Economic Growth: Solow Model

-References: There are many good presentations of the Solow Model. I have

posted three on the website.

G. McCandless The ABCs of RBCs Ch. 1;

C. Jones and D. Vollrath Introduction to Economic Growth Ch. 2;

M. Wickens Macroeconomic Theory Chapter 3

- Standards of living, size of capital stock are typically growing in the long-run.

- it is desirable to have models with long-run growth as an outcome.

- long-run growth: is it more important to well-being than cycles?

- ideally: want a macroeconomic model allowing for both long-run growth

and cyclical deviations from the long-run growth path.

(this is one aim of Real Business Cycle models)

- here we introduce a basic growth model.

- Aggregate production function: Yt = F(Kt, Nt, t)

Y = output, N=labour, K = capital, t - time

(Notational point : lower case will mean per unit labour or per worker

e.g. y = Y/N, k=K/N)

- Why might growth occur?

- accumulation: more capital and other non-labour inputs.

- labour force growth (skills too? or are they like capital?)

- technological and organization progress (“t” in the production

function might capture this).

- Growth accounting exercises take this as a starting point: choose a specific functional form for F(), obtain measures of Y, K, N; infer effect of ‘t’.

Solow Model

- Sometimes called the Solow-Swan model.

- Most widely used growth model in economics.

- Robert Solow (1956) "A Contribution to the Theory of Economic Growth".

Quarterly Journal of Economics 70 (1): 65–94.

“The Solow model is a mixture of an old-style Keynesian model and a modern dynamic macroeconomic model” D. Acemoglu.

- Let’s start with a basic version of the Solow model:

- Basic? no technological change.

- Often presented this way.

- Can allow for a quite general production function.

- Model has an interesting equilibrium outcome.


- Aggregate Production Function:

Yt = F(Kt, Nt)

- Economy-wide production function (earlier notes: firm level production)

- Form of F(): reflects technology, organization, incentives: all determine how productive of inputs.

- Other inputs? e.g. natural resources; assume fixed in this version: built into F; another possibility: part of K.

- Make our usual technical assumptions about the production function:

Positive marginal products: ∂F/∂L≡FN>0, ∂F/∂K≡FK>0

Diminishing returns: ∂2F/∂K2≡FKK and ∂2F/∂N2≡FNN<0

- Also assume “constant returns to scale”

i.e. raise all inputs by the same proportion and output rises by that

proportion.

So where z is some constant: z∙Yt = F(z∙Kt, z∙Nt)

Let: z=1/Nt

Then the production function can be written in per worker terms:

Yt/Nt = F(Kt/Nt ,1)

or: yt = F(kt,1) (y=Y/N, k=K/N)

notice : ∂y/∂k = FK

- Lastly assume the “Inada conditions” hold:

lim FK = 0 lim FK = ∞

k→∞ k→0

- Consequences of technical assumptions:

Inada conditions will push outcomes away from extremes

(0 or infinite k): convenient!

Are they true? No empirical evidence! (don’t see these extremes)

Diminishing returns: return to k investment declines in level of k.

- Plausibility? Two stories:

- high ‘k’ then each unit of K has little N to work with

and is therefore less productive.

-aggregate production: initial K goes to highest return activities, later investment in K to remaining highest return activity, etc.

- Diminishing returns places a limit on growth in this model.

(See: ‘AK Model’ to see what happens without

diminishing returns)

- R. Allen Global Economic History : is diminishing returns empirically

sensible at the aggregate level?

an aggregate production function of this form might be a good approximation across countries and time (see his Figures 8-11).

(see next page)

- A production function that satisfies the technical assumption?

Cobb-Douglas with 0<a<1, A >0

Yt= AKtaNt1-a

per worker form (divide by N):

yt= Akta FK=aAkta-1 >0

FKK=(a-1)aAkta-2<0

(Inada conditions hold too)

- Labour force growth: Nt=N0(1+n)t (N0 - initial value of N in

period 0)

i.e. labour grows at a constant per period rate of “n”.

if you prefer: Nt+1 = Nt (1+n)

(model assumes full employment: labour force and number

employed are the same)

- National income constraint (economy-wide budget) :

Yt = Ct + It C= consumption, I=investment

- Investment and saving:

- Closed economy: investment is financed with domestic savings (S):

It=St

- Savings function: St = s∙Yt ( Ct=(1-s)Yt )

- total savings (S) in time t are a fixed proportion (s) of Y.

- “s” is the savings rate (a constant in Solow)

- No “microfoundations” for this: ‘s’ is assumed constant. ‘s’ is not derived from intertemporal optimization.

(an old-style macroeconomic assumption)


- Capital growth: (DKt+1 = Kt+1-Kt)

- Capital accumulation: DKt+1 = It - dKt,

d = depreciation rate (share of K that wears out each period)

use St = s∙Yt and St=It :

DKt+1 = s∙Yt - dKt

divide by K to express it as the growth rate in K:

DKt+1/Kt = s∙Yt/Kt - d

- Is k=K/N growing, shrinking or staying the same?

Growing if: DKt+1/Kt > DNt+1/Nt (= n)

or: DKt+1/Kt - DNt+1/Nt > 0

s∙Yt/Kt - d - n >0

multiply last condition through by Kt/Nt:

s∙Yt/Nt - (d + n)∙Kt/Nt >0

or (using definitions of y and k):

s∙yt - (d + n)∙kt >0

- So: K/N is growing if: s∙yt - (d + n)∙kt >0

K/N is shrinking if: s∙yt - (d + n)∙kt <0

K/N is constant if: s∙yt - (d + n)∙kt =0

- Look more closely at these conditions:

syt = amount of savings and investment (expressed per N)

i.e. amount of new capital being created (per N).

(d + n)∙kt = amount of new capital needed to hold K/N constant (expressed

per N)

i.e. must replace depreciation: d∙kt

must equip new workers with same level of K as old workers: n∙kt

So:

k shrinks if new capital is insufficient to replace depreciation and equip new workers.

k grows if new capital is greater than what is needed to replace

depreciation and equip new workers.

k constant if new capital and needed capital are the same.


- In diagrams: remember y=F(k,1)

- If:

k<kss then s∙yt - (d + n)∙kt >0 ↑k, ↑y

k>kss then s∙yt - (d + n)∙kt <0 ↓k, ↓y

k=kss then s∙yt - (d + n)∙kt =0 k, y unchanging.

- Long-run outcome: kss, yss

- k and y are constant in this “steady state” equilibrium.

(consumption per worker: c=(1-s)y is also constant)

- Notice: Y, K, C and N are all growing at rate “n”.

- So economy is growing in this model.

( but the standard of living measures y, c are not!)

- Another way to solve the model? Use the underlying difference equation in k

- Take: DKt+1 = It - dKt, with DKt+1=Kt+1-Kt

Kt+1= It + (1-d)Kt,

- Divide by Nt:

Kt+1Nt=ItNt+(1-δ)KtNt

- Use Nt+1=Nt(1+n) to write:

Kt+1Nt+1=ItNt+1-δKtNt/(1+n)

- Write in per N terms:

kt+1= it+(1-δ)kt1+n

- Substitute: it=syt (with yt=F(kt,1) )

kt+1= sF(kt,1)+(1-δ)kt1+n

- This is a first order non-linear difference equation in k.

- Steady state equilibrium? kt=kt+1=k*

k*= sF(k*,1)+(1-δ)k*1+n

rearrange: s F(k*,1) - (d+n)k* = 0

i.e. same equilibrium condition as above.

- Is it stable? depends on slope of the difference equation at k*.

dkt+1dkt=sFK+(1-δ)1+n<1 so it is stable.

(Inada conditions tell you slope is infinite at k=0, as k→∞ slope

approaches (1-d)/(1+n) <1, and you have FKK<0 so the difference equation will intersect the 45-degree line with

slope<1)

- Diagram: start at klow and the economy moves to k*

(same is true if start with k>k*)


- Another way of looking at dynamics:

- a common technique is to use a linear approximation to assess stability

near the equilibrium.

- Here the production function is the source of non-linearity in:

kt+1= sF(kt,1)+(1-δ)kt1+n

- Replace the production function with a first-order Taylor series

approximation evaluated at k*:

F(kt,1) ≅ F(k*,1) + FK(k*,1) (kt-k*)

this gives:

kt+1≅ sFk*,1+FKk*,1kt-k*+1-δkt1+n

or:

kt+1≅sFk*,1-sFKk*,1k*1+n+sFKk*,1+1-δ1+n kt

This linear difference equation is stable if:

sFKk*,1+1-δ1+n<1

or if: sFK(k*,1) < n+d (as in the diagram above)


- Comparative statics in the Solow model? How does the steady state change?

- Exogenous parameters: s, n, d and any parameters of F(k,1).

- Rise (fall) in s: k and y both rise (fall).

- Rise (fall) in n or d: k and y both fall (rise).

(picture: (n+d)k pivots up for a rise or down for a fall)

- Shift up (down) in F: k and y both rise (fall).

(picture: rise in F is like a rise in s except y=F(k,1)

shifts up as well)

- Why might one country have a high level of y and another a low y?

- Steady-state differences:

- If both are in a steady state higher y could be the result of high s, low

n and/or high F (high productivity).

- Disequilibrium differences:

- Another possibility? same s, n and F but maybe low y country is

further from the steady state.

(is this case convergence will occur over time)


From: Jones and Vollrath Introduction to Economic Growth (y vs. s, y vs. n)

- Consumption per worker (c=C/N): ct= yt-syt with yt=F(kt,1)

- Remember that y is not the same as c.

- What is the socially best level for “s”?

- Golden rule: choose s to maximize steady state c.

c = F(kt,1) – s y for a steady state: sy=(d+n)k

= F(kt,1)-(d+n)kt

so in picture maximize distance between F(k,1) and (d+n)k.

f.o.c. (max c with respect to k):

FK-(d+n)=0 choose s that gives a steady state at this

point (see diagram).

- Is the Golden rule savings rate what policy makers should aim

for? (see later discussion in Ramsey model)

- Open economy and Solow:

- Solow model is a closed economy model.

- Differences in ‘y’ are rooted in differences in s, n and F(K,N).

- If economies are open will differences in s, n and F be smaller?

- will savings flow from high to low k countries to take advantage of

higher FK?

- will workers migrate from poor (high n) countries to rich countries

(low n)?

- will diffusion of technology reduces inter-country differences in F?

- Is convergence between rich and poor countries more likely if economies are open?


Solow Model with Exogneous Technological Change:

- Add exogenous technological change.

- Production function shifts over time in some specified way.

- Exogenous: rate of technological progress is independent of what is

happening in the economy.

- Desirable addition since technological change happens!

- also in basic Solow model steady state y does not rise over time.

(living standards seem to grow with time)

- Possible ways of introducing technological change:

- A common approach:

- define ek and en as “efficiency” units of K and N respectively.

- efficiency units: adjust for productivity of K and N.

- so the effective amount of each input is:

ekK and enN

- inputs grow because there are more units of input or because a

given unit of input is more “efficient”.

- technological change is assumed to drive changes in ek, en. in

this presentation.

(note: en could reflect skill levels of workers)

- production function is :

Y = F(ekK,enN)

- Most common approach assumes technology works through en (let

ek=1): “labour-augmenting” technological progress.

- advantage: has a steady state much like that of the basic

model (see the note below)

- Wickens assumes “neutral” technological progress (as if ek=en and

both rise by same proportion as technology improves).

So let: ek=en=e

Then: Y = F(ekK,enN)

Y = e F(K,N) due to constant returns

- Wickens and McCandless have “e” rise at an exogenous rate: m

- So: Yt = (1+m)t F(Kt,Nt) so et = (1+m)t

- If they also assume a Cobb-Douglas production function for F:

- So: Yt = (1+m)t KtaNt1-a

- In per N form: yt = (1+m)t kta

- The Cobb-Douglas does have a steady-state with this type of

technological change

- can treat neutral progress like the labour-augmenting case.

(not all constant returns production functions do).

- Growth rate in output per worker (y), call it gy ?

Output per N: yt = (1+m)t kta

then: yt+1 /yt = (1+m) (kt+1/kt)a

(1+gy) = (1+m) (1+g(kt))a (g(k) is the growth rate of k)

take logs: ln(1+gy) = ln(1+m) + a ln(1+g(kt))

or using the approximation ln(1+x) @ x for small x:

gy = m + a g(kt)

- What determines the growth rate in k (i.e. g(k)) ?

- Savings, capital accumulation and labour force growth are the same as

earlier.

- So as before: DKt+1/Kt = s∙Yt/Kt - d = s∙yt/kt - d

and: DNt+1/Nt = n

so the growth rate in K/N=k is approximately:

g(kt)= Dkt+1/kt = s∙ yt/kt - d - n