Lecture 3
A firm turns inputs into outputs. The production function summarizes this:
\[Q = f(K, L)\]
Shows the maximum output from each input combination (efficient production).
The marginal product of an input is the additional output from one more unit of that input, holding other inputs constant:
\[MP_L = \frac{\partial f}{\partial L} = f_L, \qquad MP_K = \frac{\partial f}{\partial K} = f_K\]
Diminishing marginal productivity: We assume that marginal products eventually decrease
\[\frac{\partial^2 f}{\partial L^2} = f_{LL} < 0, \qquad \frac{\partial^2 f}{\partial K^2} = f_{KK} < 0\]
Holding capital fixed, each additional worker adds less and less output.
Average product of labor: \[AP_L = \frac{Q}{L} = \frac{f(K,L)}{L}\]
Important: \(AP_L\) depends on \(K\) too. When we say “U.S. workers are more productive than Indian workers,” much of that reflects more capital per worker, not inherently more productive labor.
Cross-productivity effects: Typically \(f_{KL} > 0\) — more capital makes labor more productive and vice versa.
| Consumer | Producer |
|---|---|
| Utility \(U(x,y)\) | Output \(f(K,L)\) |
| Budget constraint | Cost constraint |
| Indifference curves | Isoquants |
| MRS | RTS |
The RTS (rate of technical substitution) is the rate at which labor can substitute for capital, holding output constant:
\[RTS = -\frac{dK}{dL}\bigg|_{Q = \bar{Q}} = \frac{MP_L}{MP_K} = \frac{f_L}{f_K}\]
Derivation: Total differential along an isoquant (\(dQ = 0\)): \[dQ = f_L \, dL + f_K \, dK = 0 \implies \frac{dK}{dL} = -\frac{f_L}{f_K}\]
Diminishing RTS: As we move along an isoquant (more \(L\), less \(K\)), the RTS falls. This means isoquants are convex — same logic as diminishing MRS for consumers.
Along an isoquant, RTS decreases as capital-labor (\(K/L\)) ratio decreases
The elasticity of substitution \(\sigma\) measures responsiveness of RTS to changes in \(K/L\):
\[\sigma = \frac{\%\Delta(K/L)}{\%\Delta RTS} = \frac{d\ln(K/L)}{d\ln(RTS)}\]
\(\sigma\) captures how easily the firm can substitute between inputs.
| Function | Formula | \(\sigma\) |
|---|---|---|
| Linear | \(Q = \alpha K + \beta L\) | \(\infty\) |
| Leontief | \(Q = \min(\alpha K, \beta L)\) | \(0\) |
| Cobb-Douglas | \(Q = K^\alpha L^\beta\) | \(1\) |
| CES | \(Q = (\alpha K^\rho + \beta L^\rho)^{\gamma/\rho}\) | \(\frac{1}{1-\rho}\) |
CES is appropriate for most applications because it nests the other three as special cases.
The CES function can also be written as:
\[Q = \left[\alpha K^{\frac{\sigma - 1}{\sigma}} + \beta L^{\frac{\sigma - 1}{\sigma}}\right]^{\frac{\sigma}{\sigma - 1}}\]
where \(\sigma\) is the elasticity of substitution.
Applications:
How does output respond when all inputs scale proportionally?
| Condition | Returns to Scale |
|---|---|
| \(f(tK, tL) = t \cdot f(K,L)\) | Constant (CRS) |
| \(f(tK, tL) < t \cdot f(K,L)\) | Decreasing (DRS) |
| \(f(tK, tL) > t \cdot f(K,L)\) | Increasing (IRS) |
CRS implies:
Why this matters: Returns to scale determine the shape of long-run cost curves and industry structure.
Economists and accountants think about costs differently:
Accounting costs: Out-of-pocket expenses, historical prices, depreciation
Economic costs: Opportunity costs i.e. the payment required to keep an input in its present use
Key differences:
Total economic cost: \(\quad C = rK + wL\)
\[\min_{K,L} \; rK + wL \quad \text{s.t.} \quad f(K,L) = q_0\]
Lagrangian: \[\mathcal{L} = rK + wL + \lambda[q_0 - f(K,L)]\]
First-order conditions: \[r = \lambda f_K, \qquad w = \lambda f_L, \qquad f(K,L) = q_0\]
Dividing the first two: \[\frac{f_L}{f_K} = \frac{w}{r} \qquad \Longrightarrow \qquad RTS = \frac{w}{r}\]
/var/folders/tm/xj959v715k78plswzz0frfqc0000gn/T/ipykernel_74707/1879582369.py:36: UserWarning: No artists with labels found to put in legend. Note that artists whose label start with an underscore are ignored when legend() is called with no argument.
ax.legend(fontsize=13)
The optimal point is where the isoquant is tangent to the lowest isocost line.
From the first-order conditions:
\[\frac{f_L}{w} = \frac{f_K}{r} = \frac{1}{\lambda}\]
Interpretation of \(\lambda\):
The cost function \(C(r, w, Q)\) gives the minimum cost of producing \(Q\) units given input prices:
\[C(r, w, Q) = rK^*(r,w,Q) + wL^*(r,w,Q)\]
where \(K^*\) and \(L^*\) are the cost-minimizing input choices from the previous problem.
Average cost: \(\quad AC = \frac{C(Q)}{Q}\)
Marginal cost: \(\quad MC = \frac{\partial C}{\partial Q}\)
Key relationships:
Intuition: Same as batting average, if your latest at-bat (marginal) is above your average, it pulls the average up.
For Cobb-Douglas \(Q = K^{0.5} L^{0.5}\), the cost-minimizing inputs are:
\[L^* = \frac{Q}{\sqrt{w/r}}, \qquad K^* = Q \sqrt{w/r}\]
Plugging into the cost function: \[C(r, w, Q) = wL^* + rK^* = w \cdot \frac{Q}{\sqrt{w/r}} + r \cdot Q \sqrt{w/r} = 2Q\sqrt{rw}\]
Average and marginal cost:
\[AC = \frac{C}{Q} = 2\sqrt{rw}, \qquad MC = \frac{\partial C}{\partial Q} = 2\sqrt{rw}\]
\(AC=MC\) and both are constant in \(Q\) always for CRS.
Long run: All inputs are variable. The firm chooses the cost-minimizing \((K, L)\).
Short run: Capital is fixed at \(\bar{K}\). The firm can only adjust labor.
\[SC(\bar{K}, w, Q) = r\bar{K} + wL(Q, \bar{K})\]
Key result: Short-run costs \(\geq\) long-run costs, with equality only at the output level for which \(\bar{K}\) is optimal.
\[SC(Q) \geq C(Q), \qquad SC(q) = C(q) \text{ where } \bar{K} = K^*(q)\]
\[\min_L r\bar{K} + wL \quad \text{s.t.} \quad Q = \bar{K}^{0.5} L^{0.5}\]
In the short-run: \(L^* = Q^2/\bar{K}\), so the short-run cost function is: \[SC(\bar{K}, w, Q) = r\bar{K} + w \cdot \frac{Q^2}{\bar{K}}\]
Short-run average, marginal, and average variable cost:
\[SAC = \frac{SC}{Q} = \frac{r\bar{K}}{Q} + w \cdot \frac{Q}{\bar{K}}, \quad SMC = \frac{\partial SC}{\partial Q} = 2w \cdot \frac{Q}{\bar{K}}, \quad SAVC = w \cdot \frac{Q}{\bar{K}}\]
SAC is U-shaped because of the fixed cost component.
The firm chooses output \(Q\) to maximize economic profit:
\[\pi = R(Q) - C(Q) = pQ - C(Q)\]
where \(R(Q)\) is total revenue and \(C(Q)\) is total (economic) cost.
First-order condition: \[\frac{d\pi}{dQ} = \frac{dR}{dQ} - \frac{dC}{dQ} = 0 \implies MR = MC\]
Second-order condition: \[\frac{d^2\pi}{dQ^2} < 0 \implies \frac{dMR}{dQ} < \frac{dMC}{dQ}\]
At the optimum, \(MC\) must be rising faster than \(MR\).
Marginal revenue is the extra revenue from selling one more unit:
\[MR = \frac{dR}{dQ} = \frac{d(p(Q)\cdot Q)}{dQ} = p + Q\cdot\frac{dp}{dQ}\]
In terms of demand elasticity \(e_{Q,p} = \frac{dQ}{dp} \cdot \frac{p}{Q}\), we can rewrite MR as:
\[MR = p\left(1 - \frac{1}{|e_{Q,p}|}\right)\]
For a price-taking firm, demand is infinitely elastic
\[|e_{Q,p}| \to \infty \implies MR \to p\]
For a price-taking firm (\(MR = P\)), the profit-maximizing rule \(MR = MC\) becomes \[P = MC\]
Short-run supply curve: The upward-sloping portion of \(SMC\) above minimum \(SAVC\).
Producer surplus measures the benefit to the firm from producing at the market price versus not producing at all.
\[PS = \int_{P_s}^{P} Q(P') \, dP' \]
Producer surplus is the area below the market price and above the supply curve. It’s the surplus that producers earn above their minimum willingness to supply.
We can also think of the firm as choosing inputs directly:
\[\max_{K,L} \; pf(K,L) - rK - wL\]
First-order conditions:
\[p \cdot MP_L = w \qquad \text{and} \qquad p \cdot MP_K = r\]
Hire each input until its marginal revenue product equals its price.
What happens to labor demand when the wage \(w\) falls?
Two effects (parallel to consumer theory):
1. Substitution effect (holding output constant)
2. Output effect (adjusting output)
What about the effect of a wage decrease on capital demand?
\[\frac{\partial K}{\partial w} = \; ?\]
Substitution effect: \(w\) falls → substitute toward labor, away from capital → less \(K\)
Output effect: \(w\) falls → lower costs → more output → more \(K\) needed
These work in opposite directions. The net effect is ambiguous.