1 The first known anthro-capital Olduvai stone chopping tool, 1.8 – 2 million years BP

2 Agenda Introductory background Essential aspects of economic growth Aggregate production functions Neoclassical growth model Simulation of increased saving experiment Then in the last week: deficits, debt, and economic growth

Great divide of macroeconomics Aggregate demand and business cycles Aggregate supply and “economic growth”

4 4 Congressional Budget Office, March 2013, cbo.gov. Actual and potential output Economic growth: Studies growth of potential output

Potential and actual real output in the long run 5

How do they fit together? Examples of why Keynesian  Classical In long run, prices and wages are flexible. In long run, expectations are accurate (rational?). In long run, entry and exit make economy more competitive. All these make long-run look more classical than short run. 6

π Y AD Y t * = potential output AS short run AS-AD in short, medium, long run πtπt AS medium run AS long run

8 Growth trend, US, (pre-recession)

9 Historical Trends in Economic Growth in the US since Strong growth in Y 2. Strong growth in productivity (both Y/L and TFP, total factor productivity) 3. Steady “capital deepening” (increase in K/L) 4. Strong growth in real wages since early 19 th C; g(w/p) ~ g(X/L) 5. Real interest rate and profit rate basically trendless

Capital deepening in agriculture 10 Food for Commons Morocco, 2002

Mining in rich and poor countries Canada 11 D.R. Congo

12 Review of aggregate production function Y t = A t F(K t, L t ) K t = capital services (like rentals as apartment-years) L t = labor services (hours worked) A t = level of technology g x = growth rate of x = (1/x t ) dx t /dt = Δ x t /x t-1 = d[ln(x t) )]/dt g A = growth of technology = rate of technological change = Δ A t /A t-1 Constant returns to scale: λY t = A t F(λK t, λL t ), or all inputs increased by λ means output increased by λ Perfect competition in factor and product markets (for p = 1): MPK = ∂Y/∂K = R = rental price of capital; ∂Y/∂L = w = wage rate Exhaustion of product with CRTS: MPK x K + MPL x L = RK + wL = Y Alternative measures of productivity: Labor productivity = Y t /L t Total factor productivity (TFP) = A t = Y t /F(K t, L t )

13 Review: Cobb-Douglas aggregate production function Remember Cobb-Douglas production function: Y t = A t K t α L t 1-α or ln(Y t )= ln(A t ) + α ln(K t ) +(1-α) ln(L t ) Here α = ∂ln(Y t )/∂ln(K t ) = elasticity of output w.r.t. capital; (1-α ) = elasticity of output w.r.t. labor MPK = R t = α A t K t α-1 L t 1-α = α Y t /K t Share of capital in national income = R t K t /Y t = α = constant. Ditto for share of labor.

The MIT School of Economics 14 Paul Samuelson ( ) Robert Solow ( )

15 Basic neoclassical growth model Major assumptions: 1.Basic setup: - full employment -flexible wages and prices -perfect competition -closed economy 2. Capital accumulation: ΔK = sY – δK; s = investment rate = constant 3. Labor supply: Δ L/L = n = exogenous 4. Production function - constant returns to scale - two factors (K, L) - single output used for both C and I: Y = C + I - no technological change to begin with - in next model, labor-augmenting technological change 5.Change of variable to transform to one-equation model: k = K/L = capital-labor ratio Y = F(K, L) = LF(K/L,1) y = Y/L = F(K/L,1) = f(k), where f(k) is per capita production fn.

16 Major variables: Y = output (GDP) L = labor inputs K = capital stock or services I = gross investment w = real wage rate r = real rate of return on capital (rate of profit) E = efficiency units = level of labor-augmenting technology (growth of E is technological change = ΔE/E) L ~ = efficiency labor inputs = EL = similarly for other variables with “ ~ ”notation) Further notational conventions Δ x = dx/dt g x = growth rate of x = (1/x) dx/dt = Δx t /x t-1 =dln(x t )/dt s = I/Y = savings and investment rate k = capital-labor ratio = K/L c = consumption per capita = C/L i = investment per worker = I/L δ = depreciation rate on capital y = output per worker = Y/L n = rate of growth of population (or labor force) = g L = Δ L/L v = capital-output ratio = K/Y h = rate of labor-augmenting technological change

17 We want to derive “laws of motion” of the economy. To do this, start with (math on next slide): 5. Δ k/k = Δ K/K - Δ L/L With some algebra,* this becomes: 5’. Δ k/k = Δ K/K - n Y Δ k = sf(k) - (n + δ) k which in steady state is: 6. sf(k*) = (n + δ) k* In steady state, y, k, w, and r are constant. No growth in real wages, real incomes, per capita output, etc. *The algebra of the derivation: ΔK/K = (sY – δK)/K = s(Y/L)(L/K) – δ Δk/k = ΔK/K – n = s(Y/L)(L/K) – δ – n Δk = k [ s(Y/L)(L/K) – δ – n] = sy – (δ + n)k = sf(k) – (δ + n)k

Mathematical note We will use the following math fact: Define z = y/x Then (1) (Growth rate of z) = (growth rate of y) – (growth rate of x) Or g z = g y - g x Proof: Using logs: ln(z t ) = ln(y t )– ln(x t ) Taking time derivative: [dz t /dt]/z t = [dy t /dt]/y t - [dx t /dt]/x t which is the desired result. Note that we sometimes use the discrete version of (1), as we did in the last slide. This has a small error that is in the order of the size of the time step or the growth rates. For example, if g y = 5 % and g x = 3 %, then by the formula g z = 2 %, while the exact calculation is that g z = %. This is close enough for expository purposes. 18

19 k y = Y/L y = f(k) (n+δ)k y* i* = (I/Y)* k* i = sf(k)

20 Predictions of basic model: –“Steady state” –constant y, w, k, and r –g Y = n Uniqueness and stability of equilibrium. –Equilibrium is unique –Equilibrium is stable (meaning k → k* as t → ∞ for all initial k 0 ). Results of neoclassical model without TC

Economic growth (II): Technological change 21 World fastest supercomputer, 2012 (Tianhe-2 in NUDT, Guangzhou, China ) at 33.8 petaflops (34x10 15 floating point operations per second, up from 16.2 last year)

22 Introducing technological change First model omits technological change (TC). What is TC? New processes that increase TFP (assembly line, fiber optics) Improvements in quality of goods (plasma TV) New goods and services (automobile, telephone, Twitter) Analytically, TC is - Shift in production function.

The greatest technological change in history [flop = floating point operations per second, e.g., / ] Abacus master (.03 flops) IBM 1620, circa 1960 (10 4 flops) High-end PC, 2013 (10 11 flops) Tianhe-2, top supercomputer, 2013 (34x10 15 flops) Yale

Technological change in medicine Scan for lung cancer 24 African medicine man

Disappearance of polio: A benefit of growth that is not captured in the GDP statistics!

26 Introducing technological change 26 Introducing technological change We take specific form which is “labor-augmenting technological change” at rate h. For this, we need new variable called “efficiency labor units” denoted as E where E = efficiency units of labor and ~ indicates efficiency units. New production function is then Note: Redefining labor units in efficiency terms is a specific way of representing TC that makes everything work out easily. Other forms of TC will give slightly different results.

27 The math with technological change is this: The equilibrium is unique and globally stable. It has exactly the same properties as earlier one, except: Note that the growth term includes h (rate of tech. change). All natural variables are growing at h: wage rates, per capita output, capital-labor ratio etc. are growing at h rather than 0.

T.C. for the Cobb-Douglas In C-D case, labor-augmenting TC is very simple: 28

29 For C-D case, Unique and stable equilibrium under standard assumptions: Predictions of basic model: –Steady state: constant –Here output per capita, capital per capita, and wage rate grow at h. –Labor’s share of output is constant. –Hence, captures the basic trends! Results of neoclassical model with labor-augmenting TC

30 What are the contributors to growth? Growth Accounting Growth accounting is a widely used technique used to separate out the sources of growth in a country; it relies on the neoclassical growth model Derivation Start with production function and competitive assumptions. For simplicity, assume a Cobb-Douglas production function with technological change: (1)Y t = A t K t α L t 1-α Take logarithms: (2)ln(Y t ) = ln(A t )+ α ln(K t ) + (1 - α) ln(L t ) Or if we assume competitive α = sh(K) and 1- α = sh(L) g(Y) = g(A) + sh(K) g(K) + sh(L) g(L )

31 Per capita output What are the contributions to per capita output growth? g(Y/L) = g(y) = g(A) + sh(K) g(K) + sh(L) g(L ) –g(L) = g(A) + sh(K) g(K) - sh(K) g(L ) = g(A) + sh(K) (g(K) - g(L )) = g(A) + sh(K) (g(k)) Since g(y) = 1.7 % p.y.; sh(K) = ¼; g(k) =2.5% So contribution to growth is Of A: 1.1%, or 65% of total Of k: ¼ x 2.5, or 35% of total This is the very surprising results that technology contributes most of the rise in per capita output over the period. Similar in other countries/periods. Source: BLS, multifactor productivity page.

32 i*=I*/Y* Impact with labor-augmenting TC

33 ln (Y), etc. time ln (L) ); g L = n Time profiles of major variables with TC ln (K); g K = n+h ln (Y); g Y = n+h

34 Sources of TC Technological change is in some deep sense “endogenous.” The underlying theory will be discussed on Monday.

35 Several “comparative dynamics” experiments Change growth in labor force (immigration or retirement policy) Change in rate of TC Change in national savings and investment rate (tax changes, savings changes, demographic changes) Here we will investigate only a change in the national savings rate.