LH Advanced Macroeconomics

(07 33109)

Part 2: Macroeconomics with

Nominal Rigidities

Dr Ceri Davies
c.r.davies.1@bham.ac.uk

Nominal Rigidities: Introduction

• We have seen that RBC models omit monetary sources of
macroeconomic fluctuations; money would be ‘neutral’ in
that model. However, historical experience and empirical
evidence suggest that this is likely to be important.

• For monetary factors to have real effects, we need some

sort of nominal imperfection – or ‘friction’ – in the model;
otherwise the ‘classical dichotomy’ holds.

• Widely-used approach: introduce a barrier to full price or

wage adjustment; e.g. ‘menu costs’, but not the only way.

• We need to think carefully about the microfoundations of

a model with these features (e.g. market power).

1
 For an Individual Firm with Price-Setting Power
Begin from the profit maximising
Q: Is point B or
point, MR=MC (point A)
point C better now?
Cut price or not?
Fall in
Profits available by demand
cutting price and
selling more.
Must be ‘small’ for
firms to choose to
keep the original
price; otherwise,
cut price
Source: Romer (1993, p.9)

Seemingly small nominal rigidities can have large aggregate affects.

Real Rigidities
• These amplify the nominal rigidities studied in this topic;
we require both for empirically satisfactory outcomes, e.g.
to break the classical dichotomy.

• Diagrammatically, real rigidities affect the slopes and shifts

of the MR and MC curves presented on the previous slide;
this affects the size of the ‘profit triangle’.

• Examples (Romer, 1993):

– ‘Thick market externalities’ which shift the MC curve up
and down during economic cycles.
– Capital market imperfections due to asymmetric
information.
– Non-Walrasian features of the labour market.

2
 Why does this Matter at the Macro Level?

Nominal/real trade-off via the

Phillips curve; classical
dichotomy breaks down.

Friedman-Phelps: we
need an expectations-
augmented Phillips
curve to explain this
data

Source: Romer (2019, p.259)

The ‘Modern Form’ Phillips Curve

• Price or wage rigidities influence the supply-side of the
economy; affects the behaviour of firms.
• You saw last year that a simple model with nominal
rigidity leads to a Phillips curve of the form (𝜓 > 0):

𝜋𝑡 = 𝜋𝑡𝑒 + 𝜓(ln𝑌𝑡 − ln𝑌ത𝑡 ) + 𝜀𝑡𝑆

Inflation Output gap

Expected Supply
inflation shock

• For example, under adaptive expectations:

𝜋𝑡 = 𝜋𝑡−1 + 𝜓(ln𝑌𝑡 − ln𝑌ത𝑡 ) + 𝜀𝑡𝑆

3
 Different Approaches to Nominal Rigidity:
Static Models (Romer, ch.6)

• Case 1: Keynes’ Model. The nominal wage is fixed and

above the market clearing level for some reason.

• Case 2: Sticky prices, flexible wages and a competitive

labour market. Firms set P>MC, n.b. firms need price-
setting power in this case.

• Case 3: Sticky prices, flexible wages and real labour

market imperfections. The labour market doesn’t clear.

• Case 4: Sticky wages, flexible prices and imperfect

competition. Produces slightly different outcomes.

Challenge + Next Steps

• What are the microfoundations for any of this?

• Older generations of models use an ‘ad hoc’ approach. This

became less acceptable from a methodological perspective
during the 1970s/80s (e.g. the Lucas critique, 1976).

• Next: we embed nominal rigidities into a micro-founded

DSGE model; builds on the theoretical elegance/rigour of
the RBC framework but adds an important empirical
observation about price or wage adjustment; crucially, firms
behave optimally subject to usual constraints + an
additional one for price-setting; the precise details of the
model must be updated.

4
 Time-Dependent Models of Nominal Adjustment

Prices are set by multi-period contracts; each period a

certain fraction of contracts (not all) come up for renewal
and prices can then be changed.

– Fischer’s model: Prices are pre-determined by the

contract length but not necessarily fixed; price
adjustment opportunities arrive deterministically.
– Taylor’s model: Prices are fixed for the duration of the
contract; price adjustment opportunities arrive
deterministically.
– Calvo’s model: Prices are fixed for the duration of the
contract; price adjustment opportunities arrive at
random.

n.b. there are other

complications in this market
(e.g. the energy price cap)

Source: BBC (29/09/21)

5
 A Dynamic ‘New Keynesian’ Model
• Each period, imperfectly competitive firms produce output
using labour as their only input; production function Y=L.

• No capital; no government spending; closed economy.

• Households choose consumption and labour hours, taking

the real wage and the real interest rate as given.

• The firms are owned by the households (i.e. shareholders).

They maximize the present discounted value of their profits,
subject to constraints on their price-setting.

• The central bank determines the path of the real interest

rate through the monetary policy process; money is included.

NK vs. RBC
• Retains some of the fundamental methodological
innovations of Kydland + Prescott (and others).

• Three significant modifications (Galí, 2018):

i. The NK model introduces nominal variables explicitly:
prices, wages, a nominal interest rate, money.
ii. It departs from the assumption of perfect competition
in the goods market and allows for price mark-ups;
the labour market is sometimes assumed to be
imperfectly competitive as well.
iii. It introduces nominal rigidities, typically using Calvo’s
model; a constant fraction of firms, drawn randomly,
are permitted to adjust their price in each period.

6
 Infinitely-Lived Households
• Maximise lifetime utility (0<β<1 is the discount factor):
A consumption
bundle (index)
∞

𝑈 = ෍ 𝛽 𝑡 𝑈 𝐶𝑡 − 𝑉(𝐿𝑡 )
𝑡=0
Labour hours
Assume the following functional forms: (disutility)

𝐶𝑡1−𝜃
CRRA form 𝑈(𝐶𝑡 ) = 𝜃>0
1−𝜃
𝐵 𝛾
𝑉(𝐿𝑡 ) = 𝐿 𝐵 > 0; 𝛾 > 1
𝛾 𝑡

Household Maximisation Problem

(Real-Terms Part)
𝑊𝑡
ℒ = 𝛽𝑡 𝑈 𝐶𝑡 − 𝑉(𝐿𝑡 ) + Λ 𝐿 − 𝐶𝑡
𝑃𝑡 𝑡
FOCs:
𝐶𝑡 : 𝛽𝑡 𝑈′ 𝐶𝑡 − Λ = 0
𝑊𝑡
𝐿𝑡 : −𝛽𝑡 𝑉 ′ (𝐿𝑡 ) + Λ =0
𝑃𝑡
Eliminating Λ:
Functional
′ 𝑊 𝛾−1 𝑊𝑡
𝑉 (𝐿𝑡 ) = 𝑈′ 𝐶𝑡 𝑃𝑡 𝐵𝐿𝑡 = 𝐶𝑡−𝜃
𝑡 form
𝑃𝑡

7
• In this model, Y=L=C; no I, G or (X – M):

𝑊𝑡 𝑉 ′ (𝑌𝑡 )
=
𝑃𝑡 𝑈′ 𝑌𝑡

• Using the (assumed) functional forms for instantaneous

utility again:

𝛾−1
𝑊𝑡 𝐵𝑌𝑡 𝜃+𝛾−1
= −𝜃 = 𝐵𝑌𝑡
𝑃𝑡 𝑌𝑡

• This gives the real wage as a function of output; a useful

equilibrium condition in the model (real part).

Plan for the Remainder of the Topic

• Introduce nominal variables (we didn’t have these in
the RBC model).

• Derive the NKIS equation/curve (the ‘demand-side’ =

households).

• Think more carefully about the microfoundations of

nominal (and real) rigidities.

• Derive the NK Phillips equation/curve (the ‘supply-side’

= firms).

• Present the ‘canonical NK model’, then critique +

possible extensions.

8
 Nominal-Terms Part of the h.h. Problem
• There is no capital investment in this model but we do allow
households to hold nominal bonds (think of a fixed-term
savings account); these earn the nominal interest rate 𝑖𝑡 .

• Households also hold money (cash) to support consumption

purchases; the downside is, cash earns no interest.

• Choose how much cash to hold + how many bonds to hold.

• Household wealth (A) evolves as follows in nominal terms:

𝐴𝑡+1 = 𝑀𝑡 (1 + 0) + (𝐴𝑡 + 𝑊𝑡 𝐿𝑡 − 𝑃𝑡 𝐶𝑡 − 𝑀𝑡 )(1 + 𝑖𝑡 )

Deriving the NKIS Equation

Suppose the h.h. reduces Ct by ΔC and increases its bond
holdings by PtΔC (n.b. real-to-nominal switch, so P is needed).

Use the proceeds to increase Ct+1 by: (1+𝑖𝑡 )PtΔC/Pt+1

As in the previous topic, the optimising h.h. must remain on

the same indifference curve (drop population growth, also
ignore uncertainty for now):

Utility cost Utility benefit

𝑃𝑡
0 = − 𝛽𝑡 𝑈′ 𝐶𝑡 Δ𝐶 + 𝛽𝑡+1 (1 + 𝑖𝑡 ) 𝑈′ 𝐶𝑡+1 Δ𝐶
𝑃𝑡+1

9
• From the previous expression we can write:
𝑃𝑡
𝑈 ′ (𝐶𝑡 ) = 𝛽(1 + 𝑖𝑡 ) 𝑈′ 𝐶𝑡+1
𝑃𝑡+1

• Use the Fisher equation as a link between real and

nominal parts of the h.h. problem:
𝑃𝑡+1
𝐹𝑖𝑠ℎ𝑒𝑟: (1 + 𝑖𝑡 ) = (1 + 𝑟𝑡 )
𝑃𝑡
• So we have:
𝑈 ′ (𝐶𝑡 ) = 𝛽(1 + 𝑟𝑡 ) 𝑈′ 𝐶𝑡+1
Euler equation
• With a CRRA utility function:
(without uncertainty)
𝐶𝑡−𝜃 = 𝛽(1 + 𝑟𝑡 )𝐶𝑡+1
−𝜃

The New Keynesian IS Curve

• Take logs of the previous expression:

−𝜃ln𝐶𝑡 = ln𝛽 + ln 1 + 𝑟𝑡 − 𝜃ln𝐶𝑡+1

1 1
ln𝐶𝑡 = − ln𝛽 − ln 1 + 𝑟𝑡 + ln𝐶𝑡+1
𝜃 𝜃

• Use Y=C; also use the approximation ln(1+x)≈x for small x:

1
ln𝑌𝑡 = 𝑎 + ln𝑌𝑡+1 − 𝑟𝑡
𝜃
1
where: 𝑎 ≡ − ln𝛽 Interpretation…
𝜃

10
 NKIS: Interpretation
• The key element is the negative relationship between
current output and the real interest rate; recall the
‘demand for goods & services’ equation last year.

• Future output also plays a role; this would be expected

future output in a model with uncertainty. We typically
use rational expectations in NK models; n.b. this does not
mean that your expectation is always correct.

• This equation represents the ‘demand side’ of our NK

DSGE model; we can easily add expectations and a
‘demand shock’ later on. Next we turn to the supply side;
we need to consider firms…

Firms
• There are a large number of firms in the economy. Firm i
produces output in period t according to the PF:

𝑌𝑖𝑡 = 𝐿𝑖𝑡

• And faces the demand curve (η>1):

−𝜂
𝑃𝑖𝑡
𝑌𝑖𝑡 = 𝑌𝑡
𝑃𝑡

where η is the price elasticity of demand for firm i’s output;

this is important for the firm’s price setting power.

11
 The Profit Function
• Real profits (R) can be expressed as revenues minus
costs:

𝑃𝑖𝑡 𝑊𝑡
𝑅𝑡 = 𝑌𝑖𝑡 − 𝑌
𝑃𝑡 𝑃𝑡 𝑖𝑡

1−𝜂 −𝜂
𝑃𝑖𝑡 𝑊𝑡 𝑃𝑖𝑡
= 𝑌𝑡 −
𝑃𝑡 𝑃𝑡 𝑃𝑡

• Now we need to think about how firms set their prices;

this is not straightforward when there are nominal
rigidities but these are crucial to the model.

Price-Setting
• Firms are aware that the price they set today could be in
force for some time due to the multi-period contracts in
place; assume that ‘today’ means t=0.
• Let qt denote the probability that the price set today is still
in effect in period t.
• Households ultimately own the firms; we assume that the
latter act in the best interests of the former, i.e. firms’
profits are ultimately valued in terms of h.h. utility.
• The marginal utility of h.h. consumption in period t relative
to period 0 is defined as 𝜆𝑡 (𝛽 is still the discount factor):

𝜆𝑡 ≡ 𝛽𝑡 𝑈′(𝐶𝑡 )/𝑈′(𝐶0 )

12
• The firm therefore chooses its period 0 price to maximise:
∞

𝐴 ≡ 𝐸 ෍ 𝑞𝑡 𝜆𝑡 𝑅𝑡
𝑡=0

• We already have an expression for Rt (using Pi for the price

set in period 0):
∞ 1−𝜂 −𝜂
𝑃𝑖 𝑊𝑡 𝑃𝑖
= 𝐸 ෍ 𝑞𝑡 𝜆𝑡 𝑌𝑡 −
𝑃𝑡 𝑃𝑡 𝑃𝑡
𝑡=0

• Which can be written as:

∞
𝜂−1 1−𝜂 −𝜂
= 𝐸 ෍ 𝑞𝑡 𝜆𝑡 𝑌𝑡 𝑃𝑡 𝑃𝑖 − 𝑊𝑡 𝑃𝑖
𝑡=0

• We can show that a firm with market power sets price as

a mark-up over marginal cost as follows:

𝑃𝑖 𝑃𝑡∗ 𝜂 𝑊𝑡
≡ =
𝑃𝑡 𝑃𝑡 𝜂 − 1 𝑃𝑡

• Notice the importance of the price elasticity of demand;

market power is limited by this.
• Hence rewrite 𝑃𝑖1−𝜂 − 𝑊𝑡 𝑃𝑖−𝜂 on the previous slide as
follows, where 𝑝𝑖 ≡ ln𝑃𝑖 and 𝑝𝑡∗ ≡ ln𝑃𝑡∗ :
A function of
∞
𝜂−1
𝐴 = 𝐸 ෍ 𝑞𝑡 𝜆𝑡 𝑌𝑡 𝑃𝑡 𝐹 𝑝𝑖 , 𝑝𝑡∗
𝑡=0

Price set at t=0 Optimal (profit max.)

price at time t

13
 Intuition
• Firms aim to keep the expected gap between 𝑝𝑖 and 𝑝𝑡∗ as
small as possible over the full time horizon:

min ෍ 𝑞𝑡 𝐸 𝑝𝑖 − 𝑝𝑡∗ 2
𝑝𝑖
𝑡=0

Without the nominal rigidity firms would set 𝑝𝑡∗ at all times;
we would just have a model with fully flexible prices and
there would be no ‘gap’.

There is an optimisation (minimisation problem) here but

we just need to understand the result…

Solution for pi
• We state that:
∞ ∞
𝛽 𝑡 𝑞𝑡
𝑝𝑖 = ෍ ∞ 𝜏 𝐸 𝑝𝑡∗ ≡ ෍ 𝜔
෥ 𝑡 𝐸[𝑝𝑡∗ ]
σ𝜏=0 𝛽 𝑞𝜏
𝑡=0 𝑡=0

where 𝜔෥𝑡 is the probability that the price the firm sets in
period 0 will be in effect in period t divided by the expected
number of periods the price will be in effect (both discounted).

Mathematically, 𝑝𝑖 is a weighted-average price; 𝜔

෥𝑡 measures the
importance of the optimal price in any given time period.

The exact form of this equation depends on which particular

model we use for the nominal rigidity.

14
• We know that: From
households

𝑃𝑖 𝑃𝑡∗ 𝜂 𝑊𝑡 𝑊𝑡 𝜃+𝛾−1
≡ = and = 𝐵𝑌𝑡
𝑃𝑡 𝑃𝑡 𝜂 − 1 𝑃𝑡 𝑃𝑡

Then, 𝜂
𝑝𝑡∗ = ln + ln𝐵 + 𝜃 + 𝛾 − 1 𝑦𝑡 + 𝑝𝑡
in logs: 𝜂−1

Assume = 0

• We can use log nominal GDP (y+p) in our expression for 𝑝𝑖 :

∞

𝑝𝑖 = ෍ 𝜔
෥𝑡 𝐸 𝜙(𝑦𝑡 + 𝑝𝑡 ) + (1 − 𝜙)𝑝𝑡
𝑡=0
𝜙 ≡ 𝜃+𝛾−1 Real rigidity

So far we have:
1
ln𝑌𝑡 = 𝑎 + ln𝑌𝑡+1 − 𝑟𝑡
𝜃
∞

𝑝𝑖 = ෍ 𝜔
෥𝑡 𝐸 𝜙(𝑦𝑡 + 𝑝𝑡 ) + (1 − 𝜙)𝑝𝑡
𝑡=0

• We need to ‘close’ (complete) the model; we need a Central

Bank.

• The C.B. determines the real interest rate; they actually set
the nominal interest rate but with price frictions they can
affect the real interest rate as well; in turn, the real interest
rate influences economic decisions.

15
 The Central Bank

• Has a target path for nominal GDP in mind and conducts

monetary policy accordingly.

• Using the quantity theory of money (MV=PY), with velocity

set to 1. In log form: 𝑚 = 𝑝 + 𝑦, where 𝑚 is set by the C.B.

• The C.B. cares about inflation and the output gap but we do
not specify the precise ‘rule’ yet; keep it simple for now.

• This allows us to focus on the supply-side and price-setting

behaviour here, i.e. the nominal rigidity; we will be more
precise about the monetary policy process later on.

Different Forms of Nominal Rigidity

∞

𝑝𝑖 = ෍ 𝜔
෥𝑡 𝐸 𝜙𝑚𝑡 + (1 − 𝜙)𝑝𝑡
𝑡=0

We can use this equation to model several different types of

nominal rigidity; each model leads to a different ‘supply side’
form for the economy (≈Phillips curve):
– Fischer’s model: Prices are pre-determined by the contract
length but not necessarily fixed; price adjustment
opportunities arrive deterministically.
– Taylor’s model: Prices are fixed for the duration of the
contract; price adjustment opportunities arrive
deterministically.
– Calvo’s model: Prices are fixed for the duration of the
contract; price adjustment opportunities arrive at random.

16
 Calvo’s Model
• Opportunities to change price arrive stochastically; this might
sound more complicated but it actually produces a more
tractable multi-period model.

• Opportunities to change price follow a Poisson process; the

probability of a firm being able to change its price is the same
each period, no matter how long ago the previous
opportunity arrived for that firm.

• Just like Taylor’s model, prices are fixed between price-

changing opportunities.

• Firms know the structure of the model and behave optimally;

constrained maximisation again.

The Profit-Maximising Price

• Using the QTM with V=1, we can write this as:

𝑝𝑡∗ = 𝜙𝑚𝑡 + (1 − 𝜙)𝑝𝑡

where m is the stock of money and pt is the general price level.

• Consider a one-off increase in m in time period t; e.g. a

‘monetary policy shock’.

• If 𝜙<1, the profit-maximising price doesn’t full adjust to

this; real rigidity is high. (‘Large’ 𝜙 for low real rigidity).

• Result: nominal rigidity leads to gradual adjustment of the

price level and real rigidity (small 𝜙) amplifies this.

17
 Deriving the New Keynesian Phillips Curve
• In each time period, a fraction α of firms are permitted to
change their price (0 < α < 1); firms are chosen at random
so α is also the probability that a firm can change price.

• What is the average price level in period t?

𝑝𝑡 = 𝛼𝑥𝑡 + (1 − 𝛼)𝑝𝑡−1

Price chosen by those Average price level from

firms who are allowed to the previous period
change price in period t

• This is simply a weighted-average.

• Subtract pt-1 from both sides and now we have inflation on

the LHS:

𝜋𝑡 = 𝛼(𝑥𝑡 −𝑝𝑡−1 )

• When firms set prices in the Calvo model, they must look
infinitely far into the future; there is a chance, though very
small, that they never get to adjust their price again.

• Recall (but with modified notation):

∞ ∞
𝛽 𝑗 𝑞𝑗 ∗ ∗
𝑥𝑡 = ෍ ∞ 𝐸 𝑝𝑡+𝑗 ≡ ෍𝜔
෥𝑗 𝐸 𝑝𝑡+𝑗
σ𝑘=0 𝛽 𝑘 𝑞𝑘
𝑗=0 𝑗=0

where qj is the probability that the price set in period t will

still be in place in period t+j.

18
• In Calvo’s model:

𝑞𝑗 = (1 − 𝛼)𝑗

Because (1–α) is the probability that a firm cannot change

price in any given period; multiply this probability j times.

So we have (be more precise about the information set now):

∞
𝛽 𝑗 (1 − 𝛼)𝑗
𝑥𝑡 = ෍ ∞ 𝐸 𝑝∗
σ𝑘=0 𝛽 𝑘 (1 − 𝛼)𝑘 𝑡 𝑡+𝑗
𝑗=0

∞
∗
= [1 − 𝛽(1 − 𝛼)] ෍ 𝛽 𝑗 (1 − 𝛼)𝑗 𝐸𝑡 𝑝𝑡+𝑗
𝑗=0

Mathematical Note
For |z|<1:
∞
𝑎
𝑎 + 𝑎𝑧 + 𝑎𝑧 2 + 𝑎𝑧 3 + 𝑎𝑧 4 + ⋯ = ෍ 𝑎𝑧 𝑡 =
1−𝑧
𝑡=0

Geometric series infinite sum; if a=1 the result is simply

1/(1 − 𝑧)

On the previous slide we used (remember, 0<β<1 and

0<α<1):
1
= [1 − 𝛽(1 − 𝛼)]
σ∞ 𝑘
𝑘=0 𝛽 (1 − 𝛼)𝑘

19
• Moving forward one period, firms face a very similar
problem as they did in period t (assumed Poisson process).

• If:
∞
∗
𝑥𝑡 = [1 − 𝛽(1 − 𝛼)] ෍ 𝛽 𝑗 (1 − 𝛼)𝑗 𝐸𝑡 𝑝𝑡+𝑗
𝑗=0

• Then:
∞
∗
𝐸𝑡 𝑥𝑡+1 = [1 − 𝛽(1 − 𝛼)] ෍ 𝛽 𝑗 (1 − 𝛼)𝑗 𝐸𝑡 𝑝𝑡+1+𝑗
𝑗=0

Simply roll forward one period and don’t forget Et on the LHS.

• Recall:
∞
∗
𝑥𝑡 = [1 − 𝛽(1 − 𝛼)] ෍ 𝛽 𝑗 (1 − 𝛼)𝑗 𝐸𝑡 𝑝𝑡+𝑗
𝑗=0

• Let’s separate period t terms (in blue) and future terms (in
green), and note that 𝛽 0 (1 − 𝛼)0 = 1:
∞
∗
𝑥𝑡 = 1 − 𝛽 1 − 𝛼 𝑝𝑡∗ + 1 − 𝛽 1 − 𝛼 ෍ 𝛽 𝑗 (1 − 𝛼)𝑗 𝐸𝑡 𝑝𝑡+𝑗
𝑗=1

∞
∗
= 1−𝛽 1−𝛼 𝑝𝑡∗ + 𝛽 1 − 𝛼 1 − 𝛽 1 − 𝛼 ෍ 𝛽 𝑗 (1 − 𝛼)𝑗 𝐸𝑡 𝑝𝑡+1+𝑗
𝑗=0

= 1 − 𝛽 1 − 𝛼 𝑝𝑡∗ + 𝛽 1 − 𝛼 𝐸𝑡 𝑥𝑡+1

20
 In terms of Inflation
• Subtract pt from both sides:

𝑥𝑡 − 𝑝𝑡 = 1 − 𝛽 1 − 𝛼 𝑝𝑡∗ + 𝛽 1 − 𝛼 𝐸𝑡 𝑥𝑡+1 − 𝑝𝑡

(𝑥𝑡 −𝑝𝑡−1 ) − (𝑝𝑡 −𝑝𝑡−1 ) = 1 − 𝛽 1 − 𝛼 (𝑝𝑡∗ − 𝑝𝑡 ) + 𝛽 1 − 𝛼 (𝐸𝑡 𝑥𝑡+1 − 𝑝𝑡 )

Inflation

• Use: 𝜋𝑡 = 𝛼(𝑥𝑡 −𝑝𝑡−1 ) and its first lead:

𝐸𝑡 𝜋𝑡+1 = 𝛼(𝐸𝑡 𝑥𝑡+1 −𝑝𝑡 )

(𝜋𝑡 /𝛼) − 𝜋𝑡 = 1 − 𝛽 1 − 𝛼 (𝑝𝑡∗ − 𝑝𝑡 ) + 𝛽 1 − 𝛼 (𝐸𝑡 𝜋𝑡+1 / 𝛼)

Using the profit-maximising price (where mt=yt+pt):

𝑝𝑡∗ = 𝜙𝑚𝑡 + (1 − 𝜙)𝑝𝑡 𝑝𝑡∗ − 𝑝𝑡 = 𝜙(𝑦𝑡 +𝑝𝑡 ) − 𝜙𝑝𝑡

𝑝𝑡∗ − 𝑝𝑡 = 𝜙𝑦𝑡

So we can write:

1−𝛼
𝜋𝑡 = 1 − 𝛽 1 − 𝛼 𝜙𝑦𝑡 + 𝛽 1 − 𝛼 (𝐸𝑡 𝜋𝑡+1 / 𝛼)
𝛼

𝛼
𝜋𝑡 = 1 − 𝛽 1 − 𝛼 𝜙𝑦𝑡 + 𝛽𝐸𝑡 𝜋𝑡+1
1−𝛼

21
 The NK Phillips Curve

𝜋𝑡 = 𝛽𝐸𝑡 𝜋𝑡+1 + 𝑘𝑦𝑡

𝛼 1−𝛽 1−𝛼 𝜙
where: 𝑘≡
1−𝛼

• This equation represents the supply-side of the New

Keynesian model; it emerges from our assumptions about
firms, including how they set prices (Calvo model).

• Notice that it features a term in expected future inflation.

• We can easily add a supply shock (𝜀𝑡𝑆 ) and natural level of

output (𝑦)
ത to this expression…

A More Familiar Form

• Rewrite the NK Phillips curve as:

𝜋𝑡 = 𝛽𝐸𝑡 𝜋𝑡+1 + 𝑘(𝑦𝑡 − 𝑦ത𝑡 ) + 𝜀𝑡𝜋

Output gap Supply shock

• 𝛽 is usually calibrated to be close to 1 (≥0.90).

• The supply shock has an expected value of 0.
• We effectively set the natural level of output (𝑦),
ത long-
run value, to zero in our derivation.

22
 Intuition

𝜋𝑡 = 𝛽𝐸𝑡 𝜋𝑡+1 + 𝑘(𝑦𝑡 − 𝑦ത𝑡 ) + 𝜀𝑡𝜋

Inflation today (𝜋𝑡 ) depends upon:

– Expected future inflation because firms must be

forward-looking if/when setting price today.
– The output gap; partly a signal of demand-side pressure
in the economy but real rigidities (𝜙) enter here as well.
– Supply shocks (e.g. oil price shock).

Similar intuition to Year 2 but we have a richer model now.

The Canonical New Keynesian Model

The first two equations should be familiar; use rational
expectations, for example:
1
𝑦𝑡 = 𝑎 + 𝐸𝑡 𝑦𝑡+1 − 𝑟𝑡 + 𝜀𝑡𝐼𝑆 NKIS
𝜃

𝜋𝑡 = 𝛽𝐸𝑡 𝜋𝑡+1 + 𝑘(𝑦𝑡 − 𝑦ത𝑡 ) + 𝜀𝑡𝜋 NKPC

And a third, updated, equation for monetary policy:

𝑖𝑡 = 𝜙𝜋 𝐸𝑡 𝜋𝑡+1 + 𝜙𝑦 𝐸𝑡 𝑦𝑡+1 + 𝜀𝑡𝑀𝑃 MP rule

Notice, no money; we’ll return to monetary policy.

23
 The NK Model in Diagram Form
• The mathematical detail perhaps looks complicated but we
can depict the model quite simply, as follows:

Source: Galí (2018, p.91)

Canonical NK Model: Features

• It is a micro-founded DSGE model with intertemporal
optimisation, forward-looking economic agents and nominal
rigidity; several positive theoretical and empirical attributes
(but see next slide).

• Breaks the ‘classical dichotomy’ via the NK Phillips curve;

allows a role for money to some extent, a central bank and
monetary policy analysis; can be used to analyse ‘stabilisation
policy’ as well (Romer, 1993).

• Nominal rigidity produces an additional internal propagation

mechanism compared to RBC, but still insufficient to match
the empirical evidence; we still need persistent shocks.

24
 Canonical NK Model: Critique
• Although nominal rigidity appears to be important in the ‘real
world’, it is not clear that time-dependent models of price
adjustment – e.g. Calvo’s form – are consistent with the
microeconomic evidence; Chari et al. (2009).

• While RBC models rely heavily on persistent exogenous

technology shocks, perhaps we have simply moved the
arbitrary element elsewhere in the model – ‘the Calvo lottery’.

• The NK model has a strong forward-looking element and the

response to shocks can often be too abrupt; lacks sufficient
internal propagation mechanisms (just like RBC models).

• Several extensions attempt to ‘slow down’ the response/

dynamic adjustment of the model.

Possible Extensions
• The NK Phillips curve with lagged indexation (0≤Ω≤1):

𝜋𝑡 = 1 − Ω 𝜋𝑡−1 + Ω𝐸𝑡 𝜋𝑡+1 + 𝜓(𝑦𝑡 − 𝑦ത𝑡 ) + 𝜀𝑡𝜋

Generates inertia/persistence in inflation and a better

fit to the data; however, micro-foundations not
entirely clear (Chari et al., 2009).

• Add frictions in the labour market, e.g. ‘sticky wages’.

• Allow for incomplete information, e.g. ‘sticky information’
(Mankiw and Reis, 2002).
• Drop rational expectations in favour of an alternative, e.g.
‘learning’ (Evans and Honkapohja, 2008).

25
 Possible Extensions cont’d
• The NKIS is ‘too forward-looking’ to generate realistic
household reactions to shocks. We could add ‘consumption
habits’ to the utility function (e.g. Dennis, 2009); households
are then slower to adjust their consumption plans in
response to a shock.
• What about investment, government spending and net
exports?
• Financial markets and credit conditions, both on the supply
(firms) and demand (households) side of the model; e.g.
investment/capital adjustment costs + financial crises; what
about the role of money in monetary policy (‘Q.E.’)?
• Consider state-dependent models of nominal adjustment
instead, e.g. Golosov and Lucas (2007).

Summary/Conclusions
• We have developed a macroeconomic framework which
incorporates nominal rigidities in a micro-founded way.

• Optimistic view: “we are well on the way to having models of the
macroeconomy that are sufficiently well grounded in
microeconomic assumptions that their parameters can be thought
of as structural… and that are sufficiently realistic that they can be
used to obtain welfare based recommendations about the conduct
of policy.” (Romer, 2019, p.365).

• Pessimistic/realistic view: the model performs poorly against the

data (e.g. forecasting); we need to make ‘adjustments’ to obtain
empirically satisfactory outcomes, but there is a cost to this.

• In future topics we need to think in more detail about

consumption/investment, financial markets + monetary policy.

26