reserves + other liabilities. When the Fed buys a bond from a dealer, it credits the dealer’s bank with reserves: both sides grow by the same amount, and no currency is printed. Unusually, banks kept those reserves at the Fed (which since 2008 pays interest on them) instead of lending them out.

The book chapter’s “Your turn” asks for May 2022, near the peak.

Quantitative easing is not the same as printing money

“QE means the Fed printed trillions of dollars, so high inflation had to follow.”

The Fed paid mostly by crediting banks’ reserve accounts: currency paid for only 15.3% of the 2007–2015 expansion. And inflation stayed low until 2021, more than a decade after QE began.

The mistake mixes up two liabilities of the Fed. Currency circulates and is spent; reserves are deposits of banks at the Fed. At a zero interest rate and with interest paid on reserves, banks were willing to hold them.

Inflation did rise in 2021–2022 (Chapter 10), but economists still debate how much of it came from the 2020 QE, from fiscal transfers, from supply bottlenecks and from energy prices. The 2009–2015 rounds came with low inflation.

Ask: if QE is not printing money, how is it supposed to work? (Lower long-term and mortgage rates: the next slides.)

Forward guidance: promises about future rates move today’s long rates

\[i^{L}_t \approx \frac{i_t + i^e_{t+1} + \dots + i^e_{t+n-1}}{n}\]

• A 4-year bond when markets expect short rates of 0, 1, 2 and 3%: a long rate of 1.5%.
• A credible promise of zero for three years (0, 0, 0, 2): a long rate of 0.5%.

Even at a zero short rate, the Fed can lower long rates by shaping expectations.

The term structure of Chapter 12: a long rate is roughly the average of the short rates expected over its life (plus a term premium, ignored here). Firms borrow for years, so long rates matter for investment.

Numbers: the notebook’s Check 6 computes 1.50 percent and 0.50 percent.

In January 2012 the Fed said its rate would likely stay exceptionally low at least through late 2014. Whether markets believe such a promise depends on the central bank’s credibility: a promise to keep rates low after the economy recovers is a promise to allow some inflation later.

How long did markets expect zero to last?

In June 2009, with the rate at zero, how soon did financial markets expect the Fed to raise it, and how high?

Predict the rate markets expected for the end of 2011, and what it turned out to be.

Collect guesses. Many students guess “zero for a long time” with hindsight. Ask them to think like a trader in 2009, when every earlier recession had been followed by rate hikes within two years.

Markets kept expecting hikes that did not come

Data: Federal Reserve Board (FEDFUNDS); Treasury constant-maturity yields, H.15 (DGS3MO, DGS6MO, DGS1, DGS2, DGS3), via FRED; expected paths (forward rates) are our calculation. Chart inspired by Fig. 14.10 in Jones (2024), Macroeconomics, 6th ed.

How to read it: the solid line is the actual fed funds rate, monthly, in percent. Each dashed line is the path that markets expected in June of one year, read off the Treasury yield curve (forward rates). The textbook uses fed funds futures, which are not public; forward rates are a close proxy with a small term premium. The slider highlights one year’s forecast; the title compares the rate expected for the end of the forecast with the actual one.

Do live:

1. June 2009: markets expected 2.70 percent for December 2011; the actual rate was 0.07 percent.
2. Step through 2010, 2011, 2012: the June 2012 path, after the January 2012 guidance, is the flattest (0.57 percent expected for December 2014).
3. Jump to 2021: markets expected 0.88 percent for the end of 2023; the Fed had raised the rate to 5.33 percent.

Every forecast from 2009 to 2015 was too high; guidance flattened them

• June 2009: markets expected 2.70% for December 2011; the rate was 0.07%. The error was positive in every June from 2009 to 2015.
• June 2012, after the Fed’s promise: 0.57% expected for December 2014, the smallest error of those years together with 2015.
• June 2021: 0.88% expected for December 2023; the Fed was at 5.33%.

Guidance moved expectations, but markets can miss in both directions.

Numbers: the notebook’s “how wrong were the markets?” key numbers: errors +2.63, +1.62, +1.44, +0.45, +1.02, +1.16, +0.45 points for June 2009 to June 2015; and the Fig. 10 slider titles for 2021.

Why the persistent optimism? After every earlier recession the Fed had raised rates within a couple of years. The recovery after 2008 was slower, and the Fed’s own projections also kept forecasting hikes.

Why only the U.S.: fed funds futures and a free daily yield curve at five maturities exist only for the Fed and Treasuries in FRED. The ECB publishes a euro-area yield curve for anyone who wants to try the method.

Did the economy recover?

By about 2018, U.S. GDP was back at the CBO’s estimate of potential. Did the economy regain the ground lost after 2008?

Predict how far GDP in 2017 was from the potential GDP that the CBO projected in 2007, before the crisis.

Short-run output Ỹ is the gap between actual and potential GDP, and potential cannot be observed: the CBO estimates it with an extended Solow model and revises it as new data arrive. Ask: if the gap closed, which of the two lines moved?

The gap closed mostly because potential GDP was revised down

Data: BEA, real GDP (GDPC1); CBO, potential GDP (GDPPOT), latest and the June 2007 vintage from ALFRED, via FRED. Chart inspired by Fig. 14.5 in Jones (2024), Macroeconomics, 6th ed.

How to read it: trillions of chained 2017 dollars (the textbook used 2012 dollars, so its numbers are lower). Three lines: actual GDP, the CBO’s latest estimate of potential, and the estimate the CBO made in 2007, which ends in 2017, the last year it projected then.

Do live: hover over 2017 and read the three values. Then follow the 2007 estimate back to 2002, where the two estimates of potential are within half a percent of each other.

In 2017 GDP was 10.8% below the 2007 view of potential, 0.6% below today’s

• The two estimates of potential drifted apart after 2002; by 2007 the latest one was already 2.3% lower.
• After 2008 the old estimate kept its trend; the new one bent down to meet actual GDP.
• In 2017: −10.8% against the 2007 estimate, −0.6% against the latest one.

Potential was revised down by about 10%: the economy never returned to its old path.

Numbers: the notebook’s “two estimates of the output gap”: 2009 −4.66 (latest) and −8.81 (2007 estimate); 2015 −0.82 and −10.22; 2017 −0.63 and −10.78.

Two implications. For policy: the Taylor rule of Act I uses the latest estimate; with the 2007 estimate, the gap would have been much larger and the rule lower. For theory: a financial crisis may lower the long-run path, through lower investment and slower productivity, a link between the short run and the long run that research is still working out.

Beyond the U.S. (Fig. 5b in the book chapter, OECD data against the 1997–2007 trend extended): in 2019 U.S. GDP was 14.2 percent below that trend, the euro area’s 16.9, Spain’s 30.9 and Greece’s 52.4. The U.S. did not lose more than others. An extrapolated trend is a counterfactual, not a fact: for boom economies such as Latvia or Ireland it gives absurd paths.

A productivity slowdown?

The recovery after 2009 was slow. Did the crisis also slow the growth of total factor productivity, or had that started earlier?

TFP is output divided by a weighted combination of capital and labor, the Ā of the production model (Chapter 4). Ask: through which channels could a financial crisis slow it? (Credit-constrained firms with good projects cannot expand; less investment in new technology; resources stuck in weak firms.)

TFP growth slowed around 2004, before the crisis

Data: BLS productivity program, total factor productivity, private business (MFPPBS) and manufacturing (MFGPROD), via FRED, 2009 = 100. Chart inspired by Fig. 14.11 in Jones (2024), Macroeconomics, 6th ed.

How to read it: TFP for the private business sector and for manufacturing, set to 100 in 2009, on a ratio scale (equal slopes are equal growth rates). The slider picks the first year of the growth rate; the title reports average annual growth from that year to the latest, g = 100 × ln(x_end/x_start)/years.

Do live:

1. “Growth since 1995”: 1.00 percent a year in private business.
2. “Since 2004”: 0.72 percent; manufacturing 0.06 percent.
3. “Since 2010”: manufacturing turns negative (−0.21).

Since 2004, productivity has grown at half its earlier pace

• From 1995 to 2004 private-business TFP grew more than 1.5% a year: the information-technology boom.
• Since 2004: 0.72% a year in private business, 0.06% in manufacturing.
• The slowdown is wide: since 2004 TFP fell in the U.K., France and Italy, and in 56 of 120 countries it was lower in 2023 than in 2004.

The slowdown began before 2008, so the crisis cannot be its only cause.

Numbers: the notebook’s Fig. 11 key numbers; with the textbook’s data (through 2022) the numbers were 0.5 and 0.2 percent; revisions and strong years since 2022 raised them. The PWT 11.0 chart (Fig. 11b in the book chapter) uses whole-economy TFP with schooling-adjusted labor: U.S. 1.19 percent a year in 1995–2004 and 0.39 in 2004–2023; Germany 0.32, U.K. −0.22, France −0.42, Italy −0.80 since 2004; China 2.94, still catching up.

The causes are debated: fewer big ideas, mismeasurement of digital goods, more misallocation after the crisis. If the slowdown is real and lasting, it is a long-run problem (Chapters 5 and 6), and the CBO was right to lower potential.

Data hygiene: oil producers’ TFP (Kuwait −5.2 percent a year) moves with oil output and prices more than with technology.

Did quantitative easing and forward guidance work?

The Fed bought trillions of dollars of bonds and promised low rates for years. Did it lower borrowing costs and raise output, and by how much?

What the model says

• QE and guidance lower long and mortgage rates.
• That raises demand even at a zero short rate.
• The emergency lending of 2008 lowered \(\bar{f}\).

What the evidence says

• Announcements lowered long yields: clear.
• Effects on output and jobs: smaller, debated.
• Later rounds did less than the first.

The left column is Act II: QE works on the long-term and mortgage rates (the present value and term-structure equations), guidance on expected future short rates; both lower the rates firms pay beyond R^ff.

The evidence. Gagnon, Raskin, Remache and Sack (2011, International Journal of Central Banking) find that the first round of purchases lowered the 10-year term premium substantially, by tens of basis points, mostly on announcement days. Krishnamurthy and Vissing-Jorgensen (2011, Brookings Papers on Economic Activity) find large effects on the yields of the assets bought (mortgage-backed securities in QE1), smaller ones elsewhere. Swanson (2021, Journal of Monetary Economics) separates forward guidance from asset purchases in announcement data and finds that both moved yields, guidance mostly at short maturities. Engen, Laubach and Reifschneider (2015, Federal Reserve FEDS working paper) estimate, in a macro model, that the programs together lowered unemployment by about 1.25 points at their peak in 2016, the “larger estimate” the textbook mentions. Greenlaw, Hamilton, Harris and West (2018, U.S. Monetary Policy Forum) are skeptical: many announcement effects reversed, and the measured effects are small. The size of the effects on output is contested; that QE1 in a panic mattered is widely accepted.

Ask the policy students: QE leaves the central bank holding trillions of long-term bonds. What happens to its profits, and to the Treasury, when rates rise? (Since 2022 the Fed pays more interest on reserves than it earns on its bonds.)

Should the government spend more when rates are stuck at zero?

In February 2009 Congress passed a large stimulus (the Recovery Act). At the zero lower bound, how much output does each dollar of government spending buy?

What the model says

• Spending raises \(\bar{a}\); at the bound rates do not rise to offset it.
• To close the gap needs \(\Delta\bar{a} = -{\widetilde{Y}}^{\max}\): 2% of output in 2008.
• With 2% deflation it needs 6%: waiting is costly.

What the evidence says

• Some models: multipliers well above 1 at the bound.
• Historical data: near 1 or below, even at zero rates.
• Future taxes may offset part of it.

The left column is the zero-lower-bound block of Act II: government purchases are part of ā, and at the bound the Fed does not raise the rate when ā rises, so reachable output rises one for one. In the 2008 example, closing the gap needs Δā = 2 percent of potential output with 2 percent inflation, but 6 percent once deflation of 2 percent has set in.

The evidence. Christiano, Eichenbaum and Rebelo (2011, Journal of Political Economy) show that in a New Keynesian model the spending multiplier can be well above 1 when the zero bound binds, because spending also stops the fall in expected inflation. Ramey and Zubairy (2018, Journal of Political Economy), with U.S. data back to 1889, find multipliers mostly below 1, with no robust evidence that they are higher near the zero bound. Chodorow-Reich (2019, American Economic Journal: Economic Policy) surveys cross-state evidence, including the Recovery Act, and finds local multipliers around 1.8, which do not translate one for one into a national multiplier. Ramey (2019, Journal of Economic Perspectives) reviews the debate. The size of the multiplier at the bound is contested.

Ricardian equivalence (Chapter 11): spending financed by future taxes may lead households to save more today. Chapter 18 returns to fiscal policy and debt.

Ask: which would you choose at the bound, tax cuts or spending, and why?

Should central banks raise rates to prick asset bubbles?

The Fed raised rates in 1928 to cool the stock market, and kept rates below its rule in 2002–06 as home prices soared. Should monetary policy lean against bubbles?

The case for leaning

• Bursting bubbles cut wealth and \(\bar{a}\) (Act I).
• Low rates in 2002–06 may have fed the boom.
• Credit-fuelled bubbles end in the worst slumps.

The case against

• Bubbles are hard to spot in real time.
• A rate hike hits the whole economy: 1928.
• Capital rules and lending limits aim better.

The left column. Act I: the price-earnings ratio peaked before the 1929 and 2000 crashes; a small change in R − g moves prices a lot, and a collapse lowers ā. Fig. 4: from 2002 to 2006 the fed funds rate was 1.4 points below the Taylor rule (r̄ = 2). Taylor (2007, Jackson Hole symposium of the Federal Reserve Bank of Kansas City) argues that these low rates fuelled the housing boom. Jordà, Schularick and Taylor (2015, Journal of International Economics) show with 140 years of data that bubbles financed by credit are followed by much deeper recessions than equity bubbles without leverage.

The right column. Bernanke and Gertler (2001, American Economic Review Papers and Proceedings) argue that central banks should respond to asset prices only through their effect on the inflation and output forecast. Bernanke (2010, speech to the American Economic Association) argues that the Fed’s rates in 2002–2006 explain little of the boom, which happened in countries with different policies. Svensson (2017, Journal of Monetary Economics) estimates that the costs of leaning (a weaker economy now) exceed the benefits (a smaller chance of a crisis later). The 1920s are the cautionary tale of the model: the 1928 hike was step 1 of the Great Depression exercise.

The alternative is “macroprudential” policy: capital requirements and limits on loan-to-value ratios aimed at the credit that feeds the bubble (Chapter 10’s policy slide on capital). The question is contested; most central banks today lean on macroprudential tools first.

Ask the policy students: who should decide that a market is in a bubble, the central bank, a financial-stability council, or nobody?

“The model’s spiral never came”

A skeptic says: “After 2008 output stayed far below potential for years, yet U.S. inflation barely fell and never became deflation. The model’s deflationary spiral did not happen, so the model is wrong.”

Reply: the spiral assumes that people expect last year’s inflation. With a credible 2% target, expectations stayed near 2%, so the real rate did not climb. That is the model’s own way out: anchored expectations keep \(\pi\) up in \({\widetilde{Y}}^{\max}\). Why inflation fell so little is a genuine puzzle, but it is an argument for credibility, not against the model.

Take the objection seriously: it is the “missing disinflation” after 2008. Fig. 4 showed core inflation low after 2008 but not falling much, though short-run output was far below potential; a Phillips curve with adaptive expectations (Chapter 12) predicts much lower inflation.

Explanations, all debated. Hall (2011, American Economic Review, “The Long Slump”) argues that inflation behaves as if it were nearly unresponsive to slack, a near-exogenous force. Ball and Mazumder (2011, Brookings Papers on Economic Activity) find that the Phillips curve became flatter and that anchored expectations held inflation up. Coibion and Gorodnichenko (2015, American Economic Journal: Macroeconomics) argue that households’ and firms’ inflation expectations rose with oil prices in 2009–2011, which explains the missing disinflation. The textbook’s own reading: the Fed’s credibility in managing expectations.

The reply also shows why the model is useful: it tells policymakers what to protect. Japan in the 1990s and 2000s is the counterexample where expectations did slip and mild deflation lasted for years.

Ask: what could a central bank do to keep expected inflation from falling? (Announce and defend a target; forward guidance; some economists propose a higher inflation target, to stay farther from the floor.)

What to remember from this chapter

• Firms pay \(R = R^{ff} + \bar{f}\). In 2008 the spread tripled, so the Fed’s cuts did not lower borrowing costs.
• A higher premium works like a fall in demand: a recession first, and lower inflation for good.
• At the zero lower bound \(R = i - \pi\) cannot fall far enough; deflation raises it and can start a spiral.
• The Fed used money, QE and forward guidance; the gap closed partly because potential was revised down.
• Policy debates: how much QE and stimulus buy, and whether to lean against bubbles.

Ask each student to write one sentence answering the opening question: why did the 2008 crisis cause such a deep slump although the Fed cut its rate to zero? A good answer: the crisis raised the risk premium, so firms’ borrowing rates rose; to offset it the Fed needed a negative real rate below what a zero nominal rate allows, and falling inflation made the gap worse.

For the policy students, three habits: ask which interest rate a policy moves, and for whom; check whether the zero lower bound binds; and judge a recovery against more than one estimate of potential.

Walk the three branches of the mind map in about a minute each: the crisis in the data (cyan), friction, deflation and the zero lower bound (lavender), policy and the aftermath (green). Click the image to zoom in.

A jump in the risk premium is a fall in demand the Fed cannot always offset

Next: Chapter 15, DSGE Models: The Frontier of Business Cycle Research

The thesis: a jump in the risk premium works like a collapse in demand, and the zero lower bound stops the Fed from offsetting it; deflation can then turn a recession into a spiral, so central banks and governments need tools beyond the interest rate.

The next chapter looks at the modern models of business cycles, built from the decisions of households and firms, that researchers developed further after the crisis, adding financial frictions like the one in this chapter.

Sources: data and figures

• Fig. 1, the spread. Moody’s (BAA); Federal Reserve Board (GS10), via FRED. Inspired by Fig. 14.1 in Jones (2024).
• Fig. 4, the Taylor rule. Federal Reserve Board (FEDFUNDS); BEA (PCEPILFE, GDPC1); CBO (GDPPOT); NBER (USREC), via FRED. Inspired by Fig. 14.4 in Jones (2024).
• Fig. 5, potential GDP. BEA (GDPC1); CBO (GDPPOT), latest and June 2007 vintage (ALFRED). Inspired by Fig. 14.5 in Jones (2024).
• Fig. 6, money. Federal Reserve Board, H.6 (CURRSL, M1SL, M2SL), via FRED. Inspired by Fig. 14.6 in Jones (2024).
• Fig. 7, price-earnings ratio. Robert J. Shiller, CAPE, shillerdata.com. Inspired by Fig. 14.7 in Jones (2024).
• Figs. 9 and 10, the Fed’s balance sheet and expected rates. Federal Reserve Board, H.4.1 (WALCL, WSHOMCB, TREAS10Y, FEDDT, WCPFF) and H.15 (DGS3MO to DGS3), via FRED. Inspired by Figs. 14.9 and 14.10 and Table 14.1 in Jones (2024).
• Fig. 11, TFP. BLS (MFPPBS, MFGPROD), via FRED. Inspired by Fig. 14.11 in Jones (2024).
• Figs. 2, 3, 12 and 13, the model. Adapted from Figs. 14.2, 14.3, 14.12 and 14.13 in Jones (2024); baseline values and derivations are the book chapter’s own.
• Textbook. Jones, C. I. (2024). Macroeconomics, 6th ed. W. W. Norton. Book chapter, with the data, code and every figure: Data Science and Macroeconomic Models, Chapter 14.

Sources: research cited

• Money and the Depression. Friedman, M. and A. J. Schwartz (1963), A Monetary History of the United States, 1867–1960, Princeton University Press.
• QE and forward guidance. Gagnon, J., M. Raskin, J. Remache and B. Sack (2011), International Journal of Central Banking 7(1); Krishnamurthy, A. and A. Vissing-Jorgensen (2011), Brookings Papers on Economic Activity, Fall; Swanson, E. (2021), Journal of Monetary Economics 118; Engen, E., T. Laubach and D. Reifschneider (2015), Federal Reserve Board, FEDS 2015-005; Greenlaw, D., J. Hamilton, E. Harris and K. West (2018), U.S. Monetary Policy Forum.
• Fiscal multipliers. Christiano, L., M. Eichenbaum and S. Rebelo (2011), Journal of Political Economy 119(1); Ramey, V. and S. Zubairy (2018), Journal of Political Economy 126(2); Chodorow-Reich, G. (2019), American Economic Journal: Economic Policy 11(2); Ramey, V. (2019), Journal of Economic Perspectives 33(2).
• Bubbles and monetary policy. Bernanke, B. and M. Gertler (2001), American Economic Review 91(2); Taylor, J. (2007), Federal Reserve Bank of Kansas City, Jackson Hole symposium; Bernanke, B. (2010), speech to the American Economic Association, January 3; Jordà, Ò., M. Schularick and A. Taylor (2015), Journal of International Economics 96(S1); Svensson, L. (2017), Journal of Monetary Economics 90.
• The missing disinflation. Hall, R. (2011), American Economic Review 101(2); Ball, L. and S. Mazumder (2011), Brookings Papers on Economic Activity, Spring; Coibion, O. and Y. Gorodnichenko (2015), American Economic Journal: Macroeconomics 7(1). Secular stagnation. Summers, L. (2014), Business Economics 49(2).