Advanced macroeconomics - Andrea Angell

Contents
1. Introduction .................................................................................................................................... 3
2. Solow Model ....................................................................................................................................... 4
2.1 Steady state................................................................................................................................... 5
2.2 Impact of higher s on output per worker, CD function:................................................................ 8
2.3 Speed of convergence: ................................................................................................................ 10
3. Golden Rule of steady state .............................................................................................................. 11
3.1 Taylor approximation .................................................................................................................. 12
3.2 Example: CD production function ............................................................................................... 13
4. Ramsey model ............................................................................................................................... 14
4.1 Technology .................................................................................................................................. 14
4.2 Firms............................................................................................................................................ 14
4.3 Factor prices ................................................................................................................................ 14
4.4 Households’ budget constraint ................................................................................................... 15
4.5 Utility function ............................................................................................................................ 15
4.6 Lifetime household’s utility......................................................................................................... 16
4.6.1 Household’s maximization problem .................................................................................... 17
4.6.2 Solving maximization problem with a Hamiltonian ............................................................. 17
4.7 The Euler equation, consumption max ....................................................................................... 17
4.8 Dynamics ..................................................................................................................................... 18
4.8.1 Dynamics of consumption given a CD production function ................................................ 18
4.8.2 Dynamics of capital given a CD function .............................................................................. 19
4.8.3 Modified golden rule steady state ....................................................................................... 20
4.9 The effect of a fall in the discount rate ....................................................................................... 21
5. Diamond model................................................................................................................................. 21
5.1 Firms’ assumptions ..................................................................................................................... 22
5.2 Households’ assumptions ........................................................................................................... 22
5.2.1 Household maximization problem ....................................................................................... 23
5.2.2 Solving the household’s problem......................................................................................... 23
5.2.3 The Euler Equation and the savings rate ............................................................................. 23
5.2.4 The savings rate ................................................................................................................... 24
5.3 The dynamics of Capital .............................................................................................................. 25
5.3.1 The evolution of capital with Cobb-Douglas production and log utility .............................. 25
 5.4 The balanced growth path .......................................................................................................... 26
5.5 The possibility of dynamic inefficiency ....................................................................................... 27
5.5.1 Dynamic inefficiency intuition ............................................................................................. 28
6. Endogenous Growth ......................................................................................................................... 28
6.1 The research and development sector ....................................................................................... 29
6.2 Goods production ....................................................................................................................... 29
6.3 Knowledge production ................................................................................................................ 29
6.4 Solow growth model with endogenous K and A ......................................................................... 30
6.5 The relationship for the growth rate .......................................................................................... 30
6.5.1 𝜽 < 𝟏: Diminishing returns, as A increases ......................................................................... 31
6.5.2 𝜽 < 𝟏: Impact of a higher 𝒂𝑳 .............................................................................................. 32
6.5.3 𝜽 > 𝟏: Increasing returns as A increases............................................................................. 33
6.5.5 𝜃 = 1: Constant returns as A increases ............................................................................... 34
6.6 The General Case, that is including capital ................................................................................. 34
6.7 The general case – growth rate of capital................................................................................... 35
6.8 The General Case, growth rate of K ............................................................................................ 35
6.9 The General Case, Dynamics of A ............................................................................................... 36
6.10 The General Case, 𝛽 + 𝜃 < 1 ................................................................................................... 38
6.11 The General, 𝛽 + 𝜃 = 1, 𝑛 = 0................................................................................................. 41
6.12 The nature of Knowledge and the determinants of the allocation of resources to RD............ 42
6.13 Learning by Doing ..................................................................................................................... 42
6.13.1 Learning by doing, example ............................................................................................... 43
6.14 The relationship for the growth rate ........................................................................................ 46
6.13.1 ∅ < 𝟏: Diminishing returns, as K increases ....................................................................... 46
6.13.2 ∅ > 𝟏: Increasing returns, as K increases .......................................................................... 47
6.13.3 ∅ = 𝟏: Constant returns, as K increases ............................................................................ 48
7. Cross-Country Income Differences .................................................................................................. 49
7.1 The Solow Model with Human Capital ....................................................................................... 50
7.1.1 Dynamics of k ....................................................................................................................... 51
7.1.2 What we observe of the dynamics of k(t) ............................................................................ 52
7.2 Decomposing output .................................................................................................................. 53
7.3 What is A? ................................................................................................................................... 54
8. Differences between the chapters.................................................................................................... 56
9. Unemployment ................................................................................................................................. 57
9.1 Search and matching models, introduction: ............................................................................... 57
9.2 The matching function ................................................................................................................ 58
 9.3 Equilibrium conditions ................................................................................................................ 59
9.2 Wage determination ................................................................................................................... 59
9.3 Nash bargaining .......................................................................................................................... 60
9.3.1 Equilibrium conditions ......................................................................................................... 60
9.3.2 Steady state equilibrium ...................................................................................................... 61
9.4 Firm’s vacancy filling rate............................................................................................................ 64
9.4.1 Equilibrium employment level ............................................................................................. 66
9.4.2 Equilibrium employment and job finding rate ..................................................................... 66
9.3 Comparative statics, output increases........................................................................................ 68
Review session about the exam............................................................................................................ 70

1. Introduction
In class exercise

Find 𝑓(𝑘) when 𝛼 = 0,5 and draw the curve between 𝑘 = 0 and 𝑘 = 10.

k (x) 0 1 4 9
y 0 1 2 3

𝑌 = 𝑘𝛼
Because

𝛼 = 0,5

Cobb Douglas
3.5
3
2.5
2
y

1.5
1
0.5
0
0 2 4 6 8 10
k

𝑲(𝒕)
Development in capital per unit of effective labor 𝒌(𝒕) = 𝑨(𝒕)𝑳(𝒕)
k: increase over time in capital measured in efficiency of labor (if this increase=0, then we’re in a
steady state). In discrete time 𝑘𝑡+1 = 𝑘𝑡 , not the case here because of continuous time.
𝐾̇
Step 1: differentiate in respect to time. The dot means it’s in respect to time.
𝐴𝐿

𝐾̇ 𝑘̇ 𝐴𝐿 − 𝐾(𝐴̇𝐿 + 𝐴𝐿̇)
( )=
𝐴𝐿 (𝐴𝐿)2

𝐾̇ 𝐴𝐿 𝐾 𝐴̇𝐿 + 𝐴̇𝐿
= − ∗
(𝐴𝐿)(𝐴𝐿) 𝐴𝐿 𝐴𝐿
𝐾̇ 𝐴̇ 𝐿̇
= −𝑘( + )
𝐴𝐿 𝐴 𝐿

What od we know about K dot?

𝑌
Small y is 𝑦 = 𝐴𝐿 so output measured in efficiency units.
𝐾
Small k is 𝑘 = 𝐴𝐿 so capital measured in efficiency units.

𝑠𝑌 − 𝛿𝐾
= − 𝑘(𝑔 + 𝑛) = 𝑠𝑦 − 𝛿𝑘 − 𝑘(𝑔 + 𝑛)
𝐴𝐿

Model for Solow growth model:

= 𝒔𝒇(𝒌) − (𝒈 + 𝒏 + 𝜹)𝒌

2. Solow Model
𝛿𝑥(𝑡)
Continuous time is denoted with a dot. This means 𝑥̇ (𝑡) = 𝛿𝑡
.

X-axis:
𝐾(𝑡)
𝑘(𝑡) =
𝐴(𝑡)𝐿(𝑡)
y-axis:
𝛼
𝑦(𝑡) = (𝑘(𝑡))

Key equation:

𝑘(𝑡) = 𝑠𝑓(𝑘(𝑡)) − (𝑛 + 𝑔 + 𝛿)𝑘(𝑡)

Savings curve = actual investment(

Breakeven investment
In the steady state, k(t)=0, which means that

𝑠𝑓(𝑘(𝑡)) = (𝑛 + 𝑔 + 𝛿)𝑘(𝑡)

Exercise Solow growth model:

Assume a Cobb Douglas production function

𝑌 = 𝐹(𝐾, 𝐴𝐿) = (𝐾)𝛼 (𝐴𝐿)1−𝛼

1. Find f(k) when alpha=0,5 and draw the curve

→ did in lecture 1

2. Let s=0,2 and draw the actual investment curve

𝑘̇ (𝑡) = 𝑠 ∗ √𝑘(𝑡) − (𝑛 + 𝑔 + 𝛿)𝑘(𝑡)

3. Let n=0,02 (population growth of 2 percent), g=0,02 (knowledge growth of 2 percent),

depreciation rate 𝛿 = 0,06 and draw the break-even point

𝑘̇ (𝑡) = 𝑠 ∗ √𝑘(𝑡) − (𝑛 + 𝑔 + 𝛿)𝑘(𝑡)

Savings are equal to investments due to the assumption of a closed economy.

k(t) y(t) Savings breakeven

0 0 0 0
1 1 0,2 0,1
4 2 0,4 0,4
9 3 0,6 0,9

Steady state: no growth in capital measured in efficiency units of labor

𝑘 ∗ = 𝑠𝑡𝑒𝑎𝑑𝑦 𝑠𝑡𝑎𝑡𝑒 = 5

2.1 Steady state

𝒌̇(𝒕) = 𝒔𝒇(𝒌(𝒕)) − (𝒈 + 𝒏 + 𝜹)𝒌(𝒕)

𝐴̇(𝑡)
= 𝑔 ⟺ 𝐴(𝑡) = 𝐴(0) ∗ 𝑒 𝑔𝑡
𝐴(𝑡)
𝐿̇(𝑡)
= 𝑛 ⟺ 𝐿(𝑡) = 𝐿(0)𝑒 𝑛𝑡
𝐿(𝑡)
L(t) how much population you have in economy
𝐾(𝑡)
𝑘(𝑡) =
𝐴(𝑡)𝐿(𝑡)
When high population growth (g) → L(t) increases a lot, the whole term of k(t)\dot falls. We draw it
into the diagram to get it two dimensional.

1) 𝑠 ↑→ 𝐼 ↑→ 𝐾 ↑→ 𝑘 ↑

2) 𝛿 ↑→ 𝐾 ↓→ 𝑘 ↓

𝑛 ↑→ 𝐿 ↑→ 𝑘 ↓
𝑔 ↑→ 𝐴 ↑→ 𝑘 ↓

In general, we have that, in the steady state

𝑘̇ (𝑡) = 𝑠𝑓(𝑘(𝑡)) − (𝑔 + 𝑛 + 𝛿)𝑘(𝑡)

𝑘̇ (𝑡) = 0
𝑠𝑓(𝑘 ∗) = (𝑔 + 𝑛 + 𝛿)𝑘 ∗

𝑠𝑘 ∗0,5−1 = (𝑔 + 𝑛 + 𝛿)
1 (𝑔 + 𝑛 + 𝛿)
0,5
=
𝑘∗ 𝑠
𝑠 2
𝑘 ∗= ( )
(𝑔 + 𝑛 + 𝛿)

When 𝑘(𝑡) = 0 ⟺ 𝑠√𝑘 ∗ = (𝑛 + 𝑔 + 𝛿)𝑘 ∗

𝑘(𝑡) = 0 ⟺

0,2 ∗ √𝑘 ∗ = 0,1 ∗ 𝑘 ∗ ⟺
Solve for k^*
0,2
⟺ = √𝑘 ∗
0,1
⟺ 𝑘∗ = 4
Find y*
𝛼
𝑦 ∗ (𝑡) = (𝑘 ∗ (𝑡)) ⟺

𝑦 ∗ = √4 = 2
 Solow growth model
3.5

2.5

2 y(t)
y(t)

1.5 Savings

1 breakeven
steady state
0.5

0
0 2 4 6 8 10
k(t)

Interpretation: there will always be such a steady state, it’s unique.

Assumption: 𝑓 ′ (𝑘) > 0, 𝑓 ′′ (𝑘) < 0

Savings have to be higher than the break-even investment for us to have a growing economy.

If first inada condition doesn’t hold: lim 𝑓 ′ (𝑘) = ∞

Doesn’t hold, wont get equilibrium above zero

𝑠 ∗ 𝑦 < (𝑛 + 𝑔 + 𝛿)𝑘 for all k.

Second inada condition doesn’t hold: lim 𝑓 ′ (𝑘) = 0

𝑘→∞
Dynamics of model:

𝒌̇(𝒕) = 𝒔𝒇(𝒌(𝒕)) − (𝒈 + 𝒏 + 𝜹)𝒌(𝒕)

Growth rate in output per worker in the steady state:

We have that
𝑌
𝑦=
𝐴𝐿
𝑌
=𝑦∗𝐴
𝐿
𝑌
= 𝑓(𝑘) ∗ 𝐴
𝐿
The growth rate per worker is as in a steady state ′𝑘 = 0.
′𝑌
= 𝑓 ′ (𝑘)′ 𝑘𝐴 + 𝑓(𝑘)′ 𝐴 = 0 + 𝑓(𝑘)𝑔𝐴 = 𝑔𝑌/𝐿
𝐿
 ′𝑌 𝑌
/ =𝑔
𝐿 𝐿
See mock exam 2019.

Another way of deriving the same thing:

Balanced growth in steady state

𝐾
𝑘=
𝐴𝐿
𝑌 𝑌
𝑦= ⟺ 𝑦 ∗ 𝐴 = = 𝑓(𝑘) ∗ 𝐴
𝐴𝐿 𝐿
GDP per capital (L)

Y/L output per worker. Differentiation of a product. K depends on time.

′𝑌
= 𝑓 ′ (𝑘)′𝑘𝐴 + 𝑓(𝑘)𝐴̇
𝐿
= 0 + 𝑓(𝑘) ∗ 𝐴̇
The growth rate in A is equal to g (by definition)

= 𝑓(𝑘) ∗ 𝑔 ∗ 𝐴
𝑌
= ∗𝑔∗𝐴
𝐴𝐿
𝑌
= ∗𝑔
𝐿
′𝑌
⟺ 𝐿 =𝑔
𝑌
𝐿
Even when they reach steady state there will positive growth (even though it looks like there is no
growth rate), but if we measure in output per worker we get a growth rate.

2.2 Impact of higher s on output per worker, CD function:

f(k) = 𝑘 𝛼
The elasticity of output wrt capital:

𝑘 𝑑𝑦 𝑓′(𝑘) 𝛼𝑘 𝛼−1
= 𝑘= 𝑘=𝛼
𝑦 𝑑𝑘 𝑓(𝑘) 𝑘𝛼
Therefore,
𝑌′
𝐿 = (𝛼 𝑘′ + 𝑔) > 𝑔
𝑌 𝑘
𝐿
Golden rule steady state:

Consumption is output=income minus savings:

𝐶 𝑌 𝑆
𝑐= = − = 𝑦 − 𝑠𝑦 = (1 − 𝑠)𝑓(𝑘)
𝐴𝐿 𝐴𝐿 𝐴𝐿
When s increases, consumption falls immediately.

If capital, C, increases, (output), consumption increases.

Whether consumption is larger or smaller in the new steady state, depends on whether we are
above or below the Golden Rule steady state.

The golden rule saving rate is the saving rate which maximises consumption in steady state:

1. If s=0 what are k and c?

2. If s=1 what are k and c?

In steady state:

𝑐 ∗= (1 − 𝑠)𝑓(𝑘 ∗) = 𝑓(𝑘 ∗) − 𝑠𝑓(𝑘 ∗) = 𝑓(𝑘 ∗) − (𝑛 + 𝑔 + 𝛿)𝑘 ∗

, where we have used the steady state condition given in equation (2). That is, we make sure that it
is the steady state.

- Graphically: Consumption is the difference between the output curve and the saving=actual
investment curve.
- Steady state requires that the saving curve intersects the break-even curve.

The saving rate, which maximises consumption by differentiating consumption with respect to s and
equalise zero:
𝑑𝑐 ∗ 𝜕𝑘 ∗
= (𝑓′(𝑘 ∗𝐺 ) − (𝑛 + 𝑔 + 𝛿)) =0
𝑑𝑠 𝜕𝑠
The corresponding golden rule steady state capital level we denote by 𝑘 ∗𝐺

1. First term: f’(k*) is the marginal productivity of capital

2. Second term: breakeven investments

- Graphically: slope of the output curve and break even curve are equal.
- Above 𝑘 ∗𝐺 , the slope of the output curve is lower than the slope of the break-even curve,
𝑓′(𝑘 ∗) < (𝑛 + 𝑔 + 𝛿). As the marginal productivity is decreasing, consumption can increase
by reducing k*.
- Below 𝑘 ∗𝐺 , consumption is also below max consumption, here the slope of the output curve
I higher than the slope of break-even curve, 𝑓′(𝑘 ∗) > (𝑛 + 𝑔 + 𝛿). As the marginal
productivity is decreasing, consumption can increase by increasing k*.

Example of Golden rule with Cobb-Douglas function:

With a CD production function, we can obtain, 𝑦 = 𝑓(𝑘) = 𝑘 𝛼

 𝑑𝑐 ∗ 𝜕𝑘 ∗
= (𝛼(𝑘 ∗𝐺 )𝛼−1 − (𝑛 + 𝑔 + 𝛿)) =0
𝑑𝑠 𝜕𝑠
↔
1
𝛼 1−𝛼
= 𝑘 ∗𝐺
𝑛+𝑔+𝛿
Remember that the steady state solution or a CD function is
1
𝑠 1−𝛼
𝑘 ∗=
𝑛+𝑔+𝛿
Then we compare the two last functions_

1 1
𝛼 1−𝛼 𝑠 1−𝛼
= 𝑘 ∗𝐺 = 𝑘 ∗=
𝑛+𝑔+𝛿 𝑛+𝑔+𝛿
This gives us that the golden rule savings rate is:

𝑠𝐺 = 𝛼

How long time does it take the variable k to reach the steady state, k*?

As k’=k(k), k dot is a function of capital

𝑘′(𝑘) = 𝑠𝑓(𝑘) − (𝑛 + 𝑔 + 𝛿)𝑘

When k=k*, is k=0.

A first order Taylor approximation of k’=k’(k) (dot), around k=k*, gives:

𝜕𝑘′(𝑘) 𝜕𝑘′(𝑘)
𝑘′ ≈ 𝑘′(𝑘 = 𝑘 ∗) + |𝑘=𝑘∗ (𝑘 − 𝑘 ∗) = | (𝑘 − 𝑘 ∗)
𝜕𝑘 𝜕𝑘 𝑘=𝑘∗
Hence, k’ (dot), is equal to the product of the difference between k and k* and the derivative of k
with respect to k at k=k*

2.3 Speed of convergence:

𝜕𝑘′(𝑘)
1. K’ is positive when k is a bit below k* and negative when it is a bit above k*, hence <
𝜕𝑘
0.
𝜕𝑘′(𝑘)
2. Let 𝜆 = − |𝑘=𝑘∗ > 0, so that 𝑘′(𝑡) ≈ −𝜆(𝑘(𝑡) − 𝑘 ∗).
𝜕𝑘

1. Close to the Steady state, k is approaching k* at a speed which is proportional to the

distance to k*.
2. The growth rate of k(t)-k* is constant and equal to - 𝜆.
 3. The solution to the equation in 2 is:

𝑘(𝑡) − 𝑘 ∗= 𝑒 −𝜆𝑡 (𝑘(0) − 𝑘 ∗) ↔ 𝑘(𝑡) = 𝑘 ∗ +𝑒 −𝜆𝑡 (𝑘(0) − 𝑘 ∗)

K approaches k* at the speed 𝜆 = (𝑛 + 𝑔 + 𝛿)(1 − 𝛼𝐾 (𝑘 ∗))

3. Golden Rule of steady state

Golden rule save rate

𝑆𝐺 = 𝑥
Max consumption is where the distance between the production function and the savings function is
max (blue parenthesis on graph below): I THINK THIS IS WRONG

Solow growth model

3.5

2.5

2 y(t)
y(t)

1.5 Savings

1 breakeven
steady state
0.5

0
0 2 4 6 8 10
k(t)

I𝑖𝑛𝑐𝑜𝑚𝑒 = 𝑐𝑜𝑠𝑡 + 𝑠𝑎𝑣𝑖𝑛𝑔𝑠

⟺ 𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑜𝑛 = 𝑐𝑜𝑠𝑡 + 𝑠𝑎𝑣𝑖𝑛𝑔𝑠

1 1
If 𝛼 = 2 it means we have to save 50% of income, which is not realistic. If 𝛼 = 3 then save 33% of
income, which is more realistic. Once we include human capital, the function will look different.

Solow model: how long does it take to get half way to steady state? → we take a Taylor
approximation.

𝑐 = (1 − 𝑠)𝑦 = 𝑦 − 𝑠𝑦
Why important? Suppose we want to increase s: for the present generation 𝑠 ↑→ 𝑐 ↓. How long time
does it take before we are better off? → when savings increase 𝑠 ↑→ 𝑖𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡𝑠 ↑→
𝑐𝑎𝑝𝑖𝑡𝑎𝑙 𝑠𝑡𝑜𝑐𝑘 ↑→ 𝑦 ↑→ 𝑐 ↑→ 𝑓𝑢𝑡𝑢𝑟𝑒 𝑔𝑒𝑛𝑒𝑟𝑒𝑎𝑡𝑖𝑜𝑛𝑠 𝑎𝑟𝑒 𝑏𝑒𝑡𝑡𝑒𝑟 𝑜𝑓𝑓

3.1 Taylor approximation

As 𝑘̇ = 𝑘(𝑘)

= 𝑠𝑓(𝑘) − (𝑛 + 𝑔 + 𝑑)𝑘
When 𝑘 = 𝑘 ∗ then 𝑠𝑓(𝑘 ∗ ) = (𝑛 + 𝑔 + 𝑑)𝑘 ∗

A first order Taylor approximation of 𝑘̇ = 𝑘̇ (𝑘) around 𝑘̇ = 0, fthat is 𝑘 = 𝑘̇ , is given:

𝛿𝑘̇ (𝑘)
𝑘̇ = 𝑘̇ (𝑘 = 𝑘 ∗ ) + (𝑘 − 𝑘 ∗ )
𝛿𝑘
How many growth do we have in growth measured in efficiency → idea is to linear the production
function in order to get an idea how long does it take to get from the point we are right now k=0 to a
steady state.

𝛿𝑘̇ (𝑘)
𝑘̇ = 𝑘̇ (𝑘 = 𝑘 ∗ ) + (𝑘 − 𝑘 ∗ )
𝛿𝑘
So 𝑘̇ is positive when k is a bit below 𝑘 ∗ (because of increase in marginal capital, so when a little bit
below in the steady state, then an increase in S is larger than breakeven investment, why kdot
̇
𝛿𝑘 (𝑘)
negative) and above 𝑘 ∗ 𝑘̇ is negative, hence 𝛿𝑘 < 0.

𝛿𝑘̇ (𝑘)
Let 𝜆 = −
𝛿𝑘

It’s the slope measured in the steady state.

→ 𝑘̇ (𝑡) = −𝜆(𝑘(𝑡) − 𝑘 ∗ ) (equation 6)

Substitute derivative of kdot with respect to k into minus lambda.

• Close to the steady state, k is approaching 𝑘 ∗ at a speed which is proportional to the

distance to 𝑘 ∗.
• The solution to equation 6 is:

𝑘(𝑡) − 𝑘 ∗ = 𝑒 −𝜆𝑡 (𝑘(0) − 𝑘 ∗ )

⟺ 𝑘(𝑡) = 𝑘 ∗ + 𝑒 −𝜆𝑡 (𝑘(0) − 𝑘 ∗ )

Find lambda (idea of the speed of converges)

So need to find 𝜆 to find the speed of convergence:

𝛿𝑘̇ (𝑘)
𝜆=−
𝛿𝑘
= −(𝑠𝑓 ′ (𝑘) − (𝑛 + 𝑔 + 𝛿))

= (𝑛 + 𝑔 + 𝛿) − 𝑠𝑓 ′ (𝑘)
Simplify by substituting in for the s → know from the steady state condition: 𝑠𝑓(𝑘) =
(𝑛 + 𝑔 + 𝛿)𝑘
 (𝑛 + 𝑔 + 𝛿)𝑘
⟺𝑠=
𝑓(𝑘)
Substitute into expression
𝑛+𝑔+𝛿 ′
𝜆 = (𝑛 + 𝑔 + 𝛿) − 𝑘𝑓 (𝑘)
𝑓(𝑘)
𝑘𝑓 ′ (𝑘)
𝜆 = (𝑛 + 𝑔 + 𝛿) (1 − )
𝑓(𝑘)

Elasticity of output in capital

λ = (𝑛 + 𝑔 + 𝛿)(1 − 𝛼)

3.2 Example: CD production function

𝛼 = 𝛼𝐾 = 1/3
𝑛 = 0,01
𝑔 = 0,02
𝛿 = 0,03
Question: how long time does it take to get half way to steady state?
𝑘(𝑡)−𝑘 ∗ 1
➔ Figure out if 𝐾(0)−𝑘∗ = 2 (half way to steady state)
𝑘(𝑡)−𝑘 ∗
➔ This tells how long time it really takes (speed of converges) if = 𝑒 𝜆𝑡
𝐾(0)−𝑘 ∗

1
λ = (𝑛 + 𝑔 + 𝛿)(1 − 𝛼) = (0,01 + 0,02 + 0,03) (1 − ) = 𝟎, 𝟎𝟒
3
Then we solve for t when:
𝑘(𝑡) − 𝑘 ∗ 1
=
𝐾(0) − 𝑘 ∗ 2
𝑘(𝑡) − 𝑘 ∗
= 𝑒 𝜆𝑡
𝐾(0) − 𝑘 ∗
Set equal to one another:
1
= 𝑒 −𝜆𝑡
2
1
= 𝑒 −0,04∗𝑡
2
1
ln ( ) = −0,04 ∗ 𝑡
2
𝒕 = 𝟏𝟕, 𝟑𝟐
Number of years to get halfway to the steady state is 17,32 years.

Solow model assumptions

• No micro foundation
 4. Ramsey model
Look at retake exam 2019

• Micro foundation
• Household max utility
• Firms max profits
• People live forever and they are all identical
• No government

4.1 Technology
𝐴′(𝑡)
1. Technology grows at the rate g: =𝑔
𝐴(𝑡)
2. There is no depreciation of capital, 𝛿 = 0
3. Capital stock is always positive, 𝐾(𝑡) > 0
4. Output can be used for consumption or savings, which are equal to investments, hence:
𝐾′(𝑡) 𝑌(𝑡) 𝜍(𝑡)
𝐻
= 𝐻
− 𝐻
, where 𝜍(𝑡) is total consumption (C(t) is consumption per person).

4.2 Firms
- The production side is like in the Solow model
- Perfect competition
- Knowledge is for free
- We have that firms maximize profits (dropping the time index for convenience)

Max 𝜋 = 𝐹(𝐾, 𝐴𝐿) − 𝑊𝐿 − 𝑟𝐾

First Order Condition:

𝛿𝜋 𝛿𝐹(𝐾, 𝐴𝐿)
=0⟺ −𝑟 =0
𝛿𝐾 𝛿𝐾
𝛿𝜋 𝛿𝐹(𝐾, 𝐴𝐿)
=0⟺ −𝑤 = 0
𝛿𝐿 𝛿𝐿
4.3 Factor prices
Marginal productivity of production factors = price of production factors.
𝐾 𝐾
We want capital measured in efficiency units of labor: 𝑘 = 𝐴𝐿 → 𝐹(𝐾, 𝐴𝐿) = 𝐴𝐿 ∗ 𝐹 (𝐴𝐿 , 1)

⟺ 𝐹(𝐾, 𝐴𝐿) = 𝐴𝐿 ∗ 𝑓(𝑘)

First condition, differentiate in regard to K
𝛿𝜋 𝛿𝐹(𝐾, 𝐴𝐿)
=0⟺ −𝑟 =0
𝛿𝐾 𝛿𝐾
𝑓 ′ (𝑘)1
𝐴𝐿 ∗ − 𝑟 = 0 ⟺ 𝑓 ′ (𝑘) = 𝑟
𝐴𝐿
Second condition, differentiate in regard to L
𝛿𝜋 𝛿𝐹(𝐾, 𝐴𝐿)
=0⟺ −𝑤 = 0
𝛿𝐿 𝛿𝐿
(production differentiation rule)
𝑘
𝐴𝑓(𝑘) + 𝐴𝐿 ∗ 𝑓 ′ (𝑘) ∗ − 𝐴−𝑊 =0
𝐴𝐿2
𝑘
𝐴 (𝑓(𝑘) + 𝐴𝐿𝑓(𝑘) ∗ − )=𝑊
𝐴𝐿2
𝑘
𝐴 (𝑓(𝑘) + 𝑓 ′ (𝑘) − )=𝑊
𝐴𝐿
𝑊
𝑓(𝑘)𝑘 =
𝐴
4.4 Households’ budget constraint
- The budget constraint of a household in each period is

𝑐𝑜𝑛𝑠𝑢𝑚𝑝𝑡𝑖𝑜𝑛 + 𝑖𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡 = 𝑙𝑎𝑏𝑜𝑢𝑟 𝑖𝑛𝑐𝑜𝑚𝑒 + 𝑐𝑎𝑝𝑖𝑡𝑎𝑙 𝑖𝑛𝑐𝑜𝑚𝑒 + 𝑝𝑟𝑜𝑓𝑖𝑡

Left side: total expenditures
Right side: total income
𝐿(𝑡)
- Household consumption: 𝐶(𝑡) 𝐻
𝐾′(𝑡)
- Household investment:
𝐻
𝐿(𝑡)
- Household labor income: 𝑊(𝑡) 𝐻
𝐾(𝑡)
- Household capital income: 𝑟(𝑡) 𝐻
- Household profits are zero as we have free entry

𝐿(𝑡) 𝐾′(𝑡) 𝐿(𝑡) 𝐾(𝑡)

𝐶(𝑡) + = 𝑊(𝑡) + 𝑟(𝑡)
𝐻 𝐻 𝐻 𝐻
𝐾(𝑡)
Divide by 𝐴(𝑡)𝐿(𝑡), 𝑚𝑢𝑙𝑡𝑖𝑝𝑙𝑦 𝑤𝑖𝑡ℎ 𝐻 𝑎𝑛𝑑 𝑢𝑠𝑒 𝑘(𝑡) = (𝐴(𝑡)𝐿(𝑡))

Make loads of calculations and get:

𝑘′ + (𝑔 + 𝑛)𝑘 = 𝑤(𝑡) + 𝑟(𝑡)𝑘(𝑡) − 𝑐(𝑡)

To obtain the households’ budget constraint in each period we get:

𝑘′ = 𝑤(𝑡) + (𝑟(𝑡) − 𝑔 − 𝑛)𝑘(𝑡) − 𝑐(𝑡)

4.5 Utility function

The utility function is a CRRA (constant relative risk aversion) function:

𝐶(𝑡)1−𝜃
𝑢(𝐶(𝑡)) = ,𝜃 > 0
1−𝜃
1. 𝜃 measures how willing consumers are to substitute consumption between periods
2. Positive and decreasing marginal utility:
𝑢′(𝐶(𝑡)) = 𝐶(𝑡)−𝜃 > 0, 𝑢′′(𝐶(𝑡)) = −𝜃𝐶(𝑡)−𝜃−1 < 0.
3. If 𝜃 = 1 is u(C(t))=lnC(t)
 4. If 𝜃 = 0 is the utility function linear and people are indifferent between consuming now or
in another period
5. The degree of CRRA is found as
𝑢′′(𝐶(𝑡)) 𝜃𝐶(𝑡)−𝜃−1
−𝐶(𝑡) = 𝐶(𝑡) =𝜃
𝑢′(𝐶(𝑡)) 𝐶(𝑡)−𝜃
Let us now convert the utility function to one with consumption per unit of effective labour.
𝐶(𝑡)
𝑐(𝑡) = ↔ 𝐴(𝑡)𝑐(𝑡)
𝐴(𝑡)

𝐶(𝑡)1−𝜃 [𝐴(𝑡)𝑐(𝑡)]1−𝜃 [𝐴(0)𝑒 𝑔𝑡 ]1−𝜃 𝑐(𝑡)1−𝜃

= = =
1−𝜃 1−𝜃 1−𝜃
𝑐(𝑡)1−𝜃
𝐴(0)1−𝜃 𝑒 (1−𝜃)𝑔𝑡
1−𝜃

4.6 Lifetime household’s utility

The lifetime utility is found by integrating over all periods
∞
𝐶(𝑡)1−𝜃 𝐿(𝑡)
𝑈 = ∫ 𝑒 −𝜌𝑡 𝑑𝑡
1−𝜃 𝐻
0

, where 𝜌 measures by how much each household discounts the future. High 𝜌 implies little value on
the future. Rewrite per effective unit of labor:
∞
𝑐(𝑡)1−𝜃 𝐿(𝑡)
𝑈 = ∫ 𝑒 −𝜌𝑡 𝐴(0)1−𝜃 𝑒 (1−𝜃)𝑔𝑡 𝑑𝑡
1−𝜃 𝐻
0

, and insert for 𝐿(𝑡) = 𝐿(0)𝑒 𝑛𝑡

∞
𝑐(𝑡)1−𝜃 𝐿(0)𝑒 𝑛𝑡
𝑈 = ∫ 𝑒 −𝜌𝑡 𝐴(0)1−𝜃 𝑒 (1−𝜃)𝑔𝑡 𝑑𝑡
1−𝜃 𝐻
0

We can now collect terms, time invariant parts outside the integral:

∞
1−𝜃
𝐿(0) 𝑐(𝑡)1−𝜃
𝑈 = 𝐴(0) ∫ 𝑒 −𝜌𝑡+(1−𝜃)𝑔𝑡+𝑛𝑡 𝑑𝑡
𝐻 1−𝜃
0
∞
𝑐(𝑡)1−𝜃
𝑈 = 𝐵 ∫ 𝑒 −𝛽𝑡 𝑑𝑡
1−𝜃
0

𝐿(0)
, where 𝐵 = 𝐴(0)1−𝜃 𝐻
, and 𝛽 = 𝜌 − (1 − 𝜃)𝑔 − 𝑛 > 0. Positive Beta implies that the utility
value does not explode.
4.6.1 Household’s maximization problem
The household maximizes lifetime utility
∞
𝑐(𝑡)1−𝜃
𝑈 = 𝐵 ∫ 𝑒 −𝛽𝑡 𝑑𝑡
1−𝜃
0

Subject to the dynamic budget constraint:

′𝑘(𝑡) = 𝑤(𝑡) + [𝑟(𝑡) − 𝑛 − 𝑔]𝑘(𝑡) − 𝑐(𝑡)

and a transversely condition: households cannot leave the economy with debt: (No-Ponzi-Game
condition). And with positive values of capital, both initially k(0)>0 and at any point in time 𝑡, 𝑘(𝑡) >
0. Both the integration and the dynamic budget constraint are differentiable and concave.

4.6.2 Solving maximization problem with a Hamiltonian

𝑐(𝑡)1−𝜃
ℋ = 𝑒 −𝛽𝑡 + 𝜇(𝑡)[𝑤(𝑡) + [𝑟(𝑡) − 𝑛 − 𝑔]𝑘(𝑡) − 𝑐(𝑡)]
1−𝜃
Necessary and sufficient conditions for optimality:

- (1) Differentiate with respect to the control c(t)

𝜕ℋ
=0
𝜕𝑐(𝑡)

𝑒 −𝛽𝑡 𝑐(𝑡)−𝜃 − 𝜇(𝑡) = 0

- (2) Differentiate with respect to the state variable k(t)
𝜕ℋ
= −′𝜇(𝑡)
𝜕𝑘(𝑡)
𝜇(𝑡)[𝑟(𝑡) − 𝑛 − 𝑔] = −′𝜇(𝑡)
- (2) Differentiate with respect to the co-state variable 𝜇(𝑡)
𝜕ℋ
= ′𝑘(𝑡)
𝜕𝜇(𝑡)
[𝑤(𝑡) + [𝑟(𝑡) − 𝑛 − 𝑔]𝑘(𝑡) − 𝑐(𝑡)] = ′𝑘(𝑡)
- (3) Household budget constraint
′𝑘(𝑡) = 𝑤(𝑡) + [𝑟(𝑡) − 𝑛 − 𝑔]𝑘(𝑡) − 𝑐(𝑡)

Transversally condition:

lim 𝑘(𝑡)𝜇(𝑡) = 0
𝑛→∞

4.7 The Euler equation, consumption max

(1) 𝑇𝑎𝑘𝑒 𝑙𝑜𝑔 𝑜𝑓 𝑡ℎ𝑒 𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛:
−𝛽 − 𝜃𝑙𝑛𝑐(𝑡) = 𝑙𝑛𝜇(𝑡)
 𝐷𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑡𝑖𝑎𝑡𝑒 𝑤𝑟𝑡 𝑡𝑖𝑚𝑒:
′𝑐(𝑡) ′𝜇(𝑡)
−𝛽 − 𝜃 =
𝑐(𝑡) 𝜇(𝑡)
𝑊𝑒 𝑡ℎ𝑒𝑛 𝑠𝑢𝑏𝑠𝑡𝑖𝑡𝑢𝑡𝑒 𝑡ℎ𝑒 𝑟𝑖𝑔ℎ𝑡 𝑠𝑖𝑑𝑒 𝑤𝑖𝑡ℎ 𝑡ℎ𝑒 𝑑𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑡𝑖𝑎𝑡𝑖𝑜𝑛 𝑓𝑟𝑜𝑚 𝑑𝑖𝑓𝑓 𝑘(𝑡)

′𝑐(𝑡) [𝑟(𝑡) − 𝑛 − 𝑔] − 𝛽 ′𝑐(𝑡)

−𝛽 − 𝜃 = −[𝑟(𝑡) − 𝑛 − 𝑔] ↔ =
𝑐(𝑡) 𝜃 𝑐(𝑡)

𝑊𝑒 𝑡ℎ𝑒𝑛 𝑠𝑢𝑏𝑠𝑡𝑖𝑡𝑢𝑡𝑒 𝑓𝑜𝑟 𝛽 𝑎𝑛𝑑 𝑟𝑒𝑚𝑒𝑚𝑏𝑒𝑟 𝑟(𝑡) = 𝑓′(𝑘(𝑡)) 𝑡𝑜 𝑜𝑏𝑡𝑎𝑖𝑛 𝑡ℎ𝑒 𝐸𝑢𝑙𝑒𝑟 𝐸𝑞𝑢𝑎𝑡𝑖𝑜𝑛
′𝑐(𝑡) 𝑓′(𝑘(𝑡)) − 𝜌 − 𝜃𝑔
=
𝑐(𝑡) 𝜃
Evolution in consumption per worker in effective labor units, is the equation above.

Evolution in consumption per worker is, as c(t)=C(t)/A(t)

′𝐶(𝑡) ′𝑐(𝑡) ′𝐴(𝑡) 𝑟(𝑡) − 𝜌 − 𝜃𝑔 𝑓′(𝑘(𝑡)) − 𝜌
= + = =
𝐶(𝑡) 𝑐(𝑡) 𝐴(𝑡) 𝜃 𝜃
- Consumption is rising if the rental rate of capital, 𝑟(𝑡) = 𝑓′(𝑘(𝑡)) exceeds the discount rate
of future consumptions, p, and vice-versa otherwise.
- The smaller 𝜃 (very likely to substitute between periods), the larger are the changes in
consumption in response to differences between the rental rate of capital and the discount
rate.

4.8 Dynamics
Evolution in consumption per worker in effective labour units is

′𝑐(𝑡) 𝑓 ′ (𝑘(𝑡)) − 𝜌 − 𝜃𝑔
=
𝑐(𝑡) 𝜃
The third condition for the Hamiltonian was the household budget constraint

′𝑘(𝑡) = (𝑤(𝑡) + (𝑟(𝑡) − 𝑛 − 𝑔)𝑘(𝑡) − 𝑐(𝑡))

Use conditions from profitmax: f’(k(t))=r(t) and

𝑓(𝑘(𝑡)) − 𝑓′(𝑘(𝑡))𝑘(𝑡) = 𝑤(𝑡) to obtain

′𝑘(𝑡) = [𝑓(𝑘(𝑡)) − [𝑛 + 𝑔]𝑘(𝑡) − 𝑐(𝑡)]

Hence, we have two differential equations determining the dynamics of c(t) and k(t). We can analyse
the equilibrium using a phase diagram

4.8.1 Dynamics of consumption given a CD production function

𝑌 = 𝐹(𝐾, 𝐴𝐿) = (𝐾)𝛼 (𝐴𝐿)1−𝛼
Gives the production function per effective unit of labour
 𝑓(𝑘(𝑡)) = 𝑘(𝑡)𝛼

The marginal productivity of capital is then

𝑓 ′ (𝑘(𝑡)) = 𝛼𝑘(𝑡)𝛼−1

′𝑐(𝑡) 𝛼𝑘 ∗𝛼−1 − 𝜌 − 𝜃𝑔
= =0
𝑐(𝑡) 𝜃
1
𝛼 1−𝛼
𝑘 ∗= ( )
𝜌 + 𝜃𝑔
That is, a vertical line.

4.8.2 Dynamics of capital given a CD function

See retake exam 2019 or original exam 2019

The equilibrium condition giving the evolution of capital was

′𝑘(𝑡) = 𝑘(𝑡)𝛼 − (𝑛 + 𝑔)𝑘(𝑡) − 𝑐(𝑡)

The curve where ‘k(t)=0 to draw into a phase diagram is then

′𝑘(𝑡) = 0
𝑐(𝑡) = 𝑘(𝑡)𝛼 − (𝑛 + 𝑔)𝑘(𝑡)
Shape of the curve:

At c=0, k=0
𝜕𝑐
= 𝑓 ′ (𝑘) − (𝑛 + 𝑔)
𝜕𝑘
We denote the value giving
𝜕𝑐
=0
𝜕𝑘
𝑑𝑐 ∗
= 𝛼𝑘(𝑡)𝛼−1 − (𝑛 + 𝑔) = 0
𝑑𝑘
𝛼𝑘𝐺𝑅 𝛼−1 = (𝑛 + 𝑔)
1
𝛼 𝛼
𝑘𝐺𝑅 =( )
(𝑛 + 𝑔)

By 𝑘𝐺𝑅
𝜕𝑐
- For 𝑘 < 𝑘𝐺𝑅 is 𝑓′(𝑘) > (𝑛 + 𝑔) and thus 𝜕𝑘 > 0
𝜕𝑐
- For 𝑘 > 𝑘𝐺𝑅 is 𝑓′(𝑘) < (𝑛 + 𝑔) and this 𝜕𝑘 < 0
 - Under ‘k=0 is 𝑐 < 𝑓(𝑘) − (𝑛 + 𝑔)𝑘, k increases (income used for savings is larger than
break-even investments), so ′𝑘 > 0, arrows towards the right, over the curve the opposite
holds, arrow towards the left ′𝑘 < 0.

4.8.3 Modified golden rule steady state

In point E is ′𝑐 = ′𝑘 = 0. This point is left of the golden rule steady state, that is consumption is
below max.

Intuition:

- ′𝑐 = 0 where 𝑓′(𝑘 ∗) = 𝜌 + 𝜃𝑔
- We also know that 𝑓 ′ (𝑘 ∗𝐺𝑅 ) = (𝑛 + 𝑔)
- From the assumption to ensure convergence
𝛽 = 𝜌 − 𝑛 − (1 − 𝜃)𝑔 > 0
we have that

𝜌 + 𝜃𝑔 > 𝑛 + 𝑔
- Hence, 𝑓 ′ (𝑘 ∗) > 𝑓 ′ (𝑘 ∗𝐺𝑅 )
- As 𝑓 ′′ (𝑘) < 0 (decreasing marginal productivity of capital), we have 𝑘 ∗< 𝑘 ∗𝐺𝑅
- The reason is discounting 𝜌: households value current consumption higher than future
consumption
- We also call k* the modified golden rule steady state, that is when we take discounting into
account

The initial value of c

K(0) is given, we assume 𝑘(0) < 𝑘 ∗

Point F is the only stable equilibrium

1. If we start above F, then k(0) will tend towards 0 which we assume is not possible
 2. If we start below F, then k(t) will tend towards something large and consumption towards 0
which is not possible

4.9 The effect of a fall in the discount rate

Suppose the discount rate falls, 𝜌 falls, that is, household become more patient

Assume

- The change is 𝜌 is unexpected

- 𝜌 enters the ′𝑐 = 0 locus but not the one ′𝑘 = 0
- ′𝑐 = 0 where 𝑓 ′ (𝑘 ∗) = 𝜌 + 𝜃𝑔, so a lower 𝜌 corresponds to a higher k* (as 𝑓′′(𝑘) < 0)
- We obtain that ′𝑐 = 0 shifts towards right
- For a Cobb Douglas production function we have
1
𝛼 1−𝛼
𝑘 ∗= ( )
𝜌 + 𝜃𝜌
- The capital stock cannot jump, but consumption can, c falls, more is saved
- And the economy is now on a new saddle path corresponding to the new k*

4.10 Sum up
- We have considered the Ramsey growth model where consumers maximize utility and firms max
profits

- We derive two dynamic equations, one for consumption and one for capital

- We can represent the model in a phase diagram with a ′𝑐 = 0 locus and a ′𝑘 = 0 locus

- The equilibrium is saddle point stable

- A lower discount rate increases savings and leads to more consumption in the long run

5. Diamond model
Terms:

𝑌𝑜𝑢𝑛𝑔 = 𝐴𝑔𝑒𝑛𝑡𝑠 𝑖𝑛 𝑡ℎ𝑒 𝑓𝑖𝑟𝑠𝑡 𝑝𝑒𝑟𝑖𝑜𝑑 𝑜𝑓 𝑡ℎ𝑒𝑖𝑟 𝑙𝑖𝑣𝑒𝑠

𝑂𝑙𝑑 = 𝐴𝑔𝑒𝑛𝑡𝑠𝑖𝑛𝑡ℎ𝑒𝑠𝑒𝑐𝑜𝑛𝑑(𝑙𝑎𝑠𝑡)𝑝𝑒𝑟𝑖𝑜𝑑𝑜𝑓𝑡ℎ𝑒𝑖𝑟𝑙𝑖𝑣𝑒𝑠
𝐿𝑡 = 𝑇ℎ𝑒 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑖𝑛𝑑𝑖𝑣𝑖𝑑𝑢𝑎𝑙𝑠 𝑏𝑜𝑟𝑛 𝑒𝑣𝑒𝑟𝑦 𝑝𝑒𝑟𝑖𝑜𝑑
𝑛 = 𝑇ℎ𝑒 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑔𝑟𝑜𝑤𝑡ℎ 𝑟𝑎𝑡𝑒. 𝑆𝑜, 𝐿𝑡+1 = (1 + 𝑛)𝐿𝑡
𝐶1𝑡 = 𝑃𝑒𝑟𝑖𝑜𝑑 𝑡 𝑐𝑜𝑛𝑠𝑢𝑚𝑝𝑡𝑖𝑜𝑛 𝑜𝑓 𝑦𝑜𝑢𝑛𝑔
𝐶2𝑡 = 𝑃𝑒𝑟𝑖𝑜𝑑 𝑡 𝑐𝑜𝑛𝑠𝑢𝑚𝑝𝑡𝑖𝑜𝑛 𝑜𝑓 𝑜𝑙𝑑
𝜌 = 𝑡ℎ𝑒 𝑑𝑖𝑠𝑜𝑐𝑢𝑛𝑡 𝑟𝑎𝑡𝑒
 𝑟(𝑡) = 𝑓′(𝑘𝑡 ) = 𝑡ℎ𝑒 𝑟𝑒𝑛𝑡𝑎𝑙 𝑟𝑎𝑡𝑒, 𝑚𝑎𝑟𝑔𝑖𝑛𝑎𝑙 𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑣𝑖𝑡𝑦 𝑜𝑓 𝑐𝑎𝑝𝑖𝑡𝑎𝑙
𝑤(𝑡) = 𝑓(𝑘𝑡 ) − 𝑘𝑡 𝑓′(𝑘𝑡 ) = 𝑤𝑎𝑔𝑒 𝑝𝑒𝑟 𝑒𝑓𝑓𝑒𝑐𝑡𝑖𝑣𝑒 𝑢𝑛𝑖𝑡 𝑜𝑓 𝑙𝑎𝑏𝑜𝑟
𝐴𝑡 = 𝑘𝑛𝑜𝑤𝑙𝑒𝑑𝑔𝑒 𝑜𝑟 𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑣𝑖𝑡𝑦 ′
𝑔 = 𝑡ℎ𝑒 𝑔𝑟𝑜𝑤𝑡ℎ 𝑟𝑎𝑡𝑒 𝑜𝑓 𝐴𝑡 𝑠𝑜 𝑡ℎ𝑎𝑡 𝐴𝑡+1 = (1 + 𝑔)𝐴𝑡
𝑠(𝑟) = 𝑡ℎ𝑒 𝑠𝑎𝑣𝑖𝑛𝑔𝑠 𝑎𝑠 𝑎 𝑓𝑢𝑛𝑐𝑡𝑖𝑜𝑛 𝑜𝑓 𝑡ℎ𝑒 𝑟𝑒𝑛𝑡𝑎𝑙 𝑟𝑎𝑡𝑒

The model’s assumptions:

- Overlapping-generations model: in each period, new individuals are born and old individuals
are dying
- Time is discrete: t=0,1,2
- Each agent lives for 2 periods
- Lt are individuals born in every period t
- Population grows at rate n→Lt+1=(1+n)Lt
- Since agents live for only two periods, at time t there are Lt individuals in the first period of
𝐿𝑡
their lives and 𝐿𝑡−1 = (1+𝑛) individuals in the second period of their lives
- Each individual supplies 1 unit of labor when young and divides the resulting labor income
between first period consumption and saving.

5.1 Firms’ assumptions

- Perfect competition and free entry
- No depreciation, 𝛿 = 0
- Production function fulfills the same assumptions as for the Solow model and the Ramsey
model
- Initial capital, 𝐾0 > 0, is owned equally by all old individuals
- Firms are profit maximizing

5.2 Households’ assumptions

- Constant relative risk aversion (CRRA) utility
- In the second period, the individual consumes any savings and any capital rental income
- 𝐶1𝑡 is period t consumption of the young
- 𝐶2𝑡 is period t consumption of the old
- Utility for a person born in period t is then
1−𝜃 𝑡−𝜃
𝐶1𝑡 1 𝐶2𝑡+1
𝑈𝑡 = + , 𝜃 > 0, 𝜌 > −1
1−𝜃 1+𝜌1−𝜃
, where the discount rate 𝜌 > −1, ensures that second period consumption is positive

If 𝜌 > 0, individuals place greater weight on the first period utility

If 𝜌 < 0, individuals place greater weight on the second period utility

- When young individuals supply one unit of labour, earn some labour income, which is split
between first-period consumption and savings
 - When old (in the second period) individuals simply consume their savings and any rental
income that accrued.
- So, we should have that
𝑆𝑎𝑣𝑖𝑛𝑔𝑠 = 𝑤𝑡 𝐴𝑡 − 𝐶1,𝑡
And as capital earns interest payments:
𝐶2,𝑡+1 = (1 + 𝑟𝑡+1 )(𝑤𝑡 𝐴𝑡 − 𝐶1,𝑡 )
Re-arranging yields:
1
𝐶1𝑡 + 𝐶 = 𝐴𝑡 𝑤𝑡
1 + 𝑟𝑡+1 2,𝑡+1

5.2.1 Household maximization problem

1−𝜃 1−𝜃
𝐶1𝑡 1 𝐶2𝑡+1
𝑚𝑎𝑥 𝑈𝑡 = + , 𝜃 > 0, 𝜌 > −1
1 −𝜃 1+𝜌1−𝜃
It is a subject to the budget restriction:
1
𝐶1𝑡 + 𝐶 = 𝐴𝑡 𝑤𝑡
1 + 𝑟𝑡+1 2,𝑡+1

Giving the lagrangian:

1−𝜃 1−𝜃
𝐶1𝑡 1 𝐶2𝑡+1 1
𝐿= + + 𝜆[𝐴𝑡 𝑤𝑡 − (𝐶1𝑡 + 𝐶 )]
1−𝜃 1+𝜌1−𝜃 1 + 𝑟𝑡+1 2,𝑡+1

5.2.2 Solving the household’s problem

The first order conditions are:
𝑑𝐿 −𝜃
= 0 ↔ 𝐶1𝑡 −𝜆=0
𝑑𝐶1𝑡
𝑑𝐿 −𝜃
1
= 0 ↔ 𝐶2𝑡+1 −𝜆 =0
𝑑𝐶2𝑡+1 1 + 𝑟𝑡+1
𝑑𝐿 1
= 0 ↔ 𝑤𝑡 𝐴𝑡 = 𝐶1𝑡 + 𝐶
𝑑𝜆 1 + 𝑟𝑡+1 2𝑡+1
Use equation 1 in 2 to obtain the Euler Equation
1
1 + 𝑟𝑡+1 −𝜃 −𝜃
𝐶2𝑡+1 1 + 𝑟𝑡+1 𝜃
𝐶 = 𝐶1𝑡 ↔ =
1 + 𝜌 2𝑡+1 𝐶1𝑡 1+𝜌

- The larger the interest rate relative to discounting, the larger return to savings relative to
how impatient the household is
- Hence, the greater savings and thereby greater consumption in period 2 relative to in period
1.

5.2.3 The Euler Equation and the savings rate

The third FOC we have a budget constraint:
1
𝑤𝑡 𝐴𝑡 = 𝐶1𝑡 + 𝐶 ↔ (1 + 𝑟𝑡+1 )(𝑊𝑡 𝐴𝑡 − 𝐶1𝑡 ) = 𝐶2𝑡+1
1 + 𝑟𝑡+1 2𝑡+1
The Euler Equation is:
1
1 + 𝑟𝑡+1 𝜃
𝐶2𝑡+1 = 𝐶1𝑡
1+𝜌
We then insert the budget constraint and calculate loads of times and get:
1⁄
(1 + 𝑝) 𝜃
𝐶1𝑡 = (1−𝜃)⁄
𝑤𝑡 𝐴𝑡
1
(1 + 𝜌) ⁄𝜃 + (1 + 𝑟𝑡+1 ) 𝜃

- This is the fraction of income consumed in the first period

- The rental rate this determines the fraction of income that the individual consumes in the
first period
- Define s(r) as the savings rate
- Thus, 𝐶1𝑡 = [1 − 𝑠(𝑟𝑡+1 )]𝐴𝑡 𝑤𝑡

5.2.4 The savings rate

1⁄
𝑤𝑡 𝐴𝑡 − 𝐶1𝑡 (1 + 𝜌) 𝜃
𝑠(𝑟𝑡+1 ) = =1− (1−𝜃)⁄
𝑤𝑡 𝐴𝑡 1⁄
(1 + 𝜌) 𝜃 + (1 + 𝑟𝑡+1 ) 𝜃

1⁄
(1 + 𝑟𝑡+1 ) 𝜃
𝑠(𝑟𝑡+1 ) = (1−𝜃)⁄
1⁄
(1 + 𝜌) 𝜃 + (1 + 𝑟𝑡+1 ) 𝜃

- S is increasing in r if 𝜃 < 1
- A rise in r has an income and a substitution effect
- If r increases, consumption in the second period becomes cheaper, savings increase
(substitution effect)
- If r increases, a given amount of savings yields more second period consumption and thus
tends to decrease savings (income effect)

The impact on the savings rate from a higher interest rate

(1−𝜃)⁄
𝜕(1 + 𝑟) 𝜃 (1 − 𝜃) (1−2𝜃)⁄
= (1 + 𝑟) 𝜃
𝜕𝑟 𝜃
- Sign depends on whether 𝜃 ><= 1
- When agents are very willing to substitute consumption between two periods to take
advantage of rate-of-return incentives (𝑤ℎ𝑒𝑛 𝜃 <
(1−𝜃)⁄
𝜕(1+𝑟) 𝜃
1, 𝑡ℎ𝑒 𝑠𝑢𝑏𝑠𝑡𝑖𝑡𝑢𝑡𝑖𝑜𝑛 𝑒𝑓𝑓𝑒𝑐𝑡 𝑑𝑜𝑚𝑖𝑛𝑎𝑡𝑒𝑠, 𝜕𝑟
> 0)
- When individuals have a strong preference for similar levels of consumption in the two
(1−𝜃)⁄
𝜕(1+𝑟) 𝜃
periods (𝑤ℎ𝑒𝑛 𝜃 > 1), 𝑡ℎ𝑒 𝑖𝑛𝑐𝑜𝑚𝑒 𝑒𝑓𝑓𝑒𝑐𝑡 𝑑𝑜𝑚𝑖𝑛𝑎𝑡𝑒𝑠: 𝜕𝑟
< 0.
- In the special case when 𝜃 =
0 (𝑙𝑜𝑔𝑎𝑟𝑖𝑡ℎ𝑚𝑖𝑐 𝑢𝑡𝑖𝑙𝑖𝑡𝑦), 𝑡ℎ𝑒 𝑡𝑤𝑜 𝑒𝑓𝑓𝑒𝑐𝑡𝑠 𝑏𝑎𝑙𝑎𝑛𝑐𝑒 𝑎𝑛𝑑 𝑡ℎ𝑒 𝑠𝑎𝑣𝑖𝑛𝑔𝑠 𝑟𝑎𝑡𝑒 𝑖𝑠 𝑖𝑛𝑑𝑒𝑝𝑒𝑛𝑑𝑒𝑛𝑡 𝑜𝑓 𝑟,
(1−𝜃)⁄
𝜕(1+𝑟) 𝜃

𝜕𝑟
= 0.
5.3 The dynamics of Capital
- Capital in period t+1 is equal to the portion of labour income saved by the young in time t:
𝐾𝑡+1 = 𝑠(𝑟𝑡+1 )𝐿𝑡 𝐴𝑡 𝑤𝑡
𝐾𝑡+1
- Since 𝑘𝑡+1 = 𝐴 , we can obtain in effective units of labour:
𝑡+1 𝐿𝑡+1

𝑠(𝑟𝑡+1 )𝐿𝑡 𝐴𝑡 𝑤𝑡 𝑠(𝑟𝑡+1 )

𝑘𝑡+1 = = 𝑤
𝐴𝑡 (1 + 𝑔)𝐿𝑡 (1 + 𝑛) (1 + 𝑔)(1 + 𝑛) 𝑡
- Using that profit max implies that input factors are paid their marginal product
𝑟𝑡 = 𝑓′(𝑘𝑡 ), 𝑤𝑡 = 𝑓(𝑘𝑡 ) − 𝑘𝑡 𝑓′(𝑘𝑡 ), 𝑤𝑒 ℎ𝑎𝑣𝑒
1
𝑘𝑡+1 = 𝑠(𝑓′(𝑘𝑡+1 ))(𝑓(𝑘𝑡 ) − 𝑘𝑡 𝑓′(𝑘𝑡 ))
(1 + 𝑔)(1 + 𝑛)

5.3.1 The evolution of capital with Cobb-Douglas production and log utility
We have both old and young people; we will get a solution where we can move something from one
generation to another.

We have capital in period t on the first axis and period t+1 on the second axis. We have a stable
solution. If we are above the steady state, for example if we had 𝑘0 . Then we have that capital in
period t+1 as a function of capital t is below the 45-degree line. K1 will be lower than K0. (This line
illustrates where capital is equal in both periods.) K* is stable, in other words, the balanced growth
path.

- Is there a 𝑘 ∗, such that 𝑘𝑡+1 = 𝑘𝑡 satisfies

1
𝑘𝑡+1 = 𝑠(𝑓 ′ (𝑘𝑡+1 ))[𝑓(𝑘𝑡 ) − 𝑘𝑡 𝑓 ′ (𝑘𝑡 )]
(1 + 𝑛)(1 + 𝑔)
 - It depends on the functional forms
- If we assume a Cobb-Douglas and logarithmic utility, (𝜃 = 1 𝑠𝑜 𝑡ℎ𝑎𝑡 𝑈(𝑐) = 𝑙𝑛𝑐) then:
→the production is Cobb-Douglas, 𝑓(𝑘) = 𝑘 𝛼 and 𝑓′(𝑘) = 𝛼𝑘 𝛼−1
→the fraction of income saved is
(1−𝜃)⁄
(1 + 𝑟𝑡+1 ) 𝜃 1 1
𝑠(𝑟𝑡+1 ) = (1−𝜃)⁄
= =
1⁄ (1 + 𝜌) + 1 2 + 𝜌
(1 + 𝜌) 𝜃 + (1 + 𝑟𝑡+1 ) 𝜃

Same equation as previously:

1
𝑘𝑡+1 = 𝑠(𝑓 ′ (𝑘𝑡+1 ))[𝑓(𝑘𝑡 ) − 𝑘𝑡 𝑓 ′ (𝑘𝑡 )]
(1 + 𝑛)(1 + 𝑔)

↔
1 1
𝑘𝑡+1 = (1 − 𝛼)𝑘𝑡𝛼
(1 + 𝑔)(1 + 𝑛) 2 + 𝜌
Strictly concave curve:
𝑑𝑘𝑡+1 1 1
- 𝑑𝑘𝑡
= (1+𝑔)(1+𝑛) 2+𝜌 (1 − 𝛼)𝛼𝑘𝑡𝛼−1 > 0
𝑑 2 𝑘𝑡+1 1 1
- 𝑑𝑘𝑡 2
= − (1+𝑔)(1+𝑛) 2+𝜌 (1 − 𝛼)2 𝛼𝑘𝑡𝛼−2 < 0

5.4 The balanced growth path

The balanced growth is where 𝑘𝑡+1 = 𝑘𝑡 = 𝑘 ∗, and hence:
1 1
𝑘 ∗= (1 − 𝛼)(𝑘 ∗)𝛼
(1 + 𝑔)(1 + 𝑛) 2 + 𝜌
↔
(1 − 𝛼) 1 1⁄
𝑘 ∗= ( ) 1−𝛼
(1 + 𝑔)(1 + 𝑛) 2 + 𝜌
- Graphically it is where the 𝑘𝑡+1 curve crosses the 45 degree line
- K* is globally stable
- If 𝑘0 > 𝑘 ∗, then 𝑘𝑡+1 < 𝑘𝑡 and so the economy converges towards k*
- If 𝑘0 < 𝑘 ∗, then 𝑘𝑡+1 > 𝑘𝑡 again, the economy converges towards k*
- Once the economy has converged to a balances growth path, its properties are similar to the
Solow Model: the savings rate is constant, output per worker is growing at the rate of
productivity, etc.

The effect of a fall in the discount rate:

- When the discount rate changes → take the first order condition of k* wrt p.
- The answer will be lower than 0
- Hence, if the discount rate falls, k* increases.
- Graphically, the 𝑘𝑡+1 curve shifts upwards when ∆𝜌 < 0.
 -

5.5 The possibility of dynamic inefficiency

- G=0, no technological progress
- We have heterogeneity since we have two group of people: different people (young and old)
- We can make someone better of without making any worse of.
- Reason for it has to be higher than the Golden rule level of k is because then we can ask
people to consume more now and save less, and then they are happier because they can
consume more and utility is increasing in consumption. And then we can also make the older
people happier
- The balanced growth level of k is higher than the Golden rule level of k:

𝑘 ∗> 𝑘 𝐺𝑅 ↔ 𝑓′(𝑘 ∗) < 𝑓′(𝑘 𝐺𝑅 )

The golden rule level of k maximizes consumption, that is

𝑚𝑎𝑥 𝑐 ∗= 𝑓(𝑘 ∗) − 𝑛𝑘 ∗
this gives:
𝑑𝑐 ∗
= 0 ↔ 𝑓′(𝑘 𝐺𝑅 ) − 𝑛 = 0 ↔ 𝑓′(𝑘 𝐺𝑅 ) = 𝑛
𝑑𝑘 ∗
Notice, this is the condition we also found when we consider the Ramsey model, where there we
had g different from 0.

Dynamic inefficiency for the Cobb-Douglas

- Log utility and CD production function 𝑓(𝑘) = 𝑘 𝛼

1 1
𝑘𝑡+1 = (1 − 𝛼)𝑘𝑡𝛼
(1 + 𝑛) 2 + 𝜌
- The balances growth level of k:
(1 − 𝛼) 1 1⁄
𝑘 ∗= ( ) 1−𝛼
(1 + 𝑛) 2 + 𝜌
Hence, the marginal productivity of capital at k* is then 𝑓′(𝑘 ∗) = 𝛼(𝑘 ∗)𝛼−1
 (1 − 𝛼) 1 1⁄ 𝛼
𝑓′(𝑘 ∗) = 𝛼(( ) 1−𝛼 )𝛼−1 = (1 + 𝑛)(2 + 𝜌)
(1 + 𝑛) 2 + 𝜌 1−𝛼
Hence k*>kGR

(We times the potenser with each other and get -1 and hence turn 1 over the equation and turn the
brøks around)

k*>kGR if:

𝑓 ′ (𝑘 ∗) < 𝑓 ′ (𝑘 𝐺𝑅 )
𝛼
(1 + 𝑛)(2 + 𝜌) < 𝑛
1−𝛼
Hence, this will hold for a high population growth level, that is, if n is high. (relative to the discount
rate, 𝜌 and alpha 𝛼).

5.5.1 Dynamic inefficiency intuition

- Dynamic inefficiency arises from overaccumulation
- Results from that current young generation needs to save for old age
- The more they save, the lower is the rate of return and may encourage to save even more
- Effect on future rate of return to capital is an externality on next generation
- If alternative ways of providing consumption to individuals in old age were introduced,
overaccumulation could be avoided

Social security

- A way of dealing with overaccumulation

- Fully-funded system: young make contributions to the social security system and their
contributions are paid back to them in their old age
- Unfunded system or a pay-as-you-go: transfers from the young directly go to the current old.
Pay-as-you-go (unfunded) Social security discourages aggregate savings. With dynamic
inefficiency, discouraging savings may lead to a Pareto improvement.
- Requires a relative high population growth rate. As population growth has recently
decreased in many countries, more countries go for Fully-funded systems

6. Endogenous Growth

𝑌 − 𝑜𝑢𝑡𝑝𝑢𝑡
𝐾 − 𝐶𝑎𝑝𝑖𝑡𝑎𝑙
𝐿 − 𝐿𝑎𝑏𝑜𝑢𝑟
𝐴 − 𝑇𝑒𝑐ℎ𝑛𝑜𝑙𝑜𝑔𝑦 𝑜𝑟 𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑣𝑖𝑡𝑦
𝑎𝐿 − 𝑡ℎ𝑒 𝑓𝑎𝑐𝑡𝑖𝑜𝑛 𝑜𝑓 𝑙𝑎𝑏𝑜𝑢𝑟 𝑒𝑚𝑝𝑙𝑜𝑦𝑒𝑑 𝑖𝑛 𝑡ℎ𝑒 𝑟𝑒𝑠𝑒𝑎𝑟𝑐ℎ 𝑎𝑛𝑑 𝑑𝑒𝑣𝑒𝑙𝑜𝑝𝑒𝑛𝑡, 𝑅&𝐷 𝑠𝑒𝑐𝑡𝑜𝑟
(1 − 𝑎𝐿 ) − 𝑡ℎ𝑒 𝑓𝑟𝑎𝑐𝑡𝑖𝑜𝑛 𝑜𝑓 𝑙𝑎𝑏𝑜𝑢𝑟 𝑒𝑚𝑝𝑙𝑜𝑦𝑒𝑑 𝑖𝑛 𝑡ℎ𝑒 𝑔𝑜𝑜𝑑𝑠 𝑠𝑒𝑐𝑡𝑜𝑟

𝑎𝐾 − 𝑡ℎ𝑒 𝑓𝑟𝑎𝑐𝑡𝑖𝑜𝑛 𝑜𝑓 𝑐𝑎𝑝𝑖𝑡𝑎𝑙 𝑒𝑚𝑝𝑙𝑜𝑦𝑒𝑑 𝑖𝑛 𝑡ℎ𝑒 𝑅&𝐷 𝑠𝑒𝑐𝑡𝑜𝑟

 (1 − 𝑎𝐾 ) − 𝑡ℎ𝑒 𝑓𝑟𝑎𝑐𝑡𝑖𝑜𝑛 𝑜𝑓 𝑐𝑎𝑝𝑖𝑡𝑎𝑙 𝑒𝑚𝑝𝑙𝑜𝑦𝑒𝑑 𝑖𝑛 𝑡ℎ𝑒 𝑔𝑜𝑜𝑑𝑠 𝑠𝑒𝑐𝑡𝑜𝑟

𝑔𝐴 (𝑡) − 𝑡ℎ𝑒 𝑔𝑟𝑜𝑤𝑡ℎ 𝑟𝑎𝑡𝑒 𝑖𝑛 𝑡ℎ𝑒 𝑠𝑡𝑜𝑐𝑘 𝑜𝑓 𝑘𝑛𝑜𝑤𝑙𝑒𝑑𝑔𝑒

′𝑔𝐴 (𝑡) − 𝑡ℎ𝑒 𝑔𝑟𝑜𝑤𝑡ℎ 𝑟𝑎𝑡𝑒 𝑜𝑓 𝑡ℎ𝑒 𝑔𝑟𝑜𝑤𝑡ℎ 𝑟𝑎𝑡𝑒 𝑜𝑓 𝑡ℎ𝑒 𝑠𝑡𝑜𝑐𝑘 𝑜𝑓 𝑘𝑛𝑜𝑤𝑙𝑒𝑑𝑔𝑒 𝑤𝑟𝑡 𝑡𝑖𝑚𝑒

- Previous growth models: as capital converges to a balanced growth path, differences in

productivity (A) must account for cross-country income differences – and A is exogenous
- Here, we are going to make productivity, A, endogenous
- Goal of model: explain to create dynamics around this g and to create an incentive for
people to invest into the advancement of technology and knowledge.
- Specifically, we’ll introduce a separate research and development, RD sector
- There is a Goods producing sector and an RD sector
- The model is continuous time
- Variables, L,K, A and Y

6.1 The research and development sector

- Knowledge can be produced, hence it is endogenous
- RD production is Cobb-Douglas, but the sum of the exponents is not restricted to 1, that is
we can have decreasing or increasing returns to scale
- The fraction of labour and capital used in the RD sector is exogenous and constant

6.2 Goods production

- We assume we have a CRS (constant returns to scale) production function
- Fraction 𝑎𝐿 (constant, exogenous) of the labour force is used in the RD sector and 1 − 𝑎𝐿 is
used in the goods producing sector. Labour devoted to RD.
- Fraction 𝑎𝐾 (constant, exogenous) of the capital stock is used in the RD sector and 1 − 𝑎𝐾 is
used in the goods producing sector
- Assume that the use of a piece of knowledge in one sector does not prevent it from being
used in another section
- Hence, both sectors use full stock of knowledge A
- CRS (constant returns to scale) production (in K and L)

𝑌(𝑡) = [(1 − 𝑎𝐾 )𝐾(𝑡)]𝛼 [𝐴(𝑡)(1 − 𝑎𝐿 )𝐿(𝑡)]1−𝛼 , 0 < 𝛼 < 1

6.3 Knowledge production

- Knowledge production is not assumed to be CRS:

′𝐴(𝑡) = 𝐵[𝑎𝐾 𝐾(𝑡)]𝛽 [𝑎𝐿 𝐿(𝑡)]𝛾 𝐴(𝑡)𝜃 , 𝐵 > 0, 𝛽 ≥ 0, 𝛾 ≥ 0

- B is a shift parameter
- [𝑎𝐿 𝐿(𝑡)]𝛾 what the people working in research ad to the stock of knowledge
- 𝐴(𝑡)𝜃 – the more knowledge we have, the more new knowledge, we can produce
- Knowledge production: diminishing returns in RD (DRS) 𝛽 + 𝛾 < 1. A doubling of RD inputs
may not lead to a doubling of knowledge.
 - Knowledge production: increasing returns to scale (IRS) 𝛽 + 𝛾 > 1. There may be fixed costs
to RD or increased interactions between researchers may lead to extra discoveries
- 𝜃 reflects the effect of the existing stock of knowledge on RD
- 𝜃 > 0: past discoveries may provide better tools and ideas that make future discoveries
easier (e.g.: the impact of the microscope)
- 𝜃 < 0: the easiest discoveries are made first, so harder to make new discoveries when the
stock of knowledge is greater (e.g.: leaps in smartphone technology are likely to be “smaller”
going forward as all of the “easiest” innovations have been discovered)

6.4 Solow growth model with endogenous K and A

- Savings is exogenous and capital depreciation is set to zero (𝛿 = 0) → ′𝐾(𝑡) = 𝑠𝑌(𝑡)
- Population growth is exogenous ′𝐿(𝑡) = 𝑛𝐿(𝑡), 𝑛 ≥ 0
′𝐿(𝑡)
- 𝑛= 𝐿(𝑡)
- We have a Solow growth model with two endogenous variables, capital K and knowledge A

Endogenous model without capital:

Let there be no capital, hence we have only one endogenous variable:

′𝐴(𝑡) = 𝐵[𝑎𝐿 𝐿(𝑡)]𝛾 𝐴(𝑡)𝜃

Dividing both sides by A(t)

′𝐴(𝑡) 𝐵[𝑎𝐿 𝐿(𝑡)]𝛾 𝐴(𝑡)𝜃

𝑔𝐴 (𝑡) = = = 𝐵[𝑎𝐿 𝐿(𝑡)]𝛾 𝐴(𝑡)𝜃−1
𝐴(𝑡) 𝐴(𝑡)
Taking logs and differentiation both sides with respect to time yield:

𝑙𝑛𝑔𝐴 (𝑡) = 𝑙𝑛𝐵 + 𝛾(𝑙𝑛𝑎𝐿 + 𝑙𝑛𝐿(𝑡)) + (𝜃 − 1)𝑙𝑛𝐴(𝑡)

′𝑔𝐴 (𝑡) ′𝐿(𝑡) ′𝐴(𝑡)

=𝛾 + (𝜃 − 1)
𝑔𝐴 (𝑡) 𝐿(𝑡) 𝐴(𝑡)
′𝐿(𝑡) ′𝐴(𝑡)
Noting that 𝐿(𝑡)
= 𝑛 and 𝐴(𝑡)
= 𝑔𝐴 (𝑡), we have the growth rate in the growth rate A:

′𝑔𝐴 (𝑡) = 𝛾𝑛𝑔𝐴 (𝑡) + (𝜃 − 1)[𝑔𝐴 (𝑡)]2

Chinese growth is about 5%, but it has been much higher. So 𝑔𝐴 is 5% and ′𝑔𝐴 is the change in the
growth rate. E.g.: recently the g a dot has been negative.

In order to see the impact of an increase in fraction of workers, we need to differentiate 𝑔𝐴 (𝑡) wrt
𝑎𝐿 . See mock exam 2019.

6.5 The relationship for the growth rate

The growth rate in the growth rate of A is

′𝑔𝐴 (𝑡) = 𝛾𝑛𝑔𝐴 (𝑡) + (𝜃 − 1)[𝑔𝐴 (𝑡)]2

To draw the curve as a function of 𝑔𝐴 :
(for calculation see mock exam 2019)
𝑑′𝑔𝐴 (𝑡)
= 𝛾𝑛 + 2(𝜃 − 1)[𝑔𝐴 (𝑡)]
𝑑𝑔𝐴
𝑑2 ′𝑔𝐴 (𝑡)
= 2(𝜃 − 1)
𝑑𝑔𝐴 2
Hence, the slope of the curve depends on the value of 𝜃, that is, the effect of the existing stock of
knowledge on RD

To find the point where growth rate in knowledge is stable, we have to find where 𝑔𝐴 (𝑡) = 𝑔 ∗𝐴 (𝑡),
giving
′𝑔𝐴 (𝑡)
=0
𝑔𝐴 (𝑡)
′𝑔𝐴 (𝑡)
= 𝛾𝑛 + (𝜃 − 1)[𝑔𝐴 (𝑡)] = 0
𝑔𝐴
𝛾𝑛 + (𝜃 − 1)[𝑔 ∗𝐴 (𝑡)] = 0
𝛾𝑛 𝛾𝑛
𝑔 ∗𝐴 (𝑡) = − =
(𝜃 − 1) (1 − 𝜃)

6.5.1 𝜽 < 𝟏: Diminishing returns, as A increases

Theta – the slope of the curve between the change in the growth rate in knowledge and the change
in the change rate in the growth rate of knowledge.

Whether theta is smaller than 1, equal to 1 of larger than 1, is going to determine how the growth
rate of the growth rate of knowledge (ga) is depending of the growth rate of knowledge.

Growth rate in Denmark is about 1%, China about 5%. When ‘gA is negative, the growth is
decreasing.

The growth rate in the growth rate of A is

′𝑔𝐴 (𝑡) = 𝛾𝑛𝑔𝐴 (𝑡) + (𝜃 − 1)[𝑔𝐴 (𝑡)]2

When 𝜃 < 1 then the growth rate of 𝑔𝐴 is positive at first
𝑑′𝑔𝐴 (𝑡)
= 𝛾𝑛 + 2(𝜃 − 1)[𝑔𝐴 (𝑡)]
𝑑𝑔𝐴
↔
𝛾𝑛
> [𝑔𝐴 (𝑡)]
2(1 − 𝜃)
Hence, until 𝑔𝐴 given by
𝛾𝑛
= 𝑔𝐴
2(1 − 𝜃)
After this, it is negative.

The growth rate in the growth rate is then positive until

′𝑔𝐴 (𝑡)
= 0 ↔ 𝛾𝑛 + (𝜃 − 1)𝑔 ∗𝐴 (𝑡) = 0
𝑔𝐴 (𝑡)
↔
𝛾𝑛
𝛾𝑛 = (1 − 𝜃)𝑔 ∗𝐴 (𝑡) ↔ = 𝑔 ∗𝐴 (𝑡)
1−𝜃

Hence, the slope of this curve changes from positive to negative when the growth rate in knowledge
𝛾𝑛
is equal to (1−𝜃) = 𝑔 ∗𝐴 . This is at the top of the curve.

6.5.2 𝜽 < 𝟏: Impact of a higher 𝒂𝑳

- More resources to the RD sector does not lead to long term growth
- Suppose 𝑎𝐿 increases, that is, a higher fraction of the labour force works in the RD sector
- Suppose that 𝑔𝐴 = 𝑔 ∗𝐴 and there is permanent increase in 𝑎𝐿 (the fraction of labour
employed by the RD sector). We notice that there is no change in 𝑔 ∗𝐴 (𝑎𝐿 is not present in
this equation):
- Notice that there are no change in 𝑔 ∗𝐴 (𝑎𝐿 is not present)
𝛾𝑛
= 𝑔 ∗𝐴
1−𝜃
𝛾
- Since 𝑔𝐴 (𝑡) = 𝐵𝑎𝐿 𝐿(𝑡)𝛾 𝐴(𝑡)𝜃−1, we know that 𝑔𝐴 is going to instantaneously increase with
an increase in 𝑎𝐿 .
- But ′𝑔𝐴 (𝑡) = 𝛾𝑛𝑔𝐴 (𝑡) + (𝜃 − 1)[𝑔𝐴 (𝑡)]2 , so 𝑎𝐿 does not enter this expression and hence
the growth rate of A is going to eventually fall back to 𝑔 ∗𝐴 as there are diminishing returns
A increases.
6.5.3 𝜽 > 𝟏: Increasing returns as A increases
The growth rate in the rate of A is

′𝑔𝐴 (𝑡) = 𝛾𝑛𝑔𝐴 (𝑡) + (𝜃 − 1)[𝑔𝐴 (𝑡)]2

When 𝜃 > 1 then the growth rate of 𝑔𝐴 is positive
𝑑′𝑔𝐴 (𝑡)
= 𝛾𝑛 + 2(𝜃 − 1)[𝑔𝐴 (𝑡)] > 0
𝑑𝑔𝐴
And increasing:

𝑑2 ′𝑔𝐴 (𝑡)
= 2(𝜃 − 1) > 0
𝑑𝑔𝐴 2
Knowledge is so useful in the production of new knowledge that the growth rate of knowledge rises
rather than falls.

Impact of higher ga(t)

6.5.5 𝜃 = 1: Constant returns as A increases
The growth rate in knowledge is:
′𝐴(𝑡)
𝑔𝐴 (𝑡) = = 𝐵[𝑎𝐿 𝐿(𝑡)]𝛾
𝐴(𝑡)
The growth rate in the growth rate of A is when 𝜃 = 1

′𝑔𝐴 (𝑡) = 𝛾𝑛𝑔𝐴 (𝑡)

- When 𝜃 = 1 then the growth rate of 𝑔𝐴 is positive
𝑑′𝑔𝐴 (𝑡)
= 𝛾𝑛 > 0
𝑑𝑔𝐴
- And the slope is constant:
𝑑2 ′𝑔𝐴 (𝑡)
=0
𝑑𝑔𝐴 2
So, the change in knowledge growth over time will be given strictly by the growth in the population
rate, n>0. If there are more people to make discoveries than ?? we obtain growth.

6.6 The General Case, that is including capital

𝛿 = 0 → 𝐾(𝑡) = 𝑠𝑌(𝑡)
- We now have two endogenous state variables, A and K
- Substituting in for Y(t) and dividing both sides by K(t) to obtain growth rate in capital

′𝐾(𝑡) 𝑠[(1 − 𝑎𝐾 )𝐾(𝑡)]𝛼 [𝐴(𝑡)(1 − 𝑎𝐿 )𝐿(𝑡)]1−𝛼

𝑔𝐾 (𝑡) = =
𝐾(𝑡) 𝐾(𝑡)
′𝐾(𝑡) 𝐴(𝑡)𝐿(𝑡) 1−𝛼
𝑔𝐾 (𝑡) = = 𝑠(1 − 𝑎𝐾 )𝛼 ((1 − 𝑎𝐿 ) )
𝐾(𝑡) 𝐾(𝑡)
 - Taking logs and differentiating wrt time
𝑙𝑛 𝑔𝐾 (𝑡) =

𝑙𝑛𝑠 + 𝛼𝑙𝑛(1 − 𝑎𝐾 ) + (1 − 𝛼)(𝑙𝑛(1 − 𝑎𝐿 ) + 𝑙𝑛𝐴(𝑡) + 𝑙𝑛𝐿(𝑡) − 𝑙𝑛𝐾(𝑡))

′𝑔𝐾 (𝑡) ′𝐴(𝑡) ′𝐿(𝑡) ′𝐾(𝑡)

= (1 − 𝛼)( + + )
𝑔𝐾 (𝑡) 𝐴(𝑡) 𝐿(𝑡) 𝐾(𝑡)

6.7 The general case – growth rate of capital

Note that 𝑔𝐾 > 0 given the specification of the production function:
′𝐾(𝑡) 𝐴(𝑡)𝐿(𝑡) 1−𝛼
𝑔𝐾 (𝑡) = = 𝑠(1 − 𝑎𝐾 )𝛼 (1 − 𝑎𝐿 )1−𝛼 ( )
𝐾(𝑡) 𝐾(𝑡)
- Every time we divide by a term, we get the rate. In discrete time, we have
𝐾𝑡+1 − 𝐾𝑡
𝑔𝐾 =
𝐾𝑡
- This is the difference between something divided by the initial.

Hence, there is always positive growth in capital. And the growth rate in the growth rate in capital is:
′𝑔𝐾 (𝒕)
= (𝟏 − 𝜶)(𝑔𝐴 (𝑡) + 𝑛 − 𝑔𝐾 (𝑡))
𝑔𝐾 (𝒕)
(In order to get this function, we first, take logs and then differentiate with respect to time.

Where the sign depends upon relative growth rates:

- ′𝑔𝐾 > 0 𝑖𝑓 𝑔𝐴 + 𝑛 − 𝑔𝐾 > 0

- ′𝑔𝐾 < 0 𝑖𝑓 𝑔𝐴 + 𝑛 − 𝑔𝐾 < 0
- ′𝑔𝐾 = 0 𝑖𝑓 𝑔𝐴 + 𝑛 − 𝑔𝐾 = 0
Hence for sustained growth in capital we need a lot of technological progress and population
growth.

6.8 The General Case, growth rate of K

We draw the 𝑔𝐾 = 0 lucus into a diagram with 𝑔𝐾 on the first axis and 𝑔𝐴 on the second axis
′𝑔𝐾 (𝑡)
= 0 ↔ 𝑔𝐾 (𝑡) = 𝑔𝐴 (𝑡) + 𝑛
𝑔𝐾 (𝑡)
Notice that 𝑔𝐾 is a function of the growth rate in technology, which is equal to 𝑔𝐴 (𝑡). We want to
draw this curve into a diagram, with 𝑔𝐴 (𝑡) on the first axis and 𝑔𝐾 (𝑡) and therefore we need to find
a slope and the intercept with the second axis.

- Slope and intercept with second axis

𝑑𝑔𝐾 (𝑡)
| = 1, 𝑔𝐾 (𝑡)|𝑔𝐾(𝑡)=𝑔𝐴(𝑡) = 𝑛
𝑑𝑔𝐴 (𝑡) 𝑔𝑘(𝑡)=0
Notice that the intercept with the second axis is found by letting 𝑔𝐴 (𝑡)=0.
 - Arrows: Consider
′𝑔𝐾 (𝑡)
= (1 − 𝛼)(𝑔𝐴 (𝑡) + 𝑛 − 𝑔𝐾 (𝑡))
𝑔𝐾 (𝑡)
- Above 𝑔𝐾 (𝑡) = 0: 𝑖𝑓 𝑔𝐾 𝑖𝑠 𝑎𝑏𝑜𝑣𝑒 𝑤ℎ𝑒𝑟𝑒 𝑔𝐾 = 0 𝑡ℎ𝑒𝑛 𝑔𝐾 < 0, 𝑡ℎ𝑒 𝑎𝑟𝑟𝑜𝑤 𝑑𝑜𝑤𝑛𝑤𝑎𝑟𝑑𝑠
- Below 𝑔𝐾 (𝑡) = 0: 𝑖𝑓 𝑔𝐾 𝑖𝑠 𝑏𝑒𝑙𝑜𝑤 𝑤ℎ𝑒𝑟𝑒 𝑔𝐾 = 0 𝑡ℎ𝑒𝑛 𝑔𝐾 > 0, 𝑡ℎ𝑒 𝑎𝑟𝑟𝑜𝑤 𝑢𝑝𝑤𝑎𝑟𝑑𝑠

We draw 𝑔𝐾 = 0 into a diagram with 𝑔𝐴 on the first axis and 𝑔𝐾 on the second axis

The intercept is equal to n population growth, so the higher population growth, the more growth we
are going to have in capital. And that is because when there are more workers, there are more
people to produce output, which means there will be more saved. When we have a fixed savings
rate, therefore we will have more capital.

The reason for why we have these arrows is to know if we are in equilibrium, are we approaching
the equilibrium or opposite? (Stable/unstable equilibrium).

6.9 The General Case, Dynamics of A

- With capital we have the dynamics of knowledge

′𝐴(𝑡) = 𝐵[𝑎𝐾 𝐾(𝑡)]𝛽 [𝑎𝐿 𝐿(𝑡)]𝛾 𝐴(𝑡)𝜃 , 𝐵 > 0, 𝛽 ≥ 0, 𝛾 ≥ 0

- We divide by A(t) to obtain the growth rate in knowledge:

′𝐴(𝑡) 𝐵[𝑎𝐾 𝐾(𝑡)]𝛽 [𝑎𝐿 𝐿(𝑡)]𝛾 𝐴(𝑡)𝜃

𝑔𝐴 (𝑡) = =
𝐴(𝑡) 𝐴(𝑡)

𝑔𝐴 (𝑡) = 𝐵[𝑎𝐾 𝐾(𝑡)]𝛽 [𝑎𝐿 𝐿(𝑡)]𝛾 𝐴(𝑡)𝜃−1

- Taking logs and differentiating wrt time
𝑙𝑛𝑔𝐴 (𝑡) = 𝑙𝑛𝐵 + 𝛽(𝑙𝑛(𝑎𝐾 ) + 𝑙𝑛𝐾(𝑡)) + 𝛾(𝑙𝑛𝑎𝐿 − 𝑙𝑛𝐿(𝑡)) + (𝜃 − 1)𝑙𝑛𝐴(𝑡)
 ′𝑔𝐴 (𝑡) ′𝐾(𝑡) ′𝐿(𝑡) 𝐴(𝑡)
=𝛽 +𝛾 + (𝜃 − 1)
𝑔𝐴 (𝑡) 𝐾(𝑡) 𝐿(𝑡) 𝐴(𝑡)
The growth rate in the growth rate of knowledge is then:
′𝑔𝐴 (𝑡)
= 𝛽𝑔𝐾 (𝑡) + 𝛾𝑛 + (𝜃 − 1)𝑔𝐴 (𝑡)
𝑔𝐴 (𝑡)

Note that 𝑔𝐴 > 0 as

𝑔𝐴 (𝑡) = 𝐵[𝑎𝐾 𝐾(𝑡)]𝛽 [𝑎𝐿 𝐿(𝑡)]𝛾 𝐴(𝑡)𝜃−1 > 0

And the growth rate in the growth rate of knowledge is then
′𝑔𝐴 (𝑡)
= 𝛽𝑔𝐾 (𝑡) + 𝛾𝑛 + (𝜃 − 1)𝑔𝐴 (𝑡)
𝑔𝐴 (𝑡)
Where the sign depends upon relative growth rates.

- ′𝑔𝐴 > 0 𝑖𝑓 𝛽𝑔𝐾 (𝑡) + 𝛾𝑛 + (𝜃 − 1)𝑔𝐴 (𝑡) > 0

- ′𝑔𝐴 < 0 𝑖𝑓 𝛽𝑔𝐾 (𝑡) + 𝛾𝑛 + (𝜃 − 1)𝑔𝐴 (𝑡) < 0
- ′𝑔𝐴 = 0 𝑖𝑓 𝛽𝑔𝐾 (𝑡) + 𝛾𝑛 + (𝜃 − 1)𝑔𝐴 (𝑡) = 0

We draw the ′𝑔𝐴 = 0 lucus into a diagram with 𝑔𝐴 on the first axis and 𝑔𝐾 on the second axis:
′𝑔𝐴 (𝑡) 𝛾 (1 − 𝜃)
= 0 ↔ 𝑔𝐾 (𝑡) = − 𝑛 + 𝑔𝐴 (𝑡)
𝑔𝐴 (𝑡) 𝛽 𝛽
So, here we have the relationship between 𝑔𝐾 and 𝑔𝐴 which implies that ′𝑔𝐴 = 0, hence we can
draw this curve into a diagram. Notice that we need 𝜃 < 1 as otherwise the curve will have a
negative slope and then it will never cross the 𝑔𝐾 = 0 curve, which we just derived. Hence there
would be no equilibrium.

- Slope (positive for 𝜃 < 1) and intercept with second axis:

𝑑𝑔𝐾 (𝑡) 1−𝜃 𝛾
= |𝑔𝐴 (𝑡)=0 = , 𝑔𝐾 (𝑡)|′𝑔𝐴 (𝑡)=𝑔𝐴 (𝑡)=0 = − 𝑛
𝑑𝑔𝐴 (𝑡) 𝛽 𝛽
Notice, the intercept is negative if population growth is positive.

Arrows: Consider

′𝑔𝐴 (𝑡)
= 𝛽𝑔𝐾 + 𝛾𝑛 + (𝜃 − 1)𝑔𝐴 (𝑡)
𝑔𝐴 (𝑡)

- Above 𝑔𝐴 (𝑡) = 0: 𝑖𝑓 𝑔𝐴 𝑖𝑠 𝑎𝑏𝑜𝑣𝑒 𝑤ℎ𝑒𝑟𝑒 𝑔𝐴 = 0 𝑡ℎ𝑒𝑛 𝑔𝐴 > 0,

𝑡ℎ𝑒 𝑎𝑟𝑟𝑜𝑤 𝑡𝑜𝑤𝑎𝑟𝑑𝑠 𝑡ℎ𝑒 𝑟𝑖𝑔ℎ𝑡. This is the case as if 𝑔𝐴 is smaller than corresponding to
′𝑔𝐴 = 0, then we subtract less, as 𝜃 − 1 is negative (remember that we have 𝜃 < 1),
therefore 𝛽𝑔𝐾 (𝑡) + 𝛾𝑛 + (𝜃 − 1)𝑔𝐴 (𝑡) becomes positive and ′𝑔𝐴 (𝑡) > 0. Can use the
opposite argument when we are above.
 - Below 𝑔𝐴 (𝑡) = 0: 𝑖𝑓 𝑔𝐴 𝑖𝑠 𝑎𝑏𝑜𝑣𝑒 𝑤ℎ𝑒𝑟𝑒 𝑔𝐴 = 0 𝑡ℎ𝑒𝑛 𝑔𝐴 < 0,
𝑡ℎ𝑒 𝑎𝑟𝑟𝑜𝑤 𝑡𝑜𝑤𝑎𝑟𝑑𝑠 𝑡ℎ𝑒 𝑙𝑒𝑓𝑡

The General Case, the ′𝒈𝑨 = 𝟎 locus

We draw the ′𝑔𝐴 = 0 lucus into a diagram with 𝑔𝐴 on the first axis and 𝑔𝐾 on the second axis:

6.10 The General Case, 𝛽 + 𝜃 < 1

Remember, Beta tells us how productive capital is in the production of one knowledge.

If 𝛽 + 𝜃 < 1, decreasing returns to technology: the slope of the ′𝑔𝐴 = 0 locus, is greater than the
slope of the ′𝑔𝐾 = 0 locus:
𝑑𝑔𝐾 (𝑡) 𝑑𝑔𝐾 (𝑡)
= |′𝑔𝐴 (𝑡)=0 > = |′𝑔𝐾 (𝑡)=0
𝑑𝑔𝐴 (𝑡) 𝑑𝑔𝐴 (𝑡)
↔
1−𝜃
>1
𝛽
↔
1>𝜃+𝛽
𝛽+𝜃 <1

Initial value of 𝑔𝐾 , that is 𝑔𝐾 (0), and initial value of 𝑔𝐴 , that is 𝑔𝐴 (0), depend on the initial values of
K, A and L. We need to slope of ′𝑔𝐴 to be greater than the slope of ′𝑔𝐾 , other wise they will not cross
eachother. We have shown that 𝛽 + 𝜃 < 1 corresponds to that the slope of the ′𝑔𝐴 = 0 locus is
greater than the slope of the ′𝑔𝐾 = 0 locus. They are meeting where the growth rate is stable at
both a and k. The reason for why they wont cross if the slope of a is smaller than the slope of k, is
that the intercept is lower for k than a.

Equilibrium: Where the two loci cross ′𝑔𝐴 (𝑡) = ′𝑔𝐾 (𝑡) = 0

′𝑔𝐴 = 0 ↔ 0 = 𝛽𝑔 ∗𝐾 + 𝛾𝑛 + (𝜃 − 1)𝑔 ∗𝐴
′𝑔𝐾 = 0 ↔ 0 = (1 − 𝛼)(𝑔 ∗𝐴 + 𝑛 − 𝑔 ∗𝐾 )
Equal to:

(Hence, we need to solve two equations in two unknowns because both needs to hold where the
two loci cross:)

(1 − 𝜃)𝑔 ∗𝐴 = 𝛽𝑔 ∗𝐾 + 𝛾𝑛

𝑔 ∗𝐾 = 𝑔 ∗𝐴 + 𝑛
Substitute second equation into the first

(1 − 𝜃)𝑔 ∗𝐴 = 𝛽(𝑔 ∗𝐴 + 𝑛) + 𝛾𝑛
𝛽+𝛾
𝑔 ∗𝐴 = 𝑛
1−𝜃−𝛽
𝑔 ∗𝐾 = 𝑔 ∗𝐴 + 𝑛
𝛽+𝛾 1+𝛾−𝜃
𝑔 ∗𝐾 = ( + 1) 𝑛 = 𝑛
1−𝜃−𝛽 1−𝜃−𝛽
Equilibrium growth depends positively on population growth, productivity of capital, 𝛽, productivity
of labour 𝛾, and productivity of knowledge 𝜃.

Stability: The equilibrium is stable as narrows point toward the equilibrium E. We have stability both
regarding the ′𝑔𝐴 = 0 = ′𝑔𝐾 = 0. All the arrows points towards the equilibrium, in contrast to the
Ramsey model, where they all don’t point towards the equilibrium.

We differentiate with respect to 𝛽:

 𝑑𝑔𝐴∗ 1 − 𝜃 − 𝛽 − (𝛽 + 𝛾)(−1) 1 − 𝜃 − 𝛽 + (𝛽 + 𝛾) 1−𝜃+𝛾
= 2
𝑛= 2
𝑛= 𝑛>0
𝑑𝛽 (1 − 𝜃 − 𝛽) (1 − 𝜃 − 𝛽) (1 − 𝜃 − 𝛽)2

Because remember that 𝜃 < 1. Beta increases, capital increases. Capital becomes more efficient for
∗
𝑑𝑔𝐴
knowledge production, then knowledge will be produced at a higher growth rate. 𝑑𝛽
> 0.

We differentiate wrt 𝛾
𝑑𝑔𝐴∗ 1
= 𝑛>0
𝑑𝛾 1−𝜃−𝛽
When labour becomes more efficient in the production of knowledge, that is 𝛾 increases, then the
growth rate of knowledge in the steady state, increases. This happens because now labour is more
efficient for knowledge production and therefore knowledge, A, will be produced at a higher growth
rate.

What happens when 𝜃 increases, that is exisisting knowledge A becomes more efficient in
knowledge production:
𝑑𝑔 ∗𝐴 𝛽 + 𝛾(−1) 𝛽+𝛾
=− 2
𝑛= 𝑛>0
𝑑𝜃 (1 − 𝜃 − 𝛽) (1 − 𝜃 − 𝛽)2
Hence, when exisiting knowledge becomes more important for knowledge production, then the
growth rate of knowledge production increases.

Let’s now differentiate wrt n along the balances growth path:

𝛽+𝛾
𝑔 ∗𝐴 = 𝑛
1−𝜃−𝛽
𝑑𝑔𝐴∗ 𝛽+𝛾
= >0
𝑑𝑛 1−𝜃−𝛽
When there are added more people into the economy every point in time, then we have more
people to produce knowledge. And therefore, the growth rate of knowledge production increases.

𝑔𝐾∗ = 𝑔𝐴∗ + 𝑛
As the growth rate of capital is equal to the growth rate of knowledge plus population growth, then
the impact of the parameter values on the growth rate of knowledge will be similar to the impact on
the growth rate of capital. With respect to population growth, we showed that a higher n increased
𝑔𝐴∗ and therefore a higher n also increases 𝑔𝐾∗ then we have a total positive impact from population
growth on the growth rate of capital.
6.11 The General, 𝛽 + 𝜃 = 1, 𝑛 = 0
If 𝛽 + 𝜃 = 1, and no population growth, 𝑛 = 0, the slope of the ′𝑔𝐴 = 0 locus is 1, which is equal to
the slope of the ′𝑔𝐾 = 0 locus:
𝑑𝑔𝐾 (𝑡) 1−𝜃
= |′𝑔𝐴 (𝑡)=0 = =1
𝑑𝑔𝐴 (𝑡) 𝛽
↔
𝑑𝑔𝐾 (𝑡)
1 = 𝜃 + 𝛽, = |′𝑔𝐾 (𝑡)=0 = 1
𝑑𝑔𝐴 (𝑡)
Hence, if there is population growth, then as the ′𝑔𝐴 = 0 curve crosses the second acis in the
negative part of the diagram, and the ′𝑔𝐾 = 0 curve crosses the second axis in the positive part of
the diagram and as the two curves have the same slope, then they will be parallel and never cross,
UNLESS n=0, because then the two curves lie on top of each other. It is the only way we can get a
solution!!!!

Intercept with second axis, when n=0:

𝑔𝐾 (𝑡)|′𝑔𝐴(𝑡)=𝑔𝐴(𝑡)=0 = 0, 𝑔𝐾 (𝑡)|′𝑔𝐾(𝑡)=𝑔𝐴(𝑡)=0 = 0

Hence, the two loci are on top of each other. Unless they are on top of eachother, they are just
going to be parallel.

Initial value of 𝑔𝐾 , that is 𝑔𝐾 (0), and initial value of 𝑔𝐴 , that is 𝑔𝐴 (𝑡), depend on initial values of K, A
and L.

Equilibrium:

′𝑔𝐴 (𝑡) = 0 ↔ 0 = 𝛽𝑔 ∗𝐾 + (𝜃 − 1)𝑔 ∗𝐴

′𝑔𝐾 (𝑡) = 0 ↔ 0 = (1 − 𝛼)(𝑔 ∗𝐴 − 𝑔 ∗𝐾 )
Equal to as

0 = 𝛽𝑔 ∗𝐾 + (𝜃 − 1)𝑔 ∗𝐴
0 = (𝑔 ∗𝐴 − 𝑔 ∗𝐾 )
Which again is equal to:

(1 − 𝜃)𝑔 ∗𝐴 = 𝛽𝑔 ∗𝐾

𝑔 ∗𝐾 = 𝑔 ∗𝐴
As 1 = 𝜃 + 𝛽 is 1 − 𝜃 = 𝛽, and hence the two equations are identical.

Hence, any equilibria on the 45 degree line, all point will be on the balanced growth path, with equal
growth in capital and knowledge.

Stability: The equilibria are stable as arrows point toward the 45 degree line.

It is possible to derive the unique though, but not part of the syllabus.

6.12 The nature of Knowledge and the determinants of the allocation of resources to
RD
- The shares going to research and development are of course not constant
- So what determines 𝑎𝐿 and 𝑎𝐾 ?
- Knowledge comes in many forms, but the general thing is that it is a non-rival good (can be
used by everyone at the same time)
- This implies: marginal cost of supplying an item of knowledge to an additional user, once
discovered, is zero
- Rental price of knowledge in a competitive market is zero.
- Knowledge may be excludable: depends on the nature of the knowledge (recipe for Coca
Cola) and institutions (property rights, for example patent laws)
- The degree of excludability then determines how many private resources we can expect to
be allocated into the production of a certain good
- If not excludable, it could call for public funding, if there is an externality for society.
- Coronavirus: production and development ok?
- Denmark: a lot of public funding for this good. Where all education is public. People in
general are getting some? Important for other people than those who can directly use it.

6.13 Learning by Doing

- Alternative idea to knowledge accumulation: Learning-by-doing
- The idea is that while producing goods, workers think of improvements
- Suppose the production function has the form

𝑌(𝑡) = 𝐾(𝑡)𝛼 [𝐴(𝑡)𝐿(𝑡)]1−𝛼

Assume: knowledge production is a side effect to capital production:

𝐴(𝑡) = 𝐵𝐾(𝑡)∅ , 𝐵 > 0, ∅ > 0

- B is a constant measuring how efficient capital is of the production of knowledge.
- Insert A(t) into Y(t)
1−𝛼
𝑌(𝑡) = 𝐾(𝑡)𝛼 [𝐵𝐾(𝑡)∅ 𝐿(𝑡)]
 𝑌(𝑡) = 𝐾(𝑡)𝛼 𝐵1−𝛼 𝐾(𝑡)∅(1−𝛼) 𝐿(𝑡)1−𝛼
- Notice that capital does not depreciate
- Growth in capital is a fraction of output and hence ′𝐾(𝑡) = 𝑠𝑌(𝑡)

′𝐾(𝑡) = 𝑠𝐾(𝑡)𝛼+∅(1−𝛼) [𝐵𝐿(𝑡)]1−𝛼

- And hence the growth rate of capital is found by dividing by K(t)

′𝐾(𝑡) 𝑠𝐾(𝑡)𝛼+∅(1−𝛼) [𝐵𝐿(𝑡)]1−𝛼

𝑔𝐾 (𝑡) = =
𝐾(𝑡) 𝐾(𝑡)

The growth rate in capital with learning by doing

𝑔𝐾 = 𝑠𝐾(𝑡)𝛼+∅(1−𝛼) [𝐵𝐿(𝑡)]1−𝛼

- Where the exponent to K(t) can be written as

𝛼 + ∅(1 − 𝛼) − 1 = (∅ − 1)(1 − 𝛼)
- Then we get
𝑔𝐾 = 𝑠𝐾(𝑡)(∅−1)(1−𝛼) [𝐵𝐿(𝑡)]1−𝛼

Taking logs and differentiate wrt time

𝑙𝑛(𝑔𝐾 ) = 𝑙𝑛𝑠 + (∅ − 1)(1 − 𝛼)𝑙𝑛𝐾(𝑡) + (1 − 𝛼)[𝑙𝑛𝐵 + 𝑙𝑛𝐿(𝑡)]

′𝑔𝐾 (𝑡) ′𝐾(𝑡) ′𝐿(𝑡)
= (∅ − 1)(1 − 𝛼) + (1 − 𝛼)
𝑔𝐾 (𝑡) 𝐾(𝑡) 𝐿(𝑡)
′𝐿(𝑡) ′𝐾(𝑡)
Noting that 𝐿(𝑡)
= 𝑛 and 𝐾(𝑡)
= 𝑔𝐾 (𝑡), we have the growth rate of the growth rate of K:

′𝑔𝐾 (𝑡)
= (∅ − 1)(1 − 𝛼)𝑔𝐾 (𝑡) + (1 − 𝛼)𝑛
𝑔𝐾 (𝑡)

6.13.1 Learning by doing, example

The change in the growth rate of K over time with learning by doing is
2
′𝑔𝐾 (𝑡) = (1 − 𝛼) (𝑛𝑔𝐾 (𝑡) + (∅ − 1)(𝑔𝐾 (𝑡)) )

Consider productivity of capital to be 𝛼 = 0,5, population growth to be n=2%, and productivity of

capital for knowledge production to be ∅:
2
′𝑔𝑘 (𝑡) = 0,5 (2𝑔𝐾 (𝑡) + (∅ − 1)(𝑔𝐾 (𝑡)) )
 1. Draw the curve for the change in the growth rate in knowledge when ∅ = 0,5 into a diagram
where 𝑔𝐴 (𝑡) 𝑖𝑠 0,1,2,3,4,5

𝑔𝐾 0 1 2 3 4 5
′𝑔𝐾 0,5(2 ∗ 0 0,5(2 ∗ 1 0,5(2 ∗ 2 0,5(2 ∗ 3 0,5(2 ∗ 4 0,5(2 ∗ 5
− 0,5 ∗ 0) − 0,5 ∗ 12 ) − 0,5 ∗ 22 ) − 0,5 ∗ 32 ) − 0,5 ∗ 42 ) − 0,5 ∗ 52 )
0 0,75 1 0,75 0 -1,25

Change in growth rate of K

1.5

0.5

0
0 1 2 3 4 5 6

-0.5

-1

-1.5

Change in growth rate of K

It looks very much like the endogenous model without capital.

2. Find the stable growth rate for capital in this model of learning by doing, that is, find 𝑔 ∗𝐾 :
We then use the equation and find the stable point:

′𝑔𝑘 (𝑡) = 0

0,5(2𝑔 ∗𝐾 − 0,5(𝑔 ∗𝐾 )2 ) = 0
(2𝑔 ∗𝐾 − 0,5(𝑔 ∗𝐾 )2 ) = 0

2𝑔 ∗𝐾 = 0,5(𝑔 ∗𝐾 )2
2 = 0,5𝑔 ∗𝐾
𝑔 ∗𝐾 = 4

3. Consider now a change in Ø to 2 and draw the curve for 𝑔𝐾 (𝑡)=0,1,2,3. What happens?

𝑔𝐾 1 2 3
′𝑔𝐾 0,5(2 ∗ 1 + 0,5 0,5(2 ∗ 2 + 0,5 ∗ 22 ) 0,5(2 ∗ 3 + 0,5 ∗ 32 )
∗ 12 )
 1,5 4 7,5

Growth rate
8

0
1 2 3

Growth rate

′𝑔𝐾 (𝑡) is now an exponential function of 𝑔𝐾 (𝑡). This means that the growth rate never converges,
and it will explode over time.

4. Let Ø=1, then we obtain:

2
′𝑔𝑘 (𝑡) = 0,5 (2𝑔𝐾 (𝑡) + (1 − 1)(𝑔𝐾 (𝑡)) )
′𝑔𝑘 (𝑡) = 0,5(2𝑔𝐾 (𝑡))
′𝑔𝑘 (𝑡) = 𝑔𝐾 (𝑡)

Series 1
4.5

3.5

2.5

1.5

0.5

0
1 2 3 4

Series 1
6.14 The relationship for the growth rate
The change in the growth rate of K over time with learning by doing is

(multiply with it with 𝑔𝐾 on both sides)

2
′𝑔𝐾 (𝑡) = (1 − 𝛼)𝑛𝑔𝐾 (𝑡) + (∅ − 1)(1 − 𝛼)(𝑔𝐾 (𝑡))

Move the order of the numbers so you can see that it is going to look like what we had for the
endogenous model without capital.
2
′𝑔𝐾 (𝑡) = (1 − 𝛼) (𝑛𝑔𝐾 (𝑡) + (∅ − 1)(𝑔𝐾 (𝑡)) )

As the curve for the model for the dynamics for A without capital. To draw the curve as a function of
𝑔𝐾 :

We first find the slope by differentiating with respect to the growth rate in capital, 𝑔𝐾 (𝑡):

𝑑′ 𝑔𝐾 (𝑡)
= (1 − 𝛼)(𝑛 + 2(∅ − 1)𝑔𝐾 (𝑡))
𝑑𝑔𝐾 (𝑡)
And the second derivative tells us whether the slop is constant or changing with 𝑔𝐾 (𝑡)
′
𝑑2 𝑔𝐾 (𝑡)
= (2(∅ − 1)(1 − 𝛼))
𝑑𝑔𝐾 (𝑡)
Hence, the slope of the curve depends on the value of ∅, that is, the effect of the existing stock of
capital on knowledge production. Depends on Ø is smaller than, higher than or equal to 1.

6.13.1 ∅ < 𝟏: Diminishing returns, as K increases

Remember, learning by doing model: important with existing capital for knowledge production.

The change in the growth rate of K with learning by doing is:

2
′𝑔𝐾 (𝑡) = (1 − 𝛼) (𝑛𝑔𝐾 (𝑡) + (∅ − 1)(𝑔𝐾 (𝑡)) )

When ∅ < 1, then the growth rate of 𝑔𝐾 (𝑡) is increasing at first

𝑑′ 𝑔𝐾 (𝑡)
= (1 − 𝛼)(𝑛 + 2(∅ − 1)𝑔𝐾 (𝑡)) > 0
𝑑𝑔𝐾 (𝑡)
 ↔
𝑛 > 2(1 − ∅)𝑔𝐾 (𝑡)
Hence, until 𝑔𝐾 given by
𝑛
= 𝑔𝐾 (𝑡)
2(1 − ∅)
For all Ø<1, we have the second order derivative is negative:
′
𝑑2 𝑔𝐾 (𝑡)
= (2(∅ − 1)(1 − 𝛼)) < 0
𝑑𝑔𝐾 (𝑡)
The growth rate in the growth rate of capital is positive until

′𝑔𝐾 (𝑡)
=0
𝑔𝐾 (𝑡)
↔
0 = (1 − 𝛼)(∅ − 1)𝑔 ∗𝐾 + (1 − 𝛼)𝑛
𝑛
𝑔 ∗𝐾 =
1−∅
This is the stable solution.

Example:

The change in the growth rate of K over time with learning by doing is
2
′𝑔𝐾 (𝑡) = (1 − 𝛼) (𝑛𝑔𝐾 (𝑡) + (∅ − 1)(𝑔𝐾 (𝑡)) )

Consider productivity of capital to be 𝛼 = 0,5, population growth to be 𝑛 = 2 percent, and

productivity of capital for knowledge production to be ∅ = 0,5.

1. Draw the curve for the change over time in the growth rate of capital for 𝑔𝐾 (𝑡) = 0,1, … ,5
2. Find the stable growth rate for capital in this model of learning by doing, that is find 𝑔 ∗𝐾 .
3. Consider now a change in ∅ to 2 and draw the curve for 𝑔𝐾 (𝑡) = 0,1,2,3,4

6.13.2 ∅ > 𝟏: Increasing returns, as K increases

The change in growth rate of K over time with learning by doing is
2
′𝑔𝐾 (𝑡) = (1 − 𝛼) (𝑛𝑔𝐾 (𝑡) + (∅ − 1)(𝑔𝐾 (𝑡)) )

When ∅ > 1, the growth rate of 𝑔𝐾 is positive and explosive

𝑑′ 𝑔𝐾 (𝑡)
= (1 − 𝛼)(𝑛 + 2(∅ − 1)𝑔𝐾 (𝑡)) > 0
𝑑𝑔𝐾 (𝑡)
′
𝑑2 𝑔𝐾 (𝑡)
= (2(∅ − 1)(1 − 𝛼)) > 0
𝑑𝑔𝐾 (𝑡)
6.13.3 ∅ = 𝟏: Constant returns, as K increases
The change in the growth rate of K over time with learning by doing is
2
′𝑔𝐾 (𝑡) = (1 − 𝛼) (𝑛𝑔𝐾 (𝑡) + (∅ − 1)(𝑔𝐾 (𝑡)) )

When ∅ = 1, the growth rate of 𝑔𝐾 is constant

𝑑′ 𝑔𝐾 (𝑡)
= (1 − 𝛼)𝑛 > 0
𝑑𝑔𝐾 (𝑡)
′
𝑑2 𝑔𝐾 (𝑡)
=0
𝑑𝑔𝐾 (𝑡)
Here, it is not the case that the growth is constant, it is the case that increase in the growth rate is
constant. We will not have the same growth rate as the year after, it is still going to be higher and
higher.

THIS IS NOT SYLLABUS:

∅ = 𝟏: Constant returns, n=0, and no population growth

- Learning by doing, constant return and no population growth, n=0. The AK model.
- The production function is then:

𝑌(𝑡) = 𝐾(𝑡)𝛼 [𝐴(𝑡)𝐿]1−𝛼

Knowledge production is a side effect to capital production when ∅ = 1:

𝐴(𝑡) = 𝐵𝐾(𝑡), 𝐵 > 0

- Insert A(t) into Y(t)

𝑌(𝑡) = 𝐾(𝑡)𝛼 [𝐵𝐾(𝑡)𝐿]1−𝛼

𝑌(𝑡) = 𝐴𝐾(𝑡)
Where 𝐴 = [𝐵𝐿]1−𝛼 . Capital accumulation is then:

′𝐾(𝑡)
′𝐾(𝑡) = 𝑠𝐴𝐾(𝑡) ↔ 𝑔𝐾 (𝑡) = 𝑔 ∗𝐾 = = 𝑠𝐴
𝐾(𝑡)
- Hence, capital grows steadily at sA

∅ = 𝟏, n=0, and savings increase growth

- The growth rate in capital is

′𝐾(𝑡)
𝑔𝐾 (𝑡) = 𝑔 ∗𝐾 = = 𝑠𝐴
𝐾(𝑡)
- Hence, capital grows steadily at sA
- Higher savings rate, higher s will increase growth:
 𝑑𝑔 ∗𝐾
= 𝐴 = 𝐵1−𝛼 𝐿1−𝛼
𝑑𝑠
Intuition:

Increased capital raises output through

- Its direct role on production 𝐾(𝑡)𝛼

- Indirectly through new knowledge and thereby making capital more productive, 𝐾(𝑡)1−𝛼

SUM UP:

- We have considered a Solow growth model with endogenous knowledge, RD and capital
- We first considered the model without capital and increasing, diminishing and constant
return to RD
- We then considered the general case with capital and drew a ′𝑔𝐴 = 0 locus and ′𝑔𝐾 = 0
locus into a phase diagram
- We again considered different cases
- Finally, we considered a learning by doing model,
- The AK model, where we can obtain perpetual growth (not syllabus)
- Hence, this chapter served to give some understanding of what is important for long run
growth: the investments in RD

7. Cross-Country Income Differences

We are challenged in explaining cross-country income differences. So far we have investigated the
following potential explanations:

- Differences in capital stocks across countries are not able to explain differences in income
across countries. (Studied in the Solow-model, Chapter 1, and in Ramsey-Cass-Koopmans
model, Chapter 2).
- Knowledge production and dissemination of technology across countries. Since
knowledge/technology is non-rival it is unlikely that this should be an important explanation
for cross-countries income differences. (Studied in Endogenous Growth, Chapter 3).

Empirical work on Cross-Country Differences

Consider factors where the influence on income is clear and direct:

- Capital and human capital - we will focus on human capital

- Growth accounting techniques
- Advantage: High confidence in conclusions; Drawback:
- Considers only immediate determinants of income.

Other branch tries to go deeper.

- Social infrastructure: Political institutions, geography, religion → a reason for the countries
to have these. Where to start? How far back should we go?
- Statistical techniques.
 - Drawback: Lower confidence in conclusions.
- Advantage: Considers more deep determinants of income.

7.1 The Solow Model with Human Capital

Extending the Solow Model to include Human Capital

- Difference: in stead of having A times L, we will have an A times H. Remember: going to look
like Solow model, small differences. Labour force would be homogenous. Now we allow this
labour force to heterogeneous, that is; we have different kind of workers in the labour force.
- Model of growth that includes both physical and human capital
- Purpose: Study cross-country income differences.
- Solow model: exogenous savings rate. If we instead used the Ramsey model, which includes
endogenous savings, to explain cross-country differences, then we would need some
evidence on cross country for preferences, which is more difficult to obtain.

We assume we have a CRS (constant returns to scale) production function with human capital H(t):

𝑌(𝑡) = [𝐾(𝑡)]𝛼 [𝐴(𝑡)𝐻(𝑡)]1−𝛼 , 0 < 𝛼 < 1

- As before: Y(t) is output, K(t) is a physical and A(t) is knowledge
- We again assume a positive depreciating rate, 𝛿 and hence capital accumulation is given by

′𝐾(𝑡) = 𝑠𝑌(𝑡) − 𝛿𝐾(𝑡)

- S, the saving rate is exogenous (like in chapter 1)

Chapter 1: Solow Growth model, micro foundation, no endogenous growth, savings rate was
exogenous.

Chapter 2: Endogenous consumption, and therefore savings. Ramsey model – continuous time,
everyone lived forever. Diamond model, 2 periods of time where we had young who consumed and
worked and old who only consumed and lived on their savings. The savings is the capital
accumulated as savings when they were young.

Chapter 3: Disregard this micro foundation which we had in chapter 2, let us in stead return the kind
of the setup we had in chapter 1, we had micro foundation, no house hold maximization of utility or
firms maximizing profits, in stead we had exogenous saving rate. Let us now endogenize knowledge
production, that is what we did in chapter 3, part of it.

Now we are back to the same assumption as in chapter 1.

- Capital is the existing capital stock in the economy.

- Knowledge growth is again constant in order to focus on the role of human capital

′𝐴(𝑡) = 𝑔𝐴(𝑡)
Human Capital

- Worker’s human capital depends on year of education

𝐻(𝑡) = 𝐿(𝑡)𝐺(𝐸)
, where L(t) is number of workers
- G(E) is a function E, which is years of schooling. (could be measured in a different way, in
some countries education is important, some not)
 - Labour force grows at a constant exogenous rate:

′𝐿(𝑡) = 𝑛𝐿(𝑡)
- N is the growth rate of population and therefore labour force, because now everyone is
entering the labour force when they are born. Everyone is working (a difference to the
diamond model)
- Functional form of G(E). We must have 𝐺 ′ (𝐸) > 0. That is human capital then increases in
years of schooling. We need that it is increasing in education. The more education you have
the more human capital you have. When E is increasing, G is increasing, and H is increasing.
- Evidence suggests that one year extra schooling increases an individual’s wage by the same
percentage amount. Hence, if wages equal to marginal productivity

𝐺(𝐸) = 𝑒 ∅𝐸 , ∅ > 0

7.1.1 Dynamics of k
- Dynamics as in Solow model, as G(E) does not depend on t
- Let k(t) = K(t)=(A(t)H(t)) = K(t)=(A(t)G(E)L(t))
- We are interested in drawing the curve ‘k(t) (with a dot because it is growth) into a diagram,
and it will look exactly as the Solow growth model, we just now have that instead of having
k(t) = K(t)/(A(t)L(t)) we will have k(t) = K(t)/(A(t)H(t)). We reached the steady level where k is
stable and not will change, but it will be a change in output. We will have that output per
worker in the steady state will then depend on the growth rate in the economy. Here we will
have a steady state in output per worker, by increasing human capital. (Each person will be
richer and richer every year due to the increase in technology. Same way here – a possibility
of increasing output per worker by increasing human capital.)
- The development in capital per unit of effective labour is then (notice, I have taking 1/G(E)
outside the parenthesis as it does NOT depend on time, so when I differentiate wrt time, I
will not differentiate G(E) with respect to time. A fraction and a product.)

′𝐾(′ 𝑡) 1 ′𝐾(𝑡)𝐴(𝑡)𝐿(𝑡) 𝐾(𝑡)(′ 𝐴(𝑡)𝐿(𝑡) + 𝐴(𝑡)′ 𝐿(𝑡)

= ( − )
𝐴(𝑡)𝐺(𝐸)𝐿(𝑡) 𝐺(𝐸) (𝐴(𝑡)𝐿(𝑡))2 (𝐴(𝑡)𝐿(𝑡))
2

′𝐾(𝑡) 𝐾(𝑡) ′𝐴(𝑡) 𝐾(𝑡) ′𝐿(𝑡)

𝑘(′ 𝑡) = ( − − )
𝐴(𝑡)𝐺(𝐸)𝐿(𝑡) 𝐴(𝑡)𝐺(𝐸)𝐿(𝑡) 𝐴(𝑡) 𝐴(𝑡)𝐺(𝐸)𝐿(𝑡) 𝐿(𝑡)

Insert for ‘K(t)

𝐾(𝑡)
𝑘(𝑡) =
𝐴(𝑡)𝐺(𝐸)𝐿(𝑡)
′𝐴(𝑡)
=𝑔
𝐴(𝑡)
′𝐿(𝑡)
=𝑛
𝐿(𝑡)

Then we get:
𝑠𝑌(𝑡) − 𝛿𝐾(𝑡)
𝑘(′ 𝑡) = − 𝑘(𝑡)𝑔 − 𝑘(𝑡)𝑛
𝐴(𝑡)𝐺(𝐸)𝐿(𝑡)
 𝑠𝑌(𝑡) 𝛿𝐾(𝑡)
𝑘(′ 𝑡) = − − 𝑘(𝑡)𝑔 − 𝑘(𝑡)𝑛
𝐴(𝑡)𝐺(𝐸)𝐿(𝑡) 𝐴(𝑡)𝐺(𝐸)𝐿(𝑡)
𝑠𝑌(𝑡)
𝑘(′ 𝑡) = − 𝛿𝑘(𝑡) − 𝑘(𝑡)𝑔 − 𝑘(𝑡)𝑛
𝐴(𝑡)𝐺(𝐸)𝐿(𝑡)

Insert for 𝑓(𝑘(𝑡))

𝑌(𝑡)
𝑓(𝑘(𝑡)) = = 𝑘(𝑡)𝛼
𝐴(𝑡)𝐺(𝐸)𝐿(𝑡)
In total we get:

′𝒌(𝒕) = 𝒔𝒌(𝒕)𝜶 − (𝒏 + 𝒈 + 𝜹)𝒌(𝒕)

7.1.2 What we observe of the dynamics of k(t)

The dynamics of k(t) is as in the Solow Growth model

′𝑘(𝑡) = 𝑠𝑘(𝑡)𝛼 − (𝑛 + 𝑔 + 𝛿)𝑘(𝑡)

The steady is found where ′𝑘(𝑡) = 0

𝑠(𝑘 ∗)𝛼 = (𝑛 + 𝑔 + 𝛿)𝑘 ∗

↔
1
𝑠 1−𝛼
𝑘 ∗= ( )
𝑛+𝑔+𝛿
𝑌 𝑌
by that we also have 𝑦 ∗ = (𝑘 ∗)𝛼 and as 𝑦 = (G(E)AL) then 𝐿 = 𝑦𝐺(𝐸)𝐴, (notice the difference to
chapter 1 is that we divide by G(E) as well) we observe:

- G(E)A does not depend on savings: A grows at the rate g and G(E) is constant.
- Years of schooling increases output per workers on a balanced growth path by the same
𝑌 𝑌
proportion as it increases : Remember, if 𝑦 = (G(E)AL) then 𝐿 = 𝑦𝐺(𝐸)𝐴. Hence even
though in a steady state output measured in efficiency units of human capital is stable, then
when we measure output per worker, this changes with A and G(E). The latter is the new
thing is this human capital model.
𝑌 𝑌
- Remember in the Solow Growth model we had that ′ / = 𝑔 in the steady state.
𝐿 𝐿
- But even though this is still, the growth rate in output per worker, and therefore the right
measure to use if we want to consider how big the growth rate in output per worker is (as
G(E) does noe depend on time) then output per worker increases with G(E), so if education
𝑌
increases, then each person in the economy becomes richer, as then, 𝐿 = 𝑦𝐺(𝐸)𝐴
increases!!! In a steady state.
𝑌
𝑑 /𝑑𝐸 = 𝑦𝐺(𝐸)𝐴 > 0
𝐿
Implication for cross-country differences:

1. Can depend on differences in human capital (different educational level in different countries)

2. same dynamics as in Solow growth model

7.2 Decomposing output

How we measure the difference between countries:

We have the production function with human capital

𝑌(𝑡) = [𝐾(𝑡)]𝛼 [𝐴(𝑡)𝐻(𝑡)]1−𝛼 , 0 < 𝛼 < 1

If we then divide by population, remember 𝐿(𝑡)𝛼+−𝛼 = 𝐿(𝑡)
𝛼 1−𝛼
𝑌(𝑡) 1 𝐾(𝑡) 𝐻(𝑡)
= 𝛼+1−𝛼
[𝐾(𝑡)]𝛼 [𝐴(𝑡)𝐻(𝑡)]1−𝛼 = [ ] [ 𝐴(𝑡)]
𝐿(𝑡) 𝐿(𝑡) 𝐿(𝑡) 𝐿(𝑡)

Then take logs and consider a country i

𝑌𝑖 𝐾𝑖 𝐻𝑖
𝑙𝑛 = 𝛼𝑙𝑛 + (1 − 𝛼) (𝑙𝑛 + 𝑙𝑛𝐴𝑖 )
𝐿𝑖 𝐿𝑖 𝐿𝑖
𝑌𝑖 𝐾𝑖 𝐻𝑖
𝑙𝑛 = 𝛼𝑙𝑛 + (1 − 𝛼)𝑙𝑛 + (1 − 𝛼)𝑙𝑛𝐴𝑖
𝐿𝑖 𝐿𝑖 𝐿𝑖
Examining this equation we have

- RHS, except A can be measured

- A is a residual
- All else equal: if A higher than K/L is higher (all else equal then k* is the same)
- In word: high productivity level implies high capital stock per worker

Consider a country i
𝑌𝑖 𝐾𝑖 𝐻𝑖
= 𝛼𝑙𝑛 + (1 − 𝛼)𝑙𝑛 + (1 − 𝛼)𝑙𝑛𝐴𝑖
𝐿𝑖 𝐿𝑖 𝐿𝑖
- An alternative composition may be more interesting
𝑌
- Subtract 𝛼𝑙𝑛 𝐿 𝑖
𝑖
𝑌𝑖 𝐾𝑖 𝑌𝑖 𝐻𝑖
(1 − 𝛼)𝑙𝑛 = 𝛼 (𝑙𝑛 − 𝑙𝑛 ) + (1 − 𝛼)𝑙𝑛 + (1 − 𝛼)𝑙𝑛𝐴𝑖
𝐿𝑖 𝐿𝑖 𝐿𝑖 𝐿𝑖
𝑌𝑖 𝐾𝑖 𝐻𝑖
(1 − 𝛼)𝑙𝑛 = 𝛼𝑙𝑛 + (1 − 𝛼)𝑙𝑛 + (1 − 𝛼)𝑙𝑛𝐴𝑖
𝐿𝑖 𝐿𝑖 𝐿𝑖
𝑌𝑖 𝛼 𝐾𝑖 𝐻𝑖
𝑙𝑛 = 𝑙𝑛 + 𝑙𝑛 + 𝑙𝑛𝐴𝑖
𝐿𝑖 (1 − 𝛼) 𝐿𝑖 𝐿𝑖
- In words: Output per worker = capital share of output + education level of workforce +
productivity
- Capital share of output the same on balanced growth path in capital accumulation models
Results:

- Consider a country i
𝑌𝑖 𝛼 𝐾𝑖 𝐻𝑖
𝑙𝑛
= 𝑙𝑛 + 𝑙𝑛 + 𝑙𝑛𝐴𝑖
𝐿𝑖 (1 − 𝛼) 𝐿𝑖 𝐿𝑖
- 1990’s compare richest five to poorest five countries
- Data for income shares suggest 𝛼 = 1/3 (physical capital’s share in the production function)
𝑌
LHS difference 3.5 (implies 𝑌 𝑟𝑖𝑐ℎ ~32)
𝑝𝑜𝑜𝑟

- Capital share difference 0.6 (implies capital share rich 1.8 times higher)
- Education (measured as years of schooling) difference 0.8 (implies effective education level
2.2 times higher for rich vs poor countries)
- Productivity difference 2.1 (implies productivity 8 times higher for richer than poor)

A is by far the most important factor. So how do we measure A?

7.3 What is A?
A is the primary driver of differences between rich and poor countries

- Two possibilities in Romer:

1. Social Infrastructure
2. Something else (culture, geography, etc)

1. Social infrastructure
- Rules of the game, mostly the policy
- Romer splits it into:
→ Distortionary taxes
→ Poor contract enforcement and property protection
→ Government corruption
- To a degree can be measures

- Does better social infrastructure explain output gap?

- Imagine the following relationship were true:
𝑌𝑖
𝑙𝑛 = 𝑎 + 𝑏𝑆𝑙𝑖 + 𝑒𝑖
𝐿𝑖
- We expect that b is positive (and large?)
- How can we estimate b?
- Two immediate problems:
→ how to measure SI
→ OLS: correlation of SI and error term 𝑒𝑖
- Solutions:
→ Instrumental variables: something which changes 𝑆𝐼𝑖 but not 𝑒𝑖
→ Natural experiments.

- Well-known example: Hall and Jones (1999)

- Western Europeans (colonial) migration influenced social infrastructure
- Four instruments:
→ Fraction of English/West European language speaking population
 → Distance to equator (didn’t colonize the same way)
→ Geographical factors influencing trade openness
→ Size
- We find social infrastructure super important
- But instruments may affect other stuff in ei :
→ Climate/geography might directly impact output
→ West European culture, not social infrastructure

Natural experiment:

- More compelling is simply the natural experiment

- North and South Korea same country until 1945
→ similar geography, similar culture
→ social infrastructure
→ South Korea much richer than North Korea
- But also international pariah with sanctions – difficult for North Korea, because a lot of other
countries sanction them due to their political system, they lowered output per worker.

2. Something else, geography, culture

Does geography determine social infrastructure?

Two ideas related to colonialism

- Europeans died more near the tropics

- Slavery flourished near tropics because sugar and cotton

But still, geography can have direct impact

Culture:

- "In the case of culture, it seems clear that views and norms about such matters as thrift,
education, trust, and the merits of material success could directly affect economic
performance...the evidence on this issue is very limited."
- The sociological view, cultural determinism
→ If so, not clear what we can/should do
 → Plays into deserving vs undeserving poor
→ Implications for immigration debate?
→ A dangerous idea?
→ and where do we start, culture is likely to be endogenous.

- Trust increases the probability of becoming an entrepreneur

- Trust index – Scandinavia high and New Zealand
- Corruption perception index

SUM UP:

- We have tried to explain income differences between countries

- Differences in human capital, physical capital are important
- but ’something else’ is the most important factor - A
- may be political institutions, geography, religion, culture
- a lot of research in this area currently.

8. Differences between the chapters

Solow model (1) Ramsey Model and Endogenous growth Differences between
Diamond model (2) (3) countries (4)
Savings exogenous Savings endogenous Savings exogenous Savings exogenous
No heterogeneity Ramsey: No 3 models: Discussion about
heterogeneity 1. No capital social infrastructure.
2. With capital Capital explain little of
Diamond: 3. Learning by doing cross country
Heterogeneity difference in Y/L. A
1.2.: sector for explains more.
knowledge production Different
3.: knowledge would determinants of A
be a bi-product of given in the chapter
capital production in
the economy
No micro foundation Micro foundation No micro foundation No micro foundation

Household maximizing
utility and firms
maximizing profits
Diamond model:
Dynamix inefficiency
possible
Golden rule steady Ramsey: Always below
state: level where we golden rule due to
discounting
 maximize
consumption

9. Unemployment
Her research – important I guess.

Unemployment is one of the central subjects of macroeconomics Issues

- Why do we have an average (that is, natural or structural) unemployment rate in the
economy which is positive
- and why do unemployment vary across countries and over time? (related to different
unemployment rates in different countries, low in Northern, high in Southern)
- Now – rise in unemployment due to Corona virus
- OECD data: unemployment varies over time and across countries. Search and Matching
models.

9.1 Search and matching models, introduction:

- Essential: Workers and firms are heterogeneous in terms of idiosyncratic preferences, skills
and needs.
- Workers search for jobs and firms search for workers. As they are heterogeneous, this
searching process takes time
- We call this search friction
- Thereby we obtain unemployment. Will never be 0! Not possible. Always be some people
searching job
- The model is called the ’Diamond-Mortensen-Pissarides’ model – Nobel Prize
- We set the model in continuous time (also possible in descret)

Assumptions regarding the worker

- There is a continuum of workers normalised to one

- Workers can be unemployed or employed (not in the other models since we removed it, but
it could be included)
- An employed worker produces output y per unit of time and receives the wage level w(t) per
unit of time. Due to search friction
- An unemployed worker receives an exogenous constant income b>=0 per unit of time
- Workers are risk neutral
- Workers’ discount rate is r > 0

Assumptions regarding the firm

- A job can be vacant or filled (ad in newspaper etc, might take a while. Some takes longer
than others due to different qualification, tests, interviews)
- A filled job gives output y and costs w(t)
 - A job, filled or vacant, involves a constant exogenous cost c > 0 per unit of time (interviewer,
ad etc)
- There is free entry of new firms (important!!! Because it implies that you don’t have a fixed
number of firms in the economy, that is wrong, you will never have that)
- Hence the number of jobs is endogenous

Corona virus relevance:

- Exogenous shock
- Countries are in a recession
- Shut down firms → increase in unemployment → fewer jobs for all workers
- Workers are being fired

9.2 The matching function

- Let U(t) be the number of unemployed workers at time t
- Let V(t) denote the number of vacant jobs at time t
- The number of matches per unit of time is

𝑀(𝑡) = 𝑀(𝑈(𝑡), 𝑉(𝑡)), 𝑀𝑈 > 0, 𝑀𝑉 > 0

- Usually assume a Cobb-Douglas function

- 𝑀𝑈 and 𝑀𝑉 are the derivatives
- The matching function proxies for the complicated process of employer recruitment, worker
search and mutual evaluation
- More workers imply more matches and more vacancies imply more matches
- We assume the matching function exhibits constant returns to scale (between 0 and 1 in CD
function)

Example CD matching function:

𝑀(𝑡) = 𝑈(𝑡)1−𝛾 𝑉(𝑡)𝛾 , 0 < 𝛾 < 1

- Job finding rate (also called transition rate), how easy it is for a worker to find a job, then
becomes:

Equation (1)

𝑀(𝑡) 𝑈(𝑡)1−𝛾 𝑉(𝑡)𝛾 𝑉(𝑡)𝛾

𝑎(𝑡) = = = 𝑈(𝑡)−𝛾 𝑉(𝑡)𝛾 = = 𝜃(𝑡)𝛾
𝑈(𝑡) 𝑈(𝑡) 𝑈(𝑡)𝛾
𝑉(𝑡)
, where 𝜃(𝑡) = is called labour market tightness
𝑈(𝑡)

- hence more vacancies relative to unemployment, a tighter labour market, makes it easier for
a worker to get a job, hence the higher 𝜃 is.
- but as 𝛾<1 the increase in the job finding rate is less than proportional to the increase in
labour market tightness
- The vacancy filling rate is:
Equation (2)

𝑀(𝑡) 𝑈(𝑡)1−𝛾 𝑉(𝑡)𝛾 𝑉(𝑡)𝛾−1

𝛼(𝑡) = = = = 𝜃(𝑡)𝛾−1
𝑉(𝑡) 𝑉(𝑡) 𝑈(𝑡)𝛾−1

9.3 Equilibrium conditions

- We will use dynamic programming to describe the values of the various states
- Let 𝑉𝐸 (𝑡) denote the value of being employed at time t (not the same V as previously)
- That is, 𝑉𝐸 (𝑡) is the expected lifetime utility from time t and forward, discounted at time t
- 𝑉𝑈 (𝑡); 𝑉𝐹 (𝑡); and 𝑉𝑉 (𝑡) are determined similarly
- Jobs end at an exogenous rate 𝜆 (since there will always be some people being fired)
- Corona: a lot of firms are shutting down, and fire their workers → increase in 𝜆
- The return to being employed and unemployed are then (as asset equations)

Equation (3)

𝑟𝑉𝐸 (𝑡) = 𝑤(𝑡) + ′𝑉𝐸 (𝑡) − 𝜆(𝑉𝐸 (𝑡) − 𝑉𝑈 (𝑡))

𝑟𝑉𝐸 (𝑡) = 𝑡ℎ𝑒 𝑑𝑖𝑠𝑐𝑜𝑢𝑛𝑡𝑒𝑑 𝑣𝑎𝑙𝑢𝑒 𝑜𝑓 𝑏𝑒𝑖𝑛𝑔 𝑒𝑚𝑝𝑙𝑜𝑦𝑒𝑑 𝑎𝑡 𝑎 𝑝𝑜𝑖𝑛𝑡 𝑖𝑛 𝑡𝑖𝑚𝑒

𝑟𝑉𝐸 (𝑡) = 𝑤𝑎𝑔𝑒 + 𝑣𝑎𝑙𝑢𝑒 𝑜𝑓 𝑏𝑒𝑖𝑛𝑔 𝑒𝑚𝑝𝑙𝑜𝑦𝑒𝑑
− 𝑝𝑟𝑜𝑏𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝑜𝑓 𝑙𝑜𝑠𝑖𝑛𝑔 𝑦𝑜𝑢𝑟 𝑗𝑜𝑏(𝑑𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 𝑣𝑎𝑙𝑢𝑒
𝑜𝑓 𝑏𝑒𝑖𝑛𝑔 𝑒𝑚𝑝𝑙𝑜𝑦𝑒𝑑 𝑎𝑛𝑑 𝑣𝑎𝑙𝑢𝑒 𝑜𝑓 𝑏𝑒𝑖𝑛𝑔 𝑢𝑛𝑒𝑚𝑝𝑙𝑜𝑦𝑒𝑑)
Equation (4)

𝑟𝑉𝑈 (𝑡) = 𝑏 + ′𝑉𝑈 (𝑡) + 𝑎(𝑡)(𝑉𝐸 (𝑡) − 𝑉𝑈 (𝑡))

𝑟𝑉𝑈 (𝑡) = 𝑡ℎ𝑒 𝑑𝑖𝑠𝑐𝑜𝑢𝑛𝑡𝑒𝑑 𝑣𝑎𝑙𝑢𝑒 𝑜𝑓 𝑏𝑒𝑖𝑛𝑔 𝑢𝑛𝑒𝑚𝑝𝑙𝑜𝑦𝑒𝑑 𝑎𝑡 𝑎 𝑝𝑜𝑖𝑛𝑡 𝑖𝑛 𝑡𝑖𝑚𝑒

- The returns to having a filled (f) and unfilled (V - vacancy) job are

Equation (5)

𝑟𝑉𝐹 (𝑡) = 𝑦 − 𝑤(𝑡) − 𝑐 + ′𝑉𝐹 (𝑡) − 𝜆(𝑉𝐹 (𝑡) − 𝑉𝑉 (𝑡))

Equation (6)

𝑟𝑉𝑉 (𝑡) = −𝑐 + ′𝑉𝑉 (𝑡) + 𝛼(𝑡)(𝑉𝐹 (𝑡) − 𝑉𝑉 (𝑡))

9.2 Wage determination

- The evolution of the number of workers is given by

Equation (7)

′𝐸(𝑡) = 𝑈(𝑡)1−𝛾 𝑉(𝑡)𝛾 − 𝜆𝐸(𝑡)

- Due to search friction, when firm and worker meet, the worker’s alternative to accept the
offer is to continue searching and the firm’s alternative to hiring the worker is to resume
searching for a worker
 - Hence, the firm and worker are better off if they fill the position than if he or she does not
- Therefore, the workers’ earn more than their reservation wage

9.3 Nash bargaining

- We assume wages are set by Nash Bargaining
- A fraction of the surplus created by the match, Ø (bargaining power) goes to the worker and
a fraction, 1-Ø goes to the firm:
→ Worker’s surplus from the match, 𝑉𝐸 (𝑡) − 𝑉𝑈 (𝑡)
→ firm’s surplus from the match, 𝑉𝐹 (𝑡) − 𝑉𝑉 (𝑡)
- The Nash bargaining assumption implies
Ø (1−Ø)
𝑚𝑎𝑥𝑤(𝑡) (𝑉𝐸 (𝑡) − 𝑉𝑈 (𝑡)) (𝑉𝐹 (𝑡) − 𝑉𝑉 (𝑡))

𝑚𝑎𝑥𝑤(𝑡) Ø𝑙𝑛(𝑉𝐸 (𝑡) − 𝑉𝑈 (𝑡)) + (1 − Ø)𝑙𝑛(𝑉𝐹 (𝑡) − 𝑉𝑉 (𝑡))

- The value of something is not exactly the same when you take logs vs. when you don’t. But
when you maximize it, you will get the same answer.
- FOC, differentiate wrt w(t) and let equal to zero
Ø 𝑑𝑉𝐸 (𝑡) (1 − Ø) 𝑑𝑉𝐹 (𝑡)
+ =0
(𝑉𝐸 (𝑡) − 𝑉𝑈 (𝑡)) 𝑑𝑤(𝑡) (𝑉𝐹 (𝑡) − 𝑉𝑉 (𝑡)) 𝑑𝑤(𝑡)

- As 𝑉𝑈 and 𝑉𝑉 do not depend on wages, we do not have to differentiate wrt to them, they are
a constant and will be 0.
𝑑𝑉𝐸 (𝑡) 1 𝑑𝑉𝐹 (𝑡) −1
- Now substituting for 𝑑𝑤(𝑡)
= (𝑟+𝜆) and 𝑑𝑤(𝑡)
= (𝑟+𝜆) in this equation to obtain

Ø 1 (1 − Ø) 1
− =0
𝑉𝐸 (𝑡) − 𝑉𝑈 (𝑡) (𝑟 + 𝜆) 𝑉𝐹 (𝑡) − 𝑉𝑉 (𝑡) (𝑟 + 𝜆)

9.3.1 Equilibrium conditions

- Hence, wages are determined by the equation

Ø (1 − Ø)
− =0
𝑉𝐸 (𝑡) − 𝑉𝑈 (𝑡) 𝑉𝐹 (𝑡) − 𝑉𝑉 (𝑡)

Ø(𝑉𝐹 (𝑡) − 𝑉𝑉 (𝑡)) = (1 − Ø)(𝑉𝐸 (𝑡) − 𝑉𝑈 (𝑡))

↔
Equation (8)
Ø
(𝑉 (𝑡) − 𝑉𝑉 (𝑡)) = 𝑉𝐸 (𝑡) − 𝑉𝑈 (𝑡)
(1 − Ø) 𝐹
- Free entry implies, new firms come into market until there is not more profits
- Here the workers will get the marginal productivity. But we have frictions and therefore we
get different results.

𝑉𝑉 (𝑡) = 0 𝑓𝑜𝑟 𝑎𝑙𝑙 𝑡

 - Initial employment is E(0) is given – 2,6 in Denmark

9.3.2 Steady state equilibrium

- We consider the steady state, that is where V(t)’s and ‘E(t) are zero. No growth in
employment or V.
- That is, we have (3) and (4) where we removed (t) and ‘V
𝑟𝑉𝐸 = 𝑤 − 𝜆(𝑉𝐸 − 𝑉𝑈 )
𝑟𝑉𝑈 = 𝑏 + 𝑎(𝑉𝐸 − 𝑉𝑈 )
- The returns to having a filled and unfilled job are (5) and (6) where we removed (t) and ‘V
𝑟𝑉𝐹 = 𝑦 − 𝑤 + 𝑐 − 𝜆(𝑉𝐹 − 𝑉𝑉 )
𝑟𝑉𝑉 = −𝑐 + 𝛼(𝑉𝐹 − 𝑉𝑉 )

- We subtract equation (4) from (3) to obtain

𝑟𝑉𝐸 − 𝑟𝑉𝑈 = 𝑤 − 𝜆(𝑉𝐸 − 𝑉𝑈 ) − 𝑏 − 𝑎(𝑉𝐸 − 𝑉𝑈 )
𝑟(𝑉𝐸 − 𝑉𝑈 ) = 𝑤 − 𝑏 − (𝑎 + 𝜆)(𝑉𝐸 − 𝑉𝑈 )
(𝑟 + 𝑎 + 𝜆)(𝑉𝐸 − 𝑉𝑈 ) = 𝑤 − 𝑏
𝑤−𝑏
(𝑉𝐸 − 𝑉𝑈 ) =
(𝑟 + 𝜆 + 𝑎)
- Next, we subtract (6) from (5) to obtain
𝑟(𝑉𝐹 − 𝑉𝑉 ) = 𝑦 − 𝑤 − 𝑐 − 𝜆(𝑉𝐹 − 𝑉𝑉 ) − (−𝑐 + 𝛼(𝑉𝐹 − 𝑉𝑉 ))
𝑟(𝑉𝐹 − 𝑉𝑉 ) = 𝑦 − 𝑤 − 𝑐 − 𝜆(𝑉𝐹 − 𝑉𝑉 ) + 𝑐 − 𝛼(𝑉𝐹 − 𝑉𝑉 ))
𝑟(𝑉𝐹 − 𝑉𝑉 ) = 𝑦 − 𝑤 − (𝜆 + 𝛼)(𝑉𝐹 − 𝑉𝑉 )
(𝑟 + 𝜆 + 𝛼)(𝑉𝐹 − 𝑉𝑉 ) = 𝑦 − 𝑤

Equation (9)
𝑦−𝑤
(𝑉𝐹 − 𝑉𝑉 ) =
(𝑟 + 𝜆 + 𝛼)
- We insert these two expressions into the Nash Bargaining solution for wages, (8) to obtain
Ø
(𝑉 − 𝑉𝑉 ) = 𝑉𝐸 − 𝑉𝑈
(1 − Ø) 𝐹
Ø 𝑦−𝑤 𝑤−𝑏
( )=
(1 − Ø) (𝑟 + 𝜆 + 𝛼) (𝑟 + 𝜆 + 𝑎)
Ø𝑦 − Ø𝑤 𝑤−𝑏
=
(1 − Ø)(𝑟 + 𝜆 + 𝛼) (𝑟 + 𝜆 + 𝑎)
Ø𝑦(𝑟 + 𝜆 + 𝑎) − Ø𝑤(𝑟 + 𝜆 + 𝑎) = 𝑤(1 − Ø)(𝑟 + 𝜆 + 𝛼) − 𝑏(1 − Ø)(𝑟 + 𝜆 + 𝛼)
Ø𝑦(𝑟 + 𝜆 + 𝑎) + 𝑏(1 − Ø)(𝑟 + 𝜆 + 𝛼) = 𝑤(1 − Ø)(𝑟 + 𝜆 + 𝛼) + Ø𝑤(𝑟 + 𝜆 + 𝑎)

- Add and subtract a term related to b on the left hand side

Ø(𝑦 − 𝑏)(𝑟 + 𝜆 + 𝑎) + 𝑏(𝛼(1 − Ø)𝑟 + 𝜆 + 𝛼) = (𝑟 + 𝜆 + ø𝑎 + (1 − ø)𝛼)𝑤

Ø(𝑦 − 𝑏)(𝑟 + 𝜆 + 𝑎) (𝛼(1 − Ø)𝑟 + 𝜆 + 𝛼)
+ 𝑏=𝑤
(𝑟 + 𝜆 + ø𝑎 + (1 − ø)𝛼) (𝑟 + 𝜆 + ø𝑎 + (1 − ø)𝛼)
 - We solve for wages to obtain
ø(𝑦 − 𝑏)(𝑟 + 𝜆 + 𝑎)
𝑤 =𝑏+
(𝑟 + 𝜆 + ø𝑎 + (1 − ø)𝛼)
- This is the wage equation!

Intuition

- Suppose 𝑎 = 𝛼 then

𝑤 = 𝑏 + ø(𝑦 − 𝑏)
- That is, a fraction, ø of the difference between output, y, and the value of leisure, b goes to
the worker and a fraction, 1-ø goes to the firm
- The higher ø is, the higher is the wage
- Suppose 𝑎 ≠ 𝛼, then the wage equation is
ø(𝑦 − 𝑏)(𝑟 + 𝜆 + 𝑎)
𝑤 =𝑏+
(𝑟 + 𝜆 + ø𝑎 + (1 − ø)𝛼)

Wages increase in workers’ job finding rate

- The wage equation is:

ø(𝑦 − 𝑏)(𝑟 + 𝜆 + 𝑎)
𝑤 =𝑏+
(𝑟 + 𝜆 + ø𝑎 + (1 − ø)𝛼)
- When the worker’s job finding rate a increases, then workers can find jobs more easily:
𝑑𝑤 (1 − ø)(𝑟 + 𝜆 + 𝛼)
= ø(𝑦 − 𝑏) >0
𝑑𝑎 (𝑟 + 𝜆 + ø𝑎 + (1 − ø)𝛼)2
- Wages increase when it becomes easier for worker to find a job as their bargaining position
therefore improves
- Finding rate for the firm 𝛼
- Impact of higher income when unemployed: higher b on wages
- Differentiate wrt income which workers will have if they are unemployed:
𝑑𝑤 ø(𝑟 + 𝜆 + 𝑎)
=1−
𝑑𝑏 (𝑟 + 𝜆 + ø𝑎 + (1 − ø)𝛼)
(𝑟 + 𝜆) + 𝛼
= (1 − ø) >0
(𝑟 + 𝜆 + ø𝑎 + (1 − ø)𝛼)
When unemployed workers receive a higher benefit, then they are better off, corresponding
to that their bargaining position improves and therefore their wages will increase.

Example, an increase in workers bargaining power could be due to that you have stronger
unions. History: during the last decades, a lot of unions in UK and US has been weakened →
led to reduction in wages for some workers. Higher bargaining power → higher wages.
Worker’s job finding rate

From the vacancy equation, (4) in the steady state we have

𝑟𝑉𝑉 = −𝑐 + 𝛼(𝑉𝐹 − 𝑉𝑉 )
Substitute for the difference between the value of a filled job and a vacant job, equation (9)
𝑦−𝑤
𝑟𝑉𝑉 = −𝑐 + 𝛼
(𝑟 + 𝜆 + 𝛼)
Substitute for wages
ø(𝑦 − 𝑏)(𝑟 + 𝜆 + 𝑎)
𝑦 − (𝑏 + )
(𝑟 + 𝜆 + ø𝑎 + (1 − ø)𝛼)
𝑟𝑉𝑉 = −𝑐 + 𝛼
(𝑟 + 𝜆 + 𝛼)

𝛼 ø(𝑦 − 𝑏)(𝑟 + 𝜆 + 𝑎)
𝑟𝑉𝑉 = −𝑐 + (𝑦 − (𝑏 + ))
(𝑟 + 𝜆 + 𝛼) (𝑟 + 𝜆 + ø𝑎 + (1 − ø)𝛼)

𝛼 (𝑦 − 𝑏)((𝑟 + 𝜆 + ø𝑎 + (1 − ø)𝛼) − ø(𝑦 − 𝑏)(𝑟 + 𝜆 + 𝑎))

𝑟𝑉𝑉 = −𝑐 + ( )
(𝑟 + 𝜆 + 𝛼) (𝑟 + 𝜆 + ø𝑎 + (1 − ø)𝛼)

𝛼(𝑦 − 𝑏) ((𝑟 + 𝜆)(1 − ø) + ø𝑎 + (1 − ø)𝛼) − ø𝑎

𝑟𝑉𝑉 = −𝑐 + ( )
(𝑟 + 𝜆 + 𝛼) (𝑟 + 𝜆 + ø𝑎 + (1 − ø)𝛼)
𝛼(𝑦 − 𝑏)(1 − ø) (𝑟 + 𝜆) + 𝛼
𝑟𝑉𝑉 = −𝑐 + ( )
(𝑟 + 𝜆 + 𝛼) (𝑟 + 𝜆 + ø𝑎 + (1 − ø)𝛼)
𝛼(𝑦 − 𝑏)(1 − ø)
𝑟𝑉𝑉 = −𝑐 +
(𝑟 + 𝜆 + ø𝑎 + (1 − ø)𝛼)
Equation (10)
𝛼(1 − ø)
𝑟𝑉𝑉 = −𝑐 + (𝑦 − 𝑏)
(𝑟 + 𝜆 + ø𝑎 + (1 − ø)𝛼)
- Value of having a vacancy is increasing in cost, and is dependent in all these terms
- Remember that a is the job finding rate, worker’s transition rate and 𝛼 is the firm transition
rate. Both worker’s job finding rate and firm’s transition rate are endogenous here
- If b increases, the worker’s wage will increase. Higher wages is more expensive for the firms
since they have to pay their workers more and therefore there will be a low value of having
a filled job, and hence it will be a low value of having a vacancy; can see that through this
equation: 𝑟𝑉𝑉 = −𝑐 + 𝛼(𝑉𝐹 − 𝑉𝑉 ). If 𝑉𝐹 is low, 𝑉𝑉 will be low due to this direct relationship.
𝑦−𝑤
𝑟𝑉𝑉 = −𝑐 + 𝛼 (𝑟+𝜆+𝛼). Also here we can see that anything that increases the wages will give
a lower value of the vacancy.
- If you are in a situation where wages increase. Suppose after the Corona crisis, the economy
is improving: workers will be better off, maybe higher productivity of the worker → improve
the wages. Then that will have a direct negative impact on how many jobs that will
default(?) because now it is more expensive for a firm to hire a worker, this will have a
negative impact on vacancies. On the other hand, since each worker will be more
productive, this will have a positive impact on having a filled job and therefore value of
vacancy and therefore potential firm.
 - From the equation determining the dynamics of employment, equation (7), we have in the
steady state, ‘E(t)=0

Equation (11)

0 = 𝑈(𝑡)1−𝛾 𝑉(𝑡)𝛾 − 𝜆𝐸(𝑡)

Equation (12)

𝜆𝐸 = 𝑈1−𝛾 𝑉 𝛾
- Furthermore, as the number of workers is normalised to one, we have E+U=1
- Use that the worker’s job finding rate is given by equation (1)

Equation (13)

𝑎 = 𝑈(𝑡)−𝛾 𝑉(𝑡)𝛾 = 𝜃 𝛾

Hence, in a steady state the number of matches, 𝑈1−𝛾 𝑉 𝛾 is equal to the number of separations, 𝜆𝐸:

0 = 𝑈1−𝛾 𝑉 𝛾 − 𝜆𝐸
𝑉𝛾
𝑈 = 𝜆𝐸
𝑈𝛾
𝑉 𝛾
𝑈 ( ) = 𝜆𝐸
𝑈
(1 − 𝐸)𝜃 𝛾 = 𝜆𝐸
𝑉
Use that E+U=1 → U=1-E and 𝑈 = 𝜃

9.4 Firm’s vacancy filling rate

The number of vacancies is then found by using equation (13) again. Assume the match efficiency
rate is 𝛾 = 0,5 (in the exam you will only be able to derive it with 0,5 for gamma), and remember
𝜆𝐸
that 𝑎 = 1−𝐸 and U=1-E

𝑎 = 𝑈 −0,5 𝑉 0,5
𝜆𝐸
𝑉 0,5 = 𝑎𝑈 0,5 = (1 − 𝐸)0,5
1−𝐸
𝜆𝐸
𝑉 0,5 = (1 − 𝐸)0,5
1−𝐸
𝜆𝐸 1/0,5
𝑉=( ) (1 − 𝐸)0,5/0,5
1−𝐸
𝜆𝐸 2
𝑉=( ) (1 − 𝐸)
1−𝐸
𝑉 = (𝜆𝐸)2 (1 − 𝐸)−2+1
𝑉 = (𝜆𝐸)2 (1 − 𝐸)−1
And the vacancy filling rate 𝛼 is found by using equation (2) to obtain

𝛼 = 𝑈 0,5 𝑉 −0,5 = (1 − 𝐸)0,5 ((𝜆𝐸)2 (1 − 𝐸))−0,5

𝛼 = (1 − 𝐸)0,5 ((𝜆𝐸)−1 (1 − 𝐸)0,5 )

𝜆𝐸 −1 1
𝛼 = (1 − 𝐸)1 𝜆𝐸 −1 = ( ) =
1−𝐸 𝑎
1
𝛼=
𝑎
1
𝑎=
𝛼
Note:

- The job finding rate is the number of matches, which depends on vacancies and unemployed
workers divided by the number of unemployed workers. The vacancy finding rate is instead
the number of matches divided by the number of vacancies. If it becomes easier for a
worker to find a job, that is, the job finding rate a increases, then it on the other hand
becomes more difficult for a firm to find a worker, that is falls.
- Whenever the workers’ job finding rate increases, the firms’ vacancy filling rate decreases.
The more jobs you have relative to unemployed workers, the more difficult it is to each firm
to find a worker.
- Workers’ job finding rate, a; increases in the number of jobs, E
- Firms’ vacancy filling rate, 𝛼 decreases in the number of job E
- As E approaches 1 (no unemployment) a approaches infinity and 𝛼 approaches 0 and 𝑟𝑉𝑉
approaches -c. The cost of having a vacancy is -c, because there are no expected return to
open a vacancy.
- As E approaches 0 (no employment – no jobs for the workers) a approaches 0 and 𝛼
approaches infinity and 𝑟𝑉𝑉 approaches y-(b + c) (to see this rewrite (10))
- The intercept with the x-axis is where 𝑟𝑉𝑉 = 0. To the left of this there will be more and
more firms coming into the market and there is a surplus, and when it reaches 0, there are
no further jobs supplied in this market. This the steady state equilibrium in the economy.
9.4.1 Equilibrium employment level
Free entry implies that there will be firms entering until the value of supplying a vacancy is zero:

𝑟𝑉𝑉 = 0
Using this in equation (10) gives the following equation to determine E
𝛼(1 − ø)
𝑟𝑉𝑉 = −𝑐 + (𝑦 − 𝑏)
(𝑟 + 𝜆 + ø𝑎 + (1 − ø)𝛼)
Let ø=0,5
𝛼0,5
𝑟𝑉𝑉 = −𝑐 + (𝑦 − 𝑏)
(𝑟 + 𝜆 + 0,5𝑎 + 0,5𝛼)
Equation (14)
𝛼
𝑟𝑉𝑉 = −𝑐 + (𝑦 − 𝑏)
(2(𝑟 + 𝜆) + 𝑎 + 𝛼)

Let 𝛾 = 0,5 whereby a and 𝛼 are

Equation (15)
𝜆𝐸 1−𝐸
𝑎= , 𝛼 = 1/𝑎 =
1−𝐸 𝜆𝐸
Equation (14) and (15) determine a, 𝛼 and E

9.4.2 Equilibrium employment and job finding rate

For equal bargaining power, 𝜙 = 0,5 and equal match efficiency, 𝛾 = 0,5 the following two
equations can be used to determine employment E and the job finding rate a:

THESE TWO EQUATIONS ARE IMPORTANT AND RELEVANT FOR THE EXAM

Equation (16)
1/𝑎
𝑐= (𝑦 − 𝑏)
(2(𝑟 + 𝜆) + 𝑎 + 𝛼)
Equation (17)
𝜆𝐸
𝑎=
1−𝐸
Next, 𝛼 is found as 𝛼 = 1/𝑎

Impact with respect to lambda!

The left-hand side describe the cost of a vacancy, the right-hand side the benefit. Lambda is in the
denominator, meaning that when lambda increases, the benefit of posting a vacancy decreases.
Why? With a high lambda, even if the firm manages to fill the vacancy and produce something, there
is a high risk that in the period afterwards, the job is going to be destroyed (at a rate lambda).

What does this mean for the job finding rate, a? If lambda is high, firms are going to post fewer
vacancies because the cost seems too high given the uncertain benefit. a is falling.

About the last sentence in the question: y is always larger than b, otherwise no jobs would ever be
posted. If unemployment benefits are higher than the output of what the worker could produce
when working, no worker would be willing to accept a job offer. Because firms know this, they would
never post any vacancies.

- Coronavirus crisis: could be illustrated by change in 𝜆. It is a high job separation.

You explain that if lambda is low (like in Italy), then firms will want to post fewer vacancies and your
explanation makes sense. However, looking at the formula for the benefits of hiring a worker, it
seems like benefits increase with a smaller lambda. This would then mean that in countries like Italy,
firms would want to post more vacancies, as the benefits are higher?

The reason is that the model doesn't capture on crucial element that describes reality, namely firing.
(Remember that lambda is an exogenous probability that a job just evaporates, regardless of
whether the worker or the firm would like to continue the match.)

In reality, a firm might want to fire a person because they realize he or she is not good at the job. If
they can't to that because of a law, they will think twice about hiring someone in the first place. Both
aspects that I just described (the fact that different workers have different productivities, and the
endogenous wish to fire someone) do not exist in the world of this model (at least not in the version
we study here).
9.3 Comparative statics, output increases
We look at what happens if there is a change in one of the parameter values in the steady state, b, Y,
Lambda, phi.

Coronavirus: there is a change in the economy that has an impact on employment. We look at this
through a change in the job finding rate. Whenever employment falls, unemployment increases. We
have to take into account that it also has an impact on wages.

Suppose there is a positive shock to output (the economy is going into a boom)

When we differentiate an equation wrt. Y, it is about a change in productivity in the economy,

suddenly everyone becomes more productive. Economy goes from recession to boom. The way we
approach this is that we will differentiate these two equations with respect to y, and we have to
acknowledge that there is both going to be an impact on the workers job finding rate, and by that an
impact on employment.

Hopefully after Corona: increase in productivity, in y, what would be the impact on employment and
unemployment, hopefully what we will see is that; a high y will lead to more jobs in the economy,
because right now see a huge unemployment rate in the economy.

Rewrite equation (16) and (17) with respect to y; E and a:

1
𝑐(2(𝑟 + 𝜆) + 𝑎 + 1/𝑎) = (𝑦 − 𝑏)
𝑎
𝑎(1 − 𝐸) = 𝜆𝐸
Remember that a is affected, because that is endogenous, that is the only thing, because the other
are parameter values and exogenous.

Then differentiate wrt a and then Y

1 1 1
𝑐 (1 − 2
) 𝑑𝑎 = − 2 (𝑦 − 𝑏)𝑑𝑎 + 𝑑𝑦
𝑎 𝑎 𝑎
(1 − 𝐸)𝑑𝑎 − 𝑎𝑑𝐸 = 𝜆𝑑𝐸

𝑑𝑎
(1 − 𝐸)𝑑𝑦 − 𝑎𝑑𝐸 = 𝜆𝑑𝐸
𝑑𝑦
Solve for the change in e and a with respect to the change in y. da/dy is the change in the jobfinding
rate with respect to productivity. We just timed with dy in the nominator and the denominator for it
to look like that.

Similar question like this on the exam.

Equal to
1 1 1
𝑐 (1 − 2
) 𝑑𝑎 = − 2 (𝑦 − 𝑏)𝑑𝑎 + 𝑑𝑦
𝑎 𝑎 𝑎
1 1 1
(𝑐 (1 − 2
) + 2 (𝑦 − 𝑏)) 𝑑𝑎 = + 𝑑𝑦
𝑎 𝑎 𝑎
 𝑑𝑎 1 1
=
𝑑𝑦 𝑎 1 1
(𝑐 (1 − ) + 2 (𝑦 − 𝑏))
𝑎2 𝑎

Since we not yet can see if the denominator is positive, we have to simplify it a bit more.
𝑑𝑎 1 1
= >0
𝑑𝑦 𝑎 1
(𝑐 + 2 (𝑦 − 𝑏 − 𝑐))
𝑎

1 1
( 2
(𝑦 − 𝑏 − 𝑐) + 𝑐) 𝑑𝑎 = 𝑑𝑦
𝑎 𝑎
Now, we can see that the cost of opening vacancy is smaller than -b, so if this is positive (which we
assume), the expressions is positive. Higher productivity of worker → more jobs formed → worker’s
transition rate is going to increase
𝑑𝑎
(1 − 𝐸)𝑑𝑦 − 𝑎𝑑𝐸 = 𝜆𝑑𝐸
𝑑𝑦
𝑑𝑎
(1 − 𝐸)𝑑𝑦 = (𝜆 + 𝑎)𝑑𝐸
𝑑𝑦
𝑑𝑎 (𝜆 + 𝑎)𝑑𝐸
=
𝑑𝑦 (1 − 𝐸)𝑑𝑦
𝑑𝑎(1 − 𝐸) 𝑑𝐸
=
𝑑𝑦(𝜆 + 𝑎) 𝑑𝑦
𝑑𝐸 (1 − 𝐸) 𝑑𝑎
= >0
𝑑𝑦 (𝑎 + 𝜆) 𝑑𝑦
hence employment and workers’ job finding rate increases when productivity, y, increases. This is
the case as firm’s profits rise associated with employing a worker. Hence employment increases and
unemployment fall in a boom. Similarly, we could consider the impact of a higher value of leisure, b
or higher costs of maintaining a job, c.

Do the 2 exercises for the exam. This is what you need to know for the exam!

Implications of the model

Search and matching models offer a straightforward explanation for average unemployment: it may
be the result of continually matching workers and jobs in a complex and changing economy. Thus,
much of observed unemployment may reflect what is traditionally known as frictional
unemployment.

Labor markets are characterized by high rates of turnover. In U.S. manufacturing, for example, more
than 3 percent of workers leave their jobs in a typical month. Moreover, many job changes are
associated with wage increases, particularly for young workers; thus at least some of the turnover
appears to be useful. In addition, there is high turnover of jobs themselves. In U.S. manufacturing, at
least 10 percent of existing jobs disappear each year. These statistics suggest that a nonnegligible
portion of unemployment is a largely inevitable result of the dynamics of the economy and the
complexities of the labor market.
The two perspectives are not just mirror images of one another. For example, suppose the rate of
job creation is constant over the business cycle and the rate of job destruction is countercyclical.
Then on the worker side, both margins are cyclical: the rate of inflows rises in recessions because of
the increase in the rate of job destruction, and the rate of outflow falls because of the increase in
the number of unemployed workers and the constant rate of job creation.

One conclusion of this literature is that answering seemingly simple questions about the
contributions of different margins to changes in unemployment is surprisingly hard. The details of
the sample period, the precise measures used, and subtleties of the data can have large impacts on
the results. To the extent that this work has reached firmer conclusions, it is that, from either the
firm or the worker perspective, both margins are important to changes in overall unemployment.

The impact of a shift in the labor demand

Specifically, we are interested in whether it causes a shift in labor demand to have a larger impact on
employment and a smaller impact on the wage than it does in a Walrasian market

Review session about the exam

Overview of the whole course in the end.

Unemployment is quite different. It WILL COME IN THE EXAM!

Unemployment questins:

Model with friction, it takes time for workers to find a job, and jobs for find workers.

Set up the matching function and derive an equilibrium

Reduce it down to 2 equations:

𝜆𝐸
𝑎=
1−𝐸
And
1
𝑐(2(𝑟 + 𝜆) + 𝑎 + 1/𝑎) = (𝑦 − 𝑏)
𝑎

Corona: a huge negative shock, number of vacaces fell relative to unemployed people. A is increasing
in the number in vacanes. A will fall since vanaces fall.
Negative shock: fall in Y – fall in productivity (simple) – due to less demand (not really corona) but
fine to think this

More realistic: expand the model → empirical research on this – in class do not doiscuss this

Decrease in Y is what happens atm. Differentaite wrt y , and a and e. When a changes, e changes,
and Y will have an impact on a. 2 endogenous variables are not possible.

Alpha also is equal to 1/a.

A is how easy it is for a worker to find a job, and hard for firms to find a worker due to competition.

Substitute for alpha in the equations.

What happens to de when there is a change in Y.

Should we be able to find E in equilibrium? No, just as in the exercises. Not going to be more
complicated, but 3 hours in stead of 4 hours. Not a lot of derivation in the exam.

Diamond model

Obligatory assignment – relevant for the exam!!

The intuition is the most important!

Mette Frederiksen/Donald – we can do better. You need common ground, have a planner, prime
minister can do better. Ramsey model – no externalities. In diamond, we have 2 types of people –
transfer from 1 group to group 2, in order to do better. Old and young.

When? If we were in a situation where we were below the level of capital with golden rule level of
capital (max consumption) – save more – more capital – more to next – BUT then young would be
consume less – not happy.

Everyone better of: above the golden rule level of capital. Where we have saved too much, because
it is too much capital in the economy. Save less – less capital in next period.

1) Can we make young people better? Consume more/less? Have to be more

Above Golden rule – everyone better off: save less! Consume more – buy stuff – happy people as
young

Save less: when we are old less capital.

Interest rate is decreasing the capital stock. High caooiutal stock will reduce return. Dynamic
inefficiency, too much capital accumukated. Return to caouitl when we are old will be too small.In
this situation, it is better to ask people to save less. When old: get a high return!

Why depend on n? we have to have a lot of people in the economy. When we get old, a lot of people
around. The young can pay to us when we are old.
 2) ?

Older bigger than young generation: young have to work a lot to give to old.

Optimal to have a pay as you go system if high n. If everyone pays for themselves – pension. In
Denmark, low n.

Why it is different to Ramsey, 2 people.

Dynamic inefficiency: externalities: climate changes, world is going to turn to. Just for fun, not
syllabus.

People live for 30 years. n>2 is fine. It is like adopting a population in about 30 years.

Relative risk aversion:

Utility function.

Theta towards 0 - Do not care what you consume in period 1 and 2.

Corona crisis: don’t consume now – consume a lot next year – ok I don’t care., becaue theta is 0.

Richer people relatively worse off compared to poor people, when theta is bigger than 0.

𝐶(𝑡)1−𝜃
𝑢(𝐶(𝑡)) = ,𝜃 > 0
1−𝜃
Same elasticity: – consumption times the second order derivative, divided by the first order
derivative.

Growth rate in the growth rate

Corona: Growth rate will fall. Negative growth for this year. 3%-10% estimated for Denmark.
Because we are very liable of other countries. 50% import. Depends on the how small reduction they
will have in growth rate in GDP. Everyone agrees negative GDP growth. Growth in growth rate is the
change in the growth rate in the economy.

Macro economics super important now!

Most extreme in 4 decades.

Problem in China: tariffs increased to China from USA – tough situation – now it is tougher. BUT they
have had huge growth rate and growth in the growth rate, at some point it had to change. Growth
rate in growth rate negative. Growth poisutuve,. They are relative poor compared to other
economies.
Some more questions

Important that we study the material this weekend.

No live chat, but we can ask questions.

Alpha is always positive

Make sure you answer the question, and proof that you have understood!

You are suppose to write: Now I am suppose to do this and that, and give a short intuition. (SHORT!)

On the other hand, for steady state solutions and equilibrium will we have the equations in the exam
similar as we had in exercises? Yes, exactly!

But we shouldn't derive the steady state solutions unless you ask for it? No just do what you are
suppose to.

3 broad question with sub question in the exam.

Cross country differences:

Decompositng output. The reason for why we divide by L(t) is that we want to see how it is per capita.

Review session!

Solow model – nothing can determine the growth rate in this economy. No heterogeneity. No
efficiency, no dynamic inefficiency. No externalities. No consumers, no firms.

Micro foundation: agents! Consumers and firms optimizing/maximizing. Ramsey and diamond model
Diamond model – heterogeneity, efficiency, can be better, deriving golden rule

Endogenous growth model. Once countries are in a recession, what can they do to get out?
Learning by doing – why? Because previously we assume that a fraction of the people work in the
knowledge production. IRL it is not exogenous. Vaccine production – knowledge production. It is a
trick, it is a biproduct of capital.

Differences between countries – use what we have learned to understand difference between countries
in terms of growth and GDP per capita. A lot of other factors, culture, temperature.
Human capital – H: education increase H, GDP per capita increase. In poor countries – education is
lower – H is lower – (growth is lower) – GDP per capita lower

Unemployment – already talked about