Unit : Economic Growth

Lecture

February
 Outline

LR growth

The Solow model

The General Solow model

Policy analysis

In closing
 Objectives/outcomes for U

* Use Solow models of economic growth to analyse the determinants of di ferences in output levels and
output growth rates across time and space.
* Describe economic convergence
* Analyse the e fects of changes in the savings rate on economic welfare
* Contrast the assumptions and predictions of endogenous and exogenous growth models
* Use macroeconomic data to perform a growth accounting exercise
* Use macroeconomic data to analyse cross-country growth
 Literature

Sorensen & Whitta-Jacobsen (SW), Ch. , ,

/ LR growth
 LR growth

* By LR growth economists mean the average growth rate over decades/centuries

* NOT the quarterly/yearly luctuations in economic growth
* From about onwards, income per capita in most Western Economies has grown by about % p.a.
 Causes of LR growth

* We use models of LR growth to explain why income levels and growth rates di fer vastly between countries
over long periods of time
* Important sources of growth are capital accumulation (investment) and technological progress
* The so-called proximate sources of LR growth
* But what then causes capital accumulation and technological progress ...
* In our models, the only sustained source of growth in income per worker is technological progress
 Income levels and growth rates

* Very large (and sustained) variation in per capita incomes around the world
* Growth rates vary substantially between countries
* Growth rates are not constant over time OR space
* Countries can therefore become ”richer” or poorer” over time ...
 Stylized facts (Kaldor, s)

* Kaldor observed the following about Western Capitalist economies in the s:

* LR average growth rate of income/worker (or income/capita) was about % p.a. (from onwards)
* The return to capital (r) is relatively constant over time
* Labour's share of national income ( wL Y
) and capital's share of national income ( rKY ) were relatively constant over time
* These imply that the capital labour ratio (k = K/L) and wages (w) grow at the same rate as income per worker (y = Y/L),
and that the capital-output ratio (z = ky = KY ) is constant
* Models of economic growth were developed to explain these facts
* If an economy exhibits these facts, it is said to be growing along its balanced growth path (or is in its
steady-state)
* New data are leading to the questioning of these facts ... see e.g. Eggerston et al.
 The balanced growth path

* Along the balanced growth path:

* Output per worker (y), capital per worker (k), consumption per worker (c) and wages (w) grow at the same rate, g
* Output (Y), capital (K) and consumption (C) grow at the same rate, (1 + g)(1 + n), where n is the
population/labour force growth rate
* This rate is approximately equal to g + n if g and n are relatively small
* Also, if y = Y/L, then Y = yL. If y grows at g and L grows at n, then g(Y) = g(y) + g(L) = g + n (use logs...). Similar
reasoning holds for g(K) and g(C)...
* The return to capital (r), the real interest rate (ρ = r − δ), the capital-output ratio (z) and labour's & capital's shares
of national income (wL/Y and rK/Y) are constant
 Calculating growth rates

Y
* Let y = L
ln(yT )−ln(yt )
* To obtain the annual average growth rate between periods T and t, let gy = T−t
* Note: the growth rate between two successive years (for which T − t = 1) is gy = ln(yT ) − ln(yt )
* The power of compounding:
* Suppose that income per worker grows continuously, then yT = yt eg(T−t)
* It can then be shown that the time that it takes for y to double between t and T is ln(2)/g = 0.693
* The rule of seventy then states that the time that it takes for y to double is 0.7
g
70
= 100g
* Note: g is a proportional growth rate, while 100g is a percentage growth rate
* Through the power of compounding, observe what a di ference an annual growth rate of % vs. % makes ...
 Continuous v. discrete time

* Discrete time: time progresses in discrete jumps ...

* Months, quarters, years
* Continuous time: time progresses along a continuum, not in jumps...
* While economic activity is usually measured at discretely, it progresses continuously.
* Analytically, easier to work with continuous time
* But for simulations, regressions, etc., have to work with discrete time
* Take note of the following:
* Discrete time: ∆y = yt − yt−1 and g = ∆y
yt−1
dy ẏ
* Continuous time: ẏ = dt
and g = y
/ The Solow model
 The (Basic) Solow model (SW, . )

* Agents: consumers, firms, government

* One output = GDP = Y and two inputs, capital (K) and labour (L)
* Markets for output and inputs (labour and capital) are perfectly competitive
* P = MC, w = MPL, r = MPK
* Real interest rate is ρ = r − δ, wher 0 ≤ δ ≤ 1 is the depreciation rate
 The Solow model: production

* Firms use capital and labour to produce output which they sell in perfectly competitive markets
* Production is given by Y = f(K, L)
* Profits are Π = Y − rK − wL = f(K, L) − rK − wL
* Firm's objective = maximise profit
* This production function exhibits constant returns to scale (Y = f(λK, λL) = λY) and exhibits decreasing
returns to capital and decreasing returns to labour
* MPL = ∂Y∂L
= fL , while MPK = ∂Y
∂K
= fK
* fL = w, fK = r
* fLL < 0, fKK < 0
* A production function that meets the above criteria is the Cobb-Douglas production function:
Y = BKα L1−α , where B = total factor productivity (TFP), and α = share of capital in national income
* In the 'basic’ Solow model (ch. ), output at time t is Yt = BKα
t Lt
1−α
... note that B is constant...
 The Solow model: consumers

* Consumers own all capital, which they lease to firms

* Consumers supply labour inelastically
* Consumers choose consumption and saving to maximise utility, subject to an intertemporal budget
constraint: St−1 − δKt−1 = Kt − Kt−1
* Assume that the consumers maximises utility, subject to the budget constraint, by saving a constant fraction
of his/her income so that S = sY, 0 ≤ s ≤ 1
* Lastly, number of consumers/workers increase a rate n, so that Lt = (1 + n)Lt−1
* In this model, I = S ⇒ I = sY
 Basic Solow model

* Shows how LR evolution of y depends on structural model parameters like s and n

* Is a model of capital accumulation: shows how changes in s and n lead to changes in k, which lead to
changes in y
* While changes in s and n can lead to changes in rate at which y grows for very long periods, they don't lead to
permanent changes in the rate at which y grows...
* In the basic Solow model, Bt = B is constant
* This means that in the long-run, in the basic Solow model, the growth rate of income per worker =
* So the basic Solow model of LR growth does not explain positive LR growth (must be zero, according to model)!!
/ The General Solow model
 The general Solow model ( . )

* For sustained positive LR growth in Solow model, we need sustained technological progress
* Recall that Y = BKα L1−α , where B is TFP (”technology”)
1
* Now suppose that A = B 1−α and At = (1 + g)At−1
* g = rate of technological progress ... exogenous, though ...
1−α
* Then write production function as Y = Kα (AL)
* This type of technology = labour-augmenting (Harrod-neutral)
* If Y = f(AK, L), the type of technology is capital-augmenting (Solow-neutral)
* If Y = Af(K, L), then the type of technology is Hicks-neutral technology
 General Solow

* Define per worker variables as y = Y/L and k = K/L

* Then, if Y = Kα (AL)1−α : y = kα A1−α
* Take logs: ln(y) = α ln(k) + (1 − α) ln(A)
* Di ferentiate w.r.t. t (or subtract lagged logs): gy = αgk + (1 − α)gA
* The growth rate of income per worker is the weighted average of the growth rates of capital per worker and
technology
 The complete general Solow model

* Discrete time (p. ): * Continuous time (Exercise , p. )

* Yt = Kα
t (At Lt )
1−α
* Y = Kα (AL)1−α
* St = sYt * S = sY
* Kt − Kt−1 = St−1 − δKt−1 * K̇ = sY − δK
* Lt = (1 + n)Lt−1 * n = L̇L ⇒ L̇ = nL ⇒ L = nL̇ = L0 ent
* At = (1 + g)At−1
( )α−1 * g = ȦA ⇒ Ȧ = gA ⇒ A = gȦ = A0 egt
* rt = fK = α AKt Lt t ( K )α−1
( )α * r = fK = α AL
( K )α
* wt = fL = (1 − α)At AKt Lt t * w = fL = (1 − α)A AL
Dynamics: General Solow, F .
 Using the general Solow model

* In what follows, we will use the continuous time Solow model (see Exercise , p. )
* As before, y = Y/L and k = K/L
K
* Now, define technology-adjusted capital per worker (capital per e fective worker) as k̃ = k̂ = AL
Y
* And let technology-adjusted income per worker (income per e fective worker) be ỹ = ŷ = AL
k y
* Note that k̂ = A and ŷ = A
* This implies that k = k̂A and y = ŷA... useful
* These per e fective worker variables are constant in the general Solow model's steady-state. For these
variables to be constant, the per worker variables must grow at the same rate as technology, i.e. g
* So in steady-state (balanced growth):
* Growth rate of k̂, ŷ, ĉ is
* Growth rate of k, y, c, w is g
* Growth rate of K, Y, C is g + n
 The Solow equation: general model

* If Y = Kα (AL)1−α , then Y/AL = ŷ = k̂α

˙
k̂
* Now, gk̂ =
k̂
* and, gk̂ = gK − (gA + gL ) = gK − gA − gL
( )
K
* Because k̂ = AL
and ln k̂ = ln(K) − (ln(A) + ln(L))
Ȧ L̇
* Furthermore, gA = A = g and gL = L = n, while K̇ = S − δK
* Therefore, K̇ = sY − δK
˙ (Y)
* Then, k̂ = sY−δK
K −n−g=s K − δK/K − n − g
k̂
˙
* ∴ k̂
k = gK − gA − gL = s ŷ − δ − g − n
k̂
˙
* Multiply with k̂: k̂ = sŷ − δ k̂ − nk̂ − gk̂ = sk̂α − (n + g + δ)k̂
* This is the Solow equation for the general Solow model
 The Solow diagram ( . )

˙
* The Solow equation for the general Solow model tells us that the change in k̂ = investment per e fective worker (sk̂α )
minus break-even investment per e fective worker ((n + g + δ)k̂)
˙
* If sk̂α > (n + g + δ)k̂ ⇒ k̂ > 0
˙
* If sk̂α < (n + g + δ)k̂ ⇒ k̂ < 0
˙
* If sk̂α = (n + g + δ)k̂ ⇒ k̂ = 0
˙
* Steady-state: k̂ = 0
˙
* In steady-state, income per e fective worker (ŷ = k̂α ) is constant if k̂ = 0
* And, y and k grow at g!
 Modified Solow diagram

˙
* Solow equation: k̂ = sk̂α − (n + g + δ)k̂
˙
k̂
* Divide by k̂: = sk̂α−1 − (n + g + δ)
k̂
˙
* Modified Solow equation tells us that growth rate of capital per e fective worker ( k̂ ) = di ference between
k̂
growth rate of investment per e fective worker (sk̂α−1 ) and break-even investment per e fective worker
(n + g + δ)
 solving for steady-state

* Finding k̂∗
˙
* In steady-state, k̂ = 0 ⇒ sk̂α − (n + g + δ) k̂ = 0
∴ sk̂α = (n + g + δ) k̂
∴ sk̂ = (n + g + δ)
α

k̂
∴ = k̂α−1 = n+g+δ
k̂α
s
k̂
( ) α−1
1 ( ) α−1
1

k̂ = k̂α−1 = n+g+δ
s
(( )−1 ) α−1
1
( ) 1−α
1
∗ s s
∴ k̂ = n+g+δ = n+g+δ
(( 1 )α
) 1−α ( )α/(1−α)
∗ ∗α s s
* Then ŷ = k̂ = n+g+δ = n+g+δ
 Steady-state in general Solow model ( . )

* For discrete time model: * For continuous time model:

˙
* In steady-state, ∆k̂ = 0 ⇒ sk̂α = * In steady-state, k̂ = 0 ⇒ sk̂α = (n + g + δ)k̂ ⇔
( )1/(1−α) ( )1/(1−α)
(n + g + ng + δ)k̂ ⇔ k̂∗ = n+ng+δ s
k̂∗ = n+g+δs

* Since ( )α/(1−α) * Since ( )α/(1−α)

ŷ = k̂α ⇒ ŷ∗ = (k̂)∗ = n+ng+δ
s
ŷ = k̂α ⇒ ŷ∗ = (k̂∗ )α = n+g+δs

* Since ( )α/(1−α) * Since ( )α/(1−α)

ĉ = (1 − s)ŷ ⇒ ĉ∗ = (1 − s) n+ng+δ
s
ĉ = (1 − s)ŷ ⇒ ĉ∗ = (1 − s) n+g+δs
 Steady-state: per worker variables

* Discrete time: * Continuous time:

( )1/(1−α) ( )1/(1−α)
∗
* Since k = Ak̂ ⇒ k = A s
n+g+ng+δ
* Since k = Ak̂ ⇒ k∗ = A n+g+δ s

( )α/(1−α) ( )α/(1−α)
* Since y = Aŷ ⇒ y∗ = A n+g+ng+δ
s
* Since y = Aŷ ⇒ y∗ = A n+g+δs

( )α/(1−α) ( )α/(1−α)
* c∗ = A(1 − s) n+g+ng+δ
s
* c∗ = A(1 − s) n+g+δs

( )−1 ( )−1
* r∗ = α n+g+ng+δ
s
* r∗ = α n+g+δs

( )α/(1−α) ( )α/(1−α)
* w∗ = A(1 − α) n+g+ng+δs
* w∗ = A(1 − α) n+g+δ s
 More steady-state results, implications

( )α/(1−α)
s
* Given that ŷ∗ = n+g+δ , a country with a higher (per e fective worker) income level has:
* A higher savings rate
* A lower depreciation rate
* A lower population growth rate
* And a lower rate of technological progress .... ???
* Recall that per worker variables are equal to per e fective worker variables times technology, so
* y = Aŷ (and k = Ak̂), etc.
* Let At = A0 egt
* Or At = (1 + g)t A0 in the discrete case
( )α/(1−α)
s
* Then, y∗ = A0 egt n+g+δ

* Change in g now a fects y in two ways:

( )α/(1−α)
s
* Lower ŷ = n+g+δ
* Higher A0 egt
* Which e fect is greater???
* The e fect of g on ŷ is static/once-o f; the e fect of g on A is dynamic and exponential... the e fect of higher g on A
overwhelms the e fect that higher g has on ŷ... y will therefore increase if g increases...
* Another way to show this, is by taking logs: ln(y∗ ) = ln(A0 ) + gt + α
(1−α) (ln(s) − ln(n + g + δ))
* So, country with higher income per worker (per capita), has higher savings rate, lower population growth and
lower depreciation rates, and higher rate of technological progress
 Increase in s

* With increase in s:
* sk̂α > (n + g + δ)k̂
˙
* Therefore, k̂ > 0
* As k̂ increases, so does ŷ = k̂α
* But, due to diminishing returns to capital, each successive increase in k̂ leads to smaller and smaller successive
˙
increases in ŷ, until k̂ and ŷ reach their new steady-state levels, where k̂ = 0 ⇒ sk̂α = (n + g + δ)k̂
* Increase in s leads to permanent increase in level of ŷ and y, but not growth rate ...
( )α/(1−α)
* ŷ∗ = s
n+g+δ
, while y∗ = Aŷ∗
∗ ∗
* Since ln(y) = ln(A) + ln(ŷ) implies that gy∗ = g + 0 = g
Higher s
Higher s
Higher s
Higher s
Transition from old to new steady-state following increase in s (F . &
 Increase in s, modified Solow
* We can also use the modified Solow diagram to show e fects of an increase in s...
 Increase in s, modified Solow
* We can also use the modified Solow diagram to show e fects of an increase in s...
 Increase in s, modified Solow
* We can also use the modified Solow diagram to show e fects of an increase in s...
simulation: higher s
 Decrease in s

* With decrease in s:
* sk̂α < (n + g + δ)k̂
˙
* Therefore, k̂ < 0
* As k̂ decreases, so does ŷ = k̂α
* But, due to diminishing returns to capital, each successive decrease in k̂ leads to smaller and smaller successive
˙
decreases in ŷ, until k̂ and ŷ reach their new steady-state levels, where k̂ = 0 ⇒ sk̂α = (n + g + δ)k̂
* Decrease in s leads to permanent decrease in level of ŷ and y, but not growth rate ...
Lower s
Lower s
Lower s
Lower s
simulation: lower s
 Increase in n

* With increase in n:
* sk̂α < (n + g + δ)k̂
˙
* Therefore, k̂ < 0
* As k̂ decreases, so does ŷ = k̂α
* But, due to diminishing returns to capital, each successive decrease in k̂ leads to smaller and smaller successive
˙
decreases in ŷ, until k̂ and ŷ reach their new steady-state levels, where k̂ = 0 ⇒ sk̂α = (n + g + δ)k̂
* Increase in n leads to permanently lower level of ŷ∗ and y∗ , but does not a fect LR growth rate ...
( )α/(1−α)
* ŷ∗ = s
n+g+δ
, while y∗ = Aŷ∗
∗ ∗
* Since ln(y) = ln(A) + ln(ŷ) implies that gy∗ = g + 0 = g
Higher n
Higher n
Higher n
simulation: higher n
 Decrease in n

* With decrease in n:
* sk̂α > (n + g + δ)k̂
˙
* Therefore, k̂ > 0
* As k̂ increases, so does ŷ = k̂α
* But, due to diminishing returns to capital, each successive increase in k̂ leads to smaller and smaller successive
˙
increases in ŷ, until k̂ and ŷ reach their new steady-state levels, where k̂ = 0 ⇒ sk̂α = (n + g + δ)k̂
* Decrease in n leads to permanently higher level of ŷ∗ and y∗ , but does not a fect LR growth rate ...
Lower n
Lower n
Lower n
simulaton: lower n
 A note about n

* Higher n will lead to higher L which will lead to higher Y, because Y = f(K, AL)
* Solow model predicts that higher n will lead to lower SS levels of ŷ and y
* Higher n does not a fect steady-state growth rate of ŷ and y... which are 0 and g
* Recall that ŷ = Y
AL
= Ay , while y = Y
L
= Aŷ
* Also note that Y = ALŷ or Y = Ly... which implies that the steady-state growth rate of Y is...?
* So, in steady-state, higher n does not a fect per capita GDP (y) growth rate (which is g), but it can lead to
higher GDP (Y) growth... don't confuse GDP and GDP per capita... And remember that GDP per worker/per
capita is what matters for welfare...
* For y, the e fect on level is due to interaction of two assumptions built into our model ( ) diminishing returns to
capital and ( ) labour/population grows exponentially (if n = L̇/L, then Lt = L0 ent )...
* Note that k = KL a fects y = YL . Capital accumulation can't be exponential - diminishing returns PLUS upper limit
on s. So, if L is increasing exponentially and at increasing rate, then k will start decreasing, as will y
* But if population growth is too low, particularly if fertility rates are below replacement level – . , this also
creates problems, e.g. China, Japan, South Korea, parts of Western Europe, as stagnant or very slowly
growing populations tend to be ageing populations...
* How does this compare with predictions of other models/theories (e.g. AK/endogenous)? ... stay tuned
 Increase in g

* With increase in g:
* sk̂α < (n + g + δ)k̂
˙
* Therefore, k̂ < 0
* As k̂ decreases, so does ŷ = k̂α
* But, due to diminishing returns to capital, each successive decrease in k̂ leads to smaller and smaller successive
˙
decreases in ŷ, until k̂ and ŷ reach their new steady-state levels, where k̂ = 0 ⇒ sk̂α = (n + g + δ)k̂
* Increase in g leads to permanently lower level of ŷ∗ . But, y∗ will be higher, while growth rate of y is also
permanently higher ...
( )α/(1−α)
* ŷ∗ = s
n+g+δ
, while y∗ = Aŷ∗
* It can be shown that A at a particular point in time is A(t) = A0 egt , where A0 is given. So, A increases exponentially
over time, and if g increases, compound e fect will ”swamp” once-o f decrease in ŷ∗
* Since ln(y)∗ = ln(A) + ln(ŷ∗ ) implies that gy∗ = g + 0 = g (where g is higher than before)
* So, even though ŷ∗ is lower, y∗ and gy are both higher following increase in g
Higher g
Higher g
Higher g
 Higher g, e fect on y and g

* Increase in g leads to permanently lower level of ŷ∗ . But, y∗ will be higher, while growth rate of y is also
permanently higher ...
( )α/(1−α)
* ŷ∗ = s
n+g+δ
, while y∗ = Aŷ∗
* It can be shown that A at a particular point in time is A(t) = A0 egt , where A0 is given. So, A increases exponentially
over time, and if g increases, compound e fect will ”swamp” once-o f decrease in ŷ∗
* Since ln(y)∗ = ln(A) + ln(ŷ∗ ) implies that gy∗ = g + 0 = g (where g is higher than before)
* So, even though ŷ∗ is lower, y∗ and gy are both higher following increase in g
* OK, what does this look like diagrammatically?
Higher g: per capita income (y), Solow diagram
Higher g: per capita income (y), Solow diagram
Higher g: per capita income (y), Solow diagram
Higher g: per capita income (y), Solow diagram
 Decrease in g

* With decrease in g:
* sk̂α > (n + g + δ)k̂
˙
* Therefore, k̂ > 0
* As k̂ increases, so does ŷ = k̂α
* But, due to diminishing returns to capital, each successive increase in k̂ leads to smaller and smaller successive
˙
increases in ŷ, until k̂ and ŷ reach their new steady-state levels, where k̂ = 0 ⇒ sk̂α = (n + g + δ)k̂
* Decrease in g leads to permanently higher level of ŷ∗ . But, y∗ will be lower, while growth rate of y is also
permanently lower ...
( )α/(1−α)
* ŷ∗ = s
n+g+δ
, while y∗ = Aŷ∗
* It can be shown that A at a particular point in time is A(t) = A0 egt , where A0 is given. So, A increases exponentially
over time, but at a slower rate, and if g increases, this compound e fect on A will ”swamp” once-o f increase in ŷ∗
* So, even though ŷ∗ is higher, y∗ and gy are both lower following decrease in g
simulation: higher g
Lower g
Lower g
Lower g
 Lower g, modified Solow

* We can also show e fects of changes in g and n using the modified Solow diagram
* As an example, consider decrease in g
 Lower g, modified Solow

* We can also show e fects of changes in g and n using the modified Solow diagram
* As an example, consider decrease in g
 Lower g, modified Solow

* We can also show e fects of changes in g and n using the modified Solow diagram
* As an example, consider decrease in g
simulation: lower g
 Government spending and the savings rate (SW, p. )

* Equilibrium in ouptput market is Y = Cp + Ip + G

* Let G = Cg + Ig
* Then Y = Cp + Cg + Ip + Ig
* This is the same as (Y − Cp − T) + (T − Cg ) = Ip + Ig , which states that Sp + Sg = Ip + Ig
* Total capital in the economy is K and capital accumulation is K̇ = Sp + Sg − δK, where Sp + Sg = S = sY,
so that K̇ = sY − δK, same as before
* All that is needed is to assume that s is constant...
* Also, implies that Cp /Y + Cg /Y = C/Y is constant, meaning Cp /Y and Cg /Y are inversely related ... some
empirical evidence for this
* Also possible to incorporate government by introducing income tax τ Y. Tax increases lower disposable
income, and therefore savings and investment... Similar to how higher Cg leads to lower s ...
 Structural policies for growth (SW, . )

* In LR, increase in government consumption will fully crowd out investment spending (lead to lower national
savings rate, s)
* But, increase in government investment (ig ), leads to increase in savings rate
* Increase in cg might be needed to correct for market failures – provision of public goods (law and order, etc.)
* If public consumption spending on health and education are perfect substitutes for private spending on
these items, overall saving will be unchanged
* Distributional concerns about spending on health and education may motivate increased government
spending on these items, even if not perfect substitutes for private spending
* A lot of public spending serves as essential inputs for private sector (again, education and health)
 Incentive policies

* Our model tells us that higher s and g, as well as lower n will lead to higher levels of income per worker. But
how can these changes be achieved?
* Policies to promote private savings:
* Tax incentives
* System of well-defined, secure property rights
* Well functioning and properly regulated financial system
* Policies to promote technological progress:
* These are very elusive...
* Tax incentives for R& D spending
* Patent and copyright laws
* Education
* Policies to lower population growth rate
* Readily available and a fordable contraceptive and family planning services
* Better pre- and postnatal services (to reduce infant mortality)
* Good labour market and educational opportunities for women and girls
/ In closing
 In closing

* Work through the list of homework questions posted on BB. Really important for assessment purposes
* Next time, we wrap up our discussion of economic growth, and focus on:
* Economic convergence (SW . & . & CS . )
* Growth accounting (SW, . - . )
* Simulations using the (discrete) time general Solow model (see SW . , Ex. . - & Ex. . )
* The golden rule of savings (See SW . & Ex. . )
* A model of endogenous growth (SW, . , . - . )
* Yesterday a ternoon ( / / ): supposed to be Budget Speech... delayed until (?) March –
unprecedented. Documents will be available here when speech is delivered...