Lesson 8 told us the allocation, \(x^i = e^i\). It did not tell us the prices. The proposition held for any positive prices at each step, so prices must be pinned down by something else: each generation must choose its endowment when facing them. In equilibrium, autarky is not imposed. It is optimal at the equilibrium prices.
Normalize \(p_1 = 1\). If \(v^i\) is differentiable and the endowment is interior, the first-order conditions are necessary at the optimum \(x^i = e^i\) II p.9:
$$v^i_1(e^i_i,e^i_{i+1}) = \lambda^i p_i,\qquad v^i_2(e^i_i,e^i_{i+1}) = \lambda^i p_{i+1} \quad\Longrightarrow\quad \frac{p_{i+1}}{p_i} = \frac{v^i_2(e^i_i,e^i_{i+1})}{v^i_1(e^i_i,e^i_{i+1})}\qquad i\ge1.$$Equilibrium relative prices equal each generation's marginal rate of substitution evaluated at its endowment. Geometrically, the budget line through the endowment is tangent to the indifference curve there.
The net interest rate between \(t\) and \(t+1\) is defined by $$\frac{1}{1+r_t} = \frac{p_{t+1}}{p_t}.$$ In the no-trade equilibrium, for all \(t\ge1\), $$r_t = \frac{v^t_1(e^t_t,e^t_{t+1})}{v^t_2(e^t_t,e^t_{t+1})} - 1,\qquad p_t = \frac{1}{(1+r_1)(1+r_2)\cdots(1+r_{t-1})} .$$
Interpretation: \(p_t/p_{t+1} = 1+r_t\) is how many date-\(t+1\) goods one date-\(t\) good buys. That is exactly the gross return on a one-period loan (Exercise 8.4 makes the sequential-markets equivalence precise). The formula for \(p_t\) chains the one-period rates. Each \(r_t\) is determined by generation \(t\) alone, the only generation with access to both dates \(t\) and \(t+1\).
\(r_t\) is the interest rate at which generation \(t\) is exactly content to neither borrow nor lend. If the young are impatient relative to their endowment profile, so that they would like to borrow against a large old-age endowment, the rate must be high enough to talk them out of it. If they would like to save, the rate must be low (possibly negative) enough to discourage saving. Since nobody can actually trade, it is the price that must adjust.
All generations \(i\ge1\) have $$v(c_y,c_o) = (1-\beta)\log c_y + \beta\log c_o,\qquad \beta\in(0,1),$$ and endowments \(e^i_i = 1-\alpha\) (young), \(e^i_{i+1} = \alpha\) (old), with \(e^0_1=\alpha\) and \(\alpha\in(0,1)\). The aggregate endowment is constant: \(\bar e_t = (1-\alpha) + \alpha = 1\).
\(\beta\) is the weight on old-age consumption: a "patience" or "taste for the future" parameter. \(\alpha\) is the share of lifetime endowment received when old: a "pension-like" endowment. These two numbers alone determine everything.
Class Note II puts \(1-\alpha\) in youth and \(\alpha\) in old age. Class Note III's privatization slides flip this (young \(\alpha\), old \(1-\alpha\)); see Lesson 14. Here, and in Lessons 10–13, we follow Note II.
With log utility, \(v_1 = (1-\beta)/c_y\) and \(v_2 = \beta/c_o\). Evaluating at the endowment \((1-\alpha,\alpha)\) II p.12:
$$\frac{1}{1+r_t} = \frac{p_{t+1}}{p_t} = \frac{\beta/\alpha}{(1-\beta)/(1-\alpha)} = \frac{\beta}{1-\beta}\cdot\frac{1-\alpha}{\alpha}.$$This is the same for every \(t\), so \(r_t\equiv\bar r\) with
$$\boxed{\;1+\bar r = \frac{(1-\beta)\,\alpha}{\beta\,(1-\alpha)},\qquad \bar r = \frac{\alpha-\beta}{\beta(1-\alpha)},\qquad p_t = \left(\frac{\beta}{1-\beta}\cdot\frac{1-\alpha}{\alpha}\right)^{t-1} = (1+\bar r)^{-(t-1)}.\;}$$Check the algebra of \(\bar r\): \(\frac{(1-\beta)\alpha - \beta(1-\alpha)}{\beta(1-\alpha)} = \frac{\alpha - \beta}{\beta(1-\alpha)}\). So \(\bar r>0 \iff \alpha>\beta\), \(\bar r=0\iff\alpha=\beta\) and \(\bar r<0\iff\alpha\lt \beta\) II p.14.
The savings of generation \(t\) when young are \(e^t_t - x^t_t\). Facing a constant rate \(r\), a young agent who saves \(s\) consumes \(1-\alpha-s\) now and \(\alpha + (1+r)s\) later. Its problem (the slide question on II p.13) is
$$s(r;\alpha,\beta) = \arg\max_s\;(1-\beta)\log(1-\alpha-s) + \beta\log\big(\alpha+s(1+r)\big).$$A higher \(r\) has a substitution effect (save more) and an income effect. Because the agent receives some endowment when old, a higher \(r\) lowers the present value of lifetime wealth \((1-\alpha)+\alpha/(1+r)\), and this income effect also raises saving. With log utility the pure-substitution and pure-income effects on \(c_y\) from \(r\) cancel, leaving only this wealth effect. If \(\alpha = 0\), saving \(s=\beta\) would not depend on \(r\) at all.
Equilibrium is \(s(\bar r) = 0\). Since only generation \(t\) can be in the loan market at \(t\), the equilibrium rate is the one at which the young choose zero savings. Setting \(s=0\): \(\beta(1-\alpha)(1+\bar r) = (1-\beta)\alpha\). This gives the same \(\bar r\) as the FOC route in §2.
The slides give a "zero-savings" argument II p.14. Make it precise with the implicit function theorem applied to \(s(\bar r;\alpha,\beta)=0\):
$$\frac{\partial\bar r}{\partial\beta} = -\frac{\partial s/\partial\beta}{\partial s/\partial r} < 0,\qquad \frac{\partial\bar r}{\partial\alpha} = -\frac{\partial s/\partial\alpha}{\partial s/\partial r} > 0 .$$The same conclusions follow from differentiating \(1+\bar r = \frac{(1-\beta)\alpha}{\beta(1-\alpha)}\) directly: it is increasing in \(\alpha\) and decreasing in \(\beta\).
Past exams test this material through short true/false justifications and through one long guided problem on an OLG economy with an asset (Exercise 9.2). Work with general parameters: the exams never ask you to plug in numbers.
True. \(1+\bar r = \frac{(1-\beta)\alpha}{\beta(1-\alpha)}<1 \iff (1-\beta)\alpha\lt \beta(1-\alpha)\iff\alpha\lt \beta\). The young want old-age consumption share \(\beta\) but are endowed with only \(\alpha\). Choking off their desire to save requires a negative rate.
False in general. Log utility fixes the expenditure share: \(c_y=(1-\beta)W\) with \(W=(1-\alpha)+\alpha/(1+r)\). The present value of the old-age endowment falls with \(r\), so \(s=(1-\alpha)-c_y\) rises with \(r\). Savings are independent of \(r\) only when there is no old-age endowment, e.g. \(s_t=\beta w_t\) in the Diamond question on the Fall 2025 final (Lesson 15).
False. \(\bar r\) is the shadow price at which the young are just willing not to trade: \(1/(1+\bar r)=v_2/v_1\) at the endowment. Its sign decides welfare: \(\bar r<0\) is exactly when transfers from young to old (social security, money, debt) can make everyone better off (Lessons 10–11).
False. In a steady state, no arbitrage gives \(1+\bar r=(\bar q+\bar d)/\bar q\), so \(\bar r=\bar d/\bar q>0\) for any positive price. An asset with a positive dividend can't be priced at \(r\le0\), because its price would be infinite.
False. If savings \(\bar\sigma\) increase strictly with the return, \(\bar q\) rises, and so does \(\bar r\): the young will absorb the more valuable tree only at a higher return. With log utility, \(\bar q=\beta\bar e\) does not move at all and \(\bar r=\bar d/(\beta\bar e)\) rises. Either way \(\bar r\) rises.
False (pure endowment economy). Generation \(t\) alone trades between \(t\) and \(t+1\), and it consumes its endowment. So \(1/(1+r_t)\) is generation \(t\)'s MRS at \((e^t_t,e^t_{t+1})\) and nothing else (Exercise 9.5).
False. At the old rate, more patient young want to save more (\(s_\beta>0\)), but saving must stay at zero, so the rate must fall: \(d\bar r/d\beta=-s_\beta/s_r<0\) (Exercise 9.3).
Continue the economy of Exercise 8.2, where you showed \(a_t=1\), \(s_t=0\) and \(1+r_t=(q_{t+1}+d_{t+1})/q_t\). Define the savings function $$\sigma_t(R)\equiv\arg\max_x\; u^t\big(e^t_t-x,\;xR\big),$$ the amount a young agent saves at gross return \(R\).
Total saving of the young is \(a_tq_t+s_t=q_t\). For (d)–(h), plot the 45° line \(y=q\) against \(y=\bar\sigma(1+\bar d/q)\) and ask which curve shifts.
(a) The young save \(q_t\) in total, at gross return \((q_{t+1}+d_{t+1})/q_t\). So $$q_t=\sigma_t\!\left(\frac{q_{t+1}+d_{t+1}}{q_t}\right).$$
(b) \(\bar q=\bar\sigma\big(1+\bar d/\bar q\big)\), with \(\bar r=\bar d/\bar q\).
(c) Yes and yes. Inada conditions make the young save a positive amount, so \(\bar q>0\) and \(\bar r=\bar d/\bar q>0\). With \(p_t=(1+\bar r)^{-(t-1)}\), the value of the aggregate endowment is \(\sum_t p_t(\bar e+\bar d)=(\bar e+\bar d)(1+\bar r)/\bar r\lt \infty\). By the First Welfare Theorem for OLG economies (Lesson 12), the equilibrium is PO. Exam warning: writing "\(\bar r>0\), so the 1WT applies" is not enough. The theorem needs a finite value of the aggregate endowment, and you must say so.
(d) Yes. The left side \(y=q\) is strictly increasing. The right side \(y=\bar\sigma(1+\bar d/q)\) is strictly decreasing in \(q\), because the return \(1+\bar d/q\) falls with \(q\) and \(\bar\sigma\) is increasing. So the curves cross at most once. They do cross: near \(q=0\) the right side is positive while the left side is near zero, and at \(q=\bar e\) the right side is below \(\bar e\), because Inada keeps \(c_y>0\).
(e) The right-hand curve shifts up at every \(q\), so \(\bar q\) rises. Since \(\bar q=\bar\sigma(1+\bar r)\) and \(\bar\sigma\) is strictly increasing, a higher \(\bar q\) needs a higher return: \(\bar r\) rises too, so \(\bar q\) rises less than proportionally to \(\bar d\).
(f) The new savings function \(\tilde\sigma\ge\bar\sigma\) shifts the curve up, so \(\bar q\) rises, and with \(\bar d\) fixed, \(\bar r=\bar d/\bar q\) falls. (Write the inequality as \(\tilde\sigma(R)\ge\bar\sigma(R)\). A circulated answer key has it backwards.)
(g) If \(c_o=xR\) is normal, savings at any \(R\) increase with \(\bar e\). This is the same shift as (f): \(\bar q\) rises and \(\bar r\) falls.
(h) Homotheticity gives \(\bar\sigma(R;\lambda\bar e)=\lambda\bar\sigma(R;\bar e)\). Try \(\lambda\bar q\): the return \(1+\lambda\bar d/(\lambda\bar q)\) is unchanged and saving scales by \(\lambda\), so it solves the equation. By uniqueness, \(\bar q\) scales by \(\lambda\) and \(\bar r\) is unchanged.
(i) The FOC \((1-\beta)/(\bar e-x)=\beta/x\) gives \(x=\beta\bar e\) at every \(R\). So \(\bar q=\beta\bar e\) and \(\bar r=\bar d/(\beta\bar e)\). Then \(\bar r\) rises with \(\bar d\), falls with \(\beta\) and with \(\bar e\), and is invariant to scaling \(\bar e,\bar d\) together. \(\bar q\) does not respond to \(\bar d\): log savings are flat in \(R\), the boundary case of (e).
Let \(s(r;\alpha,\beta)=\arg\max_s (1-\beta)\log(1-\alpha-s)+\beta\log(\alpha+s(1+r))\).
(a) The FOC is \(\frac{1-\beta}{1-\alpha-s}=\frac{\beta(1+r)}{\alpha+s(1+r)}\), so \((1-\beta)(\alpha+s(1+r))=\beta(1+r)(1-\alpha-s)\). Collecting \(s\): \(s(1+r)=\beta(1+r)(1-\alpha)-(1-\beta)\alpha\), which is the stated formula. Expanding the slide's form gives \((1-\alpha)\beta-(1-\beta)\alpha/(1+r)\), the same thing.
(b) \(s_r=\frac{(1-\beta)\alpha}{(1+r)^2}>0\): a higher rate lowers the present value of the old-age endowment, so the young are poorer and consume less today. \(s_\beta=(1-\alpha)+\frac{\alpha}{1+r}>0\): more patience moves consumption toward old age. \(s_\alpha=-\beta-\frac{1-\beta}{1+r}<0\): more old-age endowment means less need to save.
(c) \(d\bar r/d\alpha=-s_\alpha/s_r>0\) and \(d\bar r/d\beta=-s_\beta/s_r<0\). A higher \(\alpha\) lowers desired saving at the old rate, so the rate must rise to restore \(s=0\). A higher \(\beta\) works the other way. Check against \(\bar r=\frac{\alpha-\beta}{\beta(1-\alpha)}\): \(\partial\bar r/\partial\alpha=\frac{1-\beta}{\beta(1-\alpha)^2}>0\) and \(\partial\bar r/\partial\beta=-\frac{\alpha}{\beta^2(1-\alpha)}<0\).
(d) With \(e=(w,0)\) the FOC is \(\frac{1-\beta}{w-s}=\frac{\beta}{s}\), so \(s=\beta w\). The term in \(r\) came only from the present value of the old-age endowment (a wealth effect); with log utility, the income and substitution effects of \(r\) on a pure saver cancel exactly.
Replace log utility with \(v(c_y,c_o) = (1-\beta)\frac{c_y^{1-\sigma}}{1-\sigma} + \beta\frac{c_o^{1-\sigma}}{1-\sigma}\), \(\sigma>0\), keeping endowments \((1-\alpha,\alpha)\).
(a) \(1+\bar r = \dfrac{v_1}{v_2}\Big|_e = \dfrac{1-\beta}{\beta}\Big(\dfrac{\alpha}{1-\alpha}\Big)^{\sigma}\).
(b) \(\bar r<0 \iff (1-\beta)\alpha^\sigma < \beta(1-\alpha)^\sigma\). For \(\sigma = 1\) this is \(\alpha\lt \beta\). In general it depends on \(\sigma\), unless \(\alpha=\tfrac12\).
(c) If \(\alpha<1/2\), then \(\alpha/(1-\alpha)<1\) and \(1+\bar r\) is decreasing in \(\sigma\). With more curvature the young care more about smoothing toward their poor old age, so the rate must fall further to stop them saving.
Keep log preferences with weight \(\beta\), but let generation \(t\)'s endowments be \(e^t_t = w_t\) and \(e^t_{t+1}=a_t\), where \(\{w_t,a_t\}\) are positive sequences bounded above and away from zero.
(a) \(1+r_t = \dfrac{(1-\beta)/w_t}{\beta/a_t} = \dfrac{1-\beta}{\beta}\cdot\dfrac{a_t}{w_t}\) and \(p_t = \prod_{s=1}^{t-1}\dfrac{1}{1+r_s}\).
(b) Each \(r_t\) is generation \(t\)'s MRS at its own endowment, and no one trades across generations.
(c) Over each two-period cycle prices are multiplied by \(\kappa\equiv\frac{1}{(1+r_{\text{odd}})(1+r_{\text{even}})}\). Since \(\bar e_t\) is bounded above and away from zero, \(\sum_tp_t\bar e_t\) behaves like a geometric series in \(\kappa\). So it is finite iff \(\kappa<1\), i.e. \((1+r_{\text{odd}})(1+r_{\text{even}})>1\). Yes: a negative rate at odd dates is fine if it is outweighed by a high enough rate at even dates. Efficiency depends on the cumulative discounting, not on the sign of any single rate (Lesson 12).
Show directly, without FOCs, that at the prices \(p_t = (1+\bar r)^{-(t-1)}\) the young of generation \(t\) find their endowment optimal: the optimal \(c_y\) at rate \(\bar r\) equals \(1-\alpha\).
Log utility spends a share \(1-\beta\) of lifetime wealth on youth: \(c_y = (1-\beta)\big[(1-\alpha) + \alpha/(1+\bar r)\big]\). Since \(1/(1+\bar r) = \beta(1-\alpha)/((1-\beta)\alpha)\), we get \(\alpha/(1+\bar r) = \beta(1-\alpha)/(1-\beta)\). So \(c_y = (1-\beta)(1-\alpha) + \beta(1-\alpha) = 1-\alpha\). ✓ Equivalently, \(s(\bar r) = 0\).
Two-period lives, a mass one per cohort, labour income only when young. Output is \(Y_t\), with \(Y_{t+1}=Y_tz_{t+1}\), where \(z>0\) is i.i.d. The young earn \(w_t=(1-\delta)Y_t\), and one tree, owned initially by the old, pays fruit \(d_t=\delta Y_t\). Preferences are \((1-\beta)\log c^t_t+\beta E_t\log c^t_{t+1}\). The young buy \(s_t\) trees at price \(q^e_t\) and \(b_t\) one-period discount bonds at price \(q^b_t\), each paying one unit next period: $$c^t_t+s_tq^e_t+b_tq^b_t=w_t,\qquad c^t_{t+1}=s_t(q^e_{t+1}+d_{t+1})+b_t .$$
The efficiency half of this problem (is the equilibrium PO even though the safe rate can be below growth?) appears in Lessons 12–13.
(a) Tree: \(\frac{(1-\beta)q^e_t}{c^t_t}=\beta E_t\Big[\frac{q^e_{t+1}+d_{t+1}}{c^t_{t+1}}\Big]\). Bond: \(\frac{(1-\beta)q^b_t}{c^t_t}=\beta E_t\Big[\frac{1}{c^t_{t+1}}\Big]\).
(b) \(s_t=1\): the old sell the tree and the young must hold it. \(b_t=0\): bonds are in zero net supply and all young are identical.
(c) In equilibrium \(c^t_{t+1}=q^e_{t+1}+d_{t+1}\), so the right side of the tree FOC is \(\beta\), and \((1-\beta)q^e_t=\beta c^t_t=\beta(w_t-q^e_t)\). Hence \(q^e_t=\beta w_t=\beta(1-\delta)Y_t\). (Log utility: the young save a fraction \(\beta\) of income, and the tree is the only thing to save in.) A higher \(\beta\) means more saving chasing one tree; a higher \(\delta\) means less labour income to save.
(d) Old consumption is \(c^t_{t+1}=[\beta(1-\delta)+\delta]Y_tz_{t+1}\). The tree return is \(R^e=c^t_{t+1}/q^e_t\), so \(E_tR^e=\Big(1+\frac{\delta}{\beta(1-\delta)}\Big)E[z]\). The bond FOC gives \(Q^b=\frac{\beta(1-\delta)}{\beta(1-\delta)+\delta}E[1/z]\), so \(R^b=\Big(1+\frac{\delta}{\beta(1-\delta)}\Big)\big/E[1/z]\). Since \(E[1/z]\ge1/E[z]\) (Jensen), \(R^b\le E R^e\): the gap is a risk premium.