14.453 Lecture 2 Notes

1 Overview Income Fluctuation problem: • — Quadratic-CEQ → Permanent Income — CARA → precuationary savings — CRRA → steady state inequality — borrowing constraints • General Equilibrium: steady state capital and interest rate 2 Certainty Equivalence and the Permanent Income Hypothesis(CEQ-PIH) 2.1 Certainty • assume βR = 1 T = ∞ for simplicity • no uncertainty: max ∞ X β t u (ct ) t=0 At+1 = (1 + r) (At + yt − ct ) • solution: " # ∞ X r At + yt + R−t yt+j ct = 1+r j=1 1 2.2 Uncertainty: Certainty Equivalence and PIH • tempting... " ¶t ∞ µ X 1 r yt+j At + yt + Et ct = 1+r 1+r j=1 # • Permanent Income Hypothesis (PIH) • Certainty Equivalence: x → E (x) • valid iff: — preferences: u (c) quadratic and c ∈ R • main insight: given “permanent” income ytp ¶t ∞ µ X 1 ≡ yt + Et yt+j 1+r j=1 • ct function of ytp and not independently of yt • innovations ∆ct ≡ ct − ct−1 ¶j ∞ µ 1 r X = [Et yt+j − Et−1 yt+j ] 1 + r j=0 1 + r → revisions in permanent income • implications: — random-walk: Et−1 [∆ct ] = 0 — no insurance... ...consumption smoothing → minimize ∆c 2 — marginal propensity to consume from wealth: r 1+r — marginal propensity to consume from innovation to current income depends on persistence of income process • example: {yt } is MA (2) yt = εt + β 1 εt−1 = β (L) εt r X −j ∆ct = R [Et yt+j − Et−1 yt+j ] 1 + r j=1 ª r © = yt − Et−1 yt + R−1 (Et yt+1 − Et−1 yt+1 ) 1+r r r εt + R−1 β 1 εt = 1+r 1+r ¤ r £ 1 + R−1 β 1 εt = 1+r where yt = εt + β 1 εt−1 , Et−1 yt = β 1 εt−1 and Et−1 yt+j = 0 for j ≥ 1 and Et yt+1 = βεt ∞ • ARMA • persistence → α (L) yt = β (L) εt r β (R−1 ) εt → ∆ct = 1 + r α (R−1 ) ∂ c ∂εt t > r 1+r • with a unit root in yt → mg propensity to consume may be greater than 1 3 Estimation and Tests 3.1 CEQ-PIH • “random walk” (martingale): ∆ct = ut Et−1 ut = 0 3 • ut perfectly correlated with news arriving at t about the expected present value of future income: ∆ct = ut = r X 1 [Et yt+j − Et−1 yt+j ] 1 + r j=0 (1 + r)j ∞ • Two main tests: (generally on aggregate data) — random walk → unpredictability of consumption violations = ‘excess sensitivity’ to predictable current income — propensity to consume too small given income perisistence = “excess smoothness” • both tests rely on persistence of income → controversial • aggregation issues: — across goods — agents: Euler equation typically non-linear Attanasio and Weber → leads to rejection on aggregate data — time aggregation: data averaged over continuous time → introduces serial correlation 3.2 Euler Equations • Hall: revolutionary idea: forget consumption function find property it satisfies → Euler equation! u0 (ct ) = β (1 + r) Et [u0 (ct+1 )] • Attanasio et al 4 4 Precautionary Savings • idea: break CEQ 4.1 Two Periods • two period savings problem: max u (c0 ) + βEU (c̃1 ) a1 + c0 = Ra0 + y0 = x c̃1 = Ra1 + ỹ1 • subsituting: max {u (x0 − a1 ) + βEU (Ra1 + ỹ1 )} a1 f.o.c. (Euler equation) u0 (x0 − a1 ) = βR EU 0 (Ra1 + ỹ1 ) U'[(1+r)k0+z0-k1] β(1+r)EU'[z1+(1+r)k1] k*1 k1 Optimum: unique intersection k*1 Figure 1: optimum: unique intersection k1∗ 5 • mean preserving spread: second order stochastic dominance replace ỹ1 with ỹ10 = ỹ1 + ε̃ with E (ε̃ | y) = 0 U'[(1+r)k0+z0-k1] β(1+r)EU'[z1+E+(1+r)k1] β(1+r)EU'[z1+(1+r)k1] k*1 k1 k*1 ' Increase in uncertainty Comparative Static with u000 > 0: Mean Preserving Spread • three possibilities: — U 0 (·) linear ⇒ a∗1 constant — U 0 (·) convex: RHS rises ⇒ a∗1 increases — U 0 (·) concave: RHS falls ⇒ a∗1 decreases • introspection: a∗1 increases ⇒ U 0 (·) is convex U 000 > 0 • CRRA: U 0 (c) = c−σ for σ > 0 is convex • somewhat unavoidable: U 0 (c) > 0 and c ≥ 0 ⇒ U 0 (c) strictly convex near 0 and ∞ 4.2 Longer Horizon • i.i.d. income shocks T =∞ 6 • Bellman equation V (x) = max {u (x − a0 ) + βEV (Ra0 + ỹ)} • FOC from Bellman u0 (c) = βR EV 0 (Ra + ỹ) • again: V 0 convex → precautionary savings • but V 000 endogneous! • result: u000 > 0 then v000 > 0 (Sibley, 1975) 4.3 CARA • CARA preferences u (c) = − exp (−γc) V (x) = max {u (x − a0 ) + βEV (Ra0 + ỹ)} a • no borrowing constraints (except No-Ponzi) no non-negativity for consumption • guess and verify: where λ ≡ V (x) = Au (λx) r 1+r • note with CARA u (a + b) = −u (a) u(b) • verifying V (x) = max {u (x − a0 ) + βAEu (λ (Ra0 + ỹ))} µ ¶ ½ µ µ ¶¶ µ µ ¶ ¶¾ r 1 1 r 0 0 V (x) = −u x max u − a − x + βAEu r a − x + ỹ 1+r R R R ³r ´ n ³ ´o r V (x) = −u x max u (−α0 ) + βAEu rα0 + ỹ R R 7 where α0 = a0 − x/R or equivalently c= r x − α0 1+r confirms guess. Solving for A : ³ n r ´o A = max u (−α0 ) + βAEu rα0 + ỹ R ³ r ´ 0 u (−α ) = rβAEu rα + ỹ R ³ r ´ 0 0 u (−α ) = rβAEu rα + ỹ R 0 0 0 where we used u0 (c) = −γu(c) ³ r ´ 0 A = u (−α ) + βAEu rα + ỹ R 0 ) u (−α 1 +r = u (−α0 ) + =− u (−α0 ) r r 0 (note A > 0) coming back... ³ 1+r r ´ (−u (−α0 )) Eu rα0 + ỹ r R ³r ´ ỹ u (−rα0 ) = β (1 + r) Eu R ³ r ´´ 1 −1 ³ −α0 = u ỹ β (1 + r) Eu r R 1 −1 1 ³r ´ = u (β (1 + r)) + Eu ỹ r r R u (−α0 ) = rβ • verifying c(x) = λx + α using Euler... u0 (ct ) 1 1 1 = = = = βR Et u0 (ct+1 ) βR Et u0 (ct+1 − ct ) βR Et u0 (c (xt+1 ) − c(xt )) βR Et u0 (λ (xt+1 − xt )) since xt+1 = Rat+1 + yt+1 and at+1 = α0 + xt /R xt+1 − xt = R (α0 + xt /R) + yt+1 − xt = Rα0 + yt+1 8 ³ ´ r 1 = βR Et u0 rα0 + yt+1 R same as before • Verifying value function (again) note that u0 (c) = −γu(c) u0 (ct ) = βR Et u0 (ct+1 ) ⇐⇒ u (ct ) = βR Et u (ct+1 ) Et u (ct+1 ) = (βR)−t u (ct ) Then welfare ⇐⇒ current consumption: Vt ≡ ∞ X β t Et u (ct+s ) = s=0 ∞ X β t (βR)−t u (ct ) = s=0 Verifying ct = λxt ⇐⇒ V (x) = 1+r u (ct ) r 1+r u (λx − α0 ) r • consumption function ∙ ¸ 1 ∗ 1 log (β (1 + r)) c (x) = λ x + y − r r γ y∗ ≡ 1 −1 u [Eu (λy)] λ • suppose βR = 1 no CEQ... ...but simple deviation: constant y ∗ • CARA, the good: — tractable — useful benchmark → helps understand other cases — good for aggregation (linearity) • CARA, the bad: — negative consumption — unbounded inequality 9 5 Income Fluctuation Problem • iid income yt • ct ≥ 0 • borrowing constraints 5.1 Borrowing Constraints: Natural and Ad Hoc • natural borrowing constraint maximize borrowing given ct ≥ 0 ct ≥ 0 + No-Ponzi ⇒ at ≥ − ymin r • ad hoc borrowing constraint: at ≥ −φ φ = min{ymin /r, b} • Bellman {u (x − a0 ) + βEV (Ra0 + ỹ)} V (x) = max 0 a ≥−φ • change of variables ât = at + φ and ât+1 ≥ 0 zt = Rât + yt − rφ zt = ât + ct • transformed problem {u (z − aˆ0 ) + βEv (Râ0 + ỹ − rφ)} v (z) = max 0 â ≥0 (droppingbnotation) {u (z − a0 ) + βEv (Ra0 + ỹ − rφ)} v (z) = max 0 a ≥0 10 5.2 Properties of Solution βR= 1 • CARA: E [at+1 ] > at and E [ct+1 ] > ct • Martingale Convergence Theorem: If xt ≥ 0 and xt ≥ E [xt+1 ] then xt → x˜ (note: x˜ < ∞ a.e.) • Euler u0 (ct ) = βRE [u0 (ct+1 )] ⇒ u0 (ct ) converges ⇒ ct → c • if c < ∞ contradiction with budget constraint equality • at → ∞ and ct → ∞ βR< 1 • Bellman equation {u (x − a0 ) + βEv (Ra0 + ỹ − rφ)} v (z) = max 0 a • v is increasing, concave and differentiable • Preview of Properties — monotonicity of c (z) and a0 (z) — borrowing constraint is binding iff z ≤ z ∗ — if u00 (c) =0 c→0 u0 (c) lim then assets bounded 11 — if u ∈ HARA class ⇒ c (z) is concave (Carrol and Kimball) Figures removed due to copyright restrictions. See figures Ia and Ib on p. 667 in Aiyagari, S. Rao. "Uninsured Idiosyncratic Risk and Aggregate Savings." Quarterly Journal of Economics 109, no. 3 (1994): 659-684. • borrowing constraints — certainty: [0, z ∗ ] large approached monotonically — uncertainty: [0, z ∗ ] relatively small not approached monotonically • concavity of v ⇒ concavity of Φ (a0 ) = βEv (Ra0 + ỹ) ⇒ standard consumption problem with two normal goods {u (c) + © (a0 )} v (z) = max 0 c,a c + a0 ≤ x a0 ≥ 0 ⇒ c (z) and a0 (z) are increasing in z 12 • FOC (Euler) u0 (x − a0 ) ≥ βREv0 (Ra0 + ỹ) equality if a0 > 0 • define u0 (z ∗ ) = βR Ev0 (ỹ) z ≤ z∗ ⇒ c=z ⇒ a0 = 0 Assets bounded above • not a technicality... ...remember CARA case • idea: take a → ∞ income uncertainty unrelated to a (i.e. absolute risk) −u00 u0 → 0 ⇒ income uncertainty unimportant βR bites ⇒ a0 < a falls Proof 0 = (1 + r) a0 (z) + ymax ≤ z for z ≥ z ∗ exist a z ∗ such that zmax Euler Eu0 (c (z 0 )) 0 u (c̄ (z)) u0 (c (z)) = β (1 + r) 0 u (c̄ (z)) 0 (z)) = c (a0 (z) + ymax − rφ) where c̄ (z) = c (zmax E [u0 (c (z 0 ))] = 1 ⇒ DONE z→∞ u0 (c̄ (z)) IF lim u0 (c (z)) u0 (c̄ (z) − (c̄ (z) − c (z))) Eu0 (c (z 0 )) ≥ 0 ≥ 1≥ 0 u (c̄ (z)) u (c̄ (z)) u0 (c̄ (z)) since a0 is increasing c̄ (z) − c (z) = c (Ra0 (z) + ymax − rφ) − c (Ra0 (z) + ymin − rφ) < ymax − ymin 13 1≥ u0 (c̄ (z) − (ymax − ymin )) Eu0 (c (z 0 )) ≥ u0 (c̄ (z)) u0 (c̄ (z)) Since z → ∞ ⇒ a0 (z) , c (z) → ∞ then c̄ (z) = c (a0 (z) + ymax − rφ) → ∞. Apply Lemma below. ¥ Lemma. for A > 0 u0 (c − A) →1 u0 (c) Proof. 1 ≤ u0 (c − A) = 1+ u0 (c) Z A u00 (c − s) ds u0 (c) 0 Z A 0 u (c − s) −u00 (c − s) = 1− ds u0 (c) u0 (c − s) 0 Z A 0 u (c − s) = 1− γ (c − s) ds u0 (c) 0 Z A ≤ 1− γ (c − s) ds 0 since u0 (c−s) u0 (c) > 1 for all t > 0 Z 0 so 6 u0 (c−A) u0 (c) A γ (c − s) ds → 0 → 1. ¥ Lessons from Simulations From Deaton’s “Saving and Liquidity Constraints” (1991) paper: • important borrowing constraint may bind infrequently (wealth endogenous) • marginal propensity to consume higher than in PIH 14 Figure removed due to copyright restrictions. See Figure 1 on p. 1228 in Deaton, Angus. “Saving and Liquidity Constraints.” Econometrica 59, no. 5 (1991): 1221-1248. 15 Figure removed due to copyright restrictions. See Figure 2 on p. 1230 in Deaton, Angus. “Saving and Liquidity Constraints.” Econometrica 59, no. 5 (1991): 1221-1248. 16 Figure removed due to copyright restrictions. See Figure 4 on p. 1234 in Deaton, Angus. “Saving and Liquidity Constraints.” Econometrica 59, no. 5 (1991): 1221-1248. 17 • consumption — smoother temporary shocks — harder with permanent shocks 7 Invariant Distributions • initial distribution F0 (z0 ) • laws of motion z 0 = Ra0 (z) + y 0 generate F0 (z0 ) → F1 (z1 ) F1 (z1 ) → F2 (z2 ) .. . • steady state: invariant distribution F (z) → F (z) • result: 1. exists 2. unique 3. stable • key: bound on assets and monotonicity • A (r) ≡ E (a0 (z)) — continuous — not necessarily monotonically increasing in r income vs. substitution; and w(r) effect typically: monotonically increasing — A (r) → ∞ as R → β −1 18 8 General Equilibrium • GE effects of precuationary savings? →more k, lower r • how much? 8.1 Huggett: Endowment • endowment economy • no government • zero net supply of assets • idea: any precuationary saving translates to lower equilibrium interest rate • computational GE exercise: — CRRA preferences — borrowing constraints 8.2 Aiygari • adds capital • yt = wlt and lt is random; w is economy-wide wage P • N is given by N = li pi • define steady state equilibrium: 3 equations / 3 unknowns: (K, r, w) Z A (z, r, w) dF (z; r, w) − φ = K r = Fk (K, N) − δ w = FN (K, N) 19 • solve w (r) and substitute: Z GE A (r) = a (z, r, w (r)) dμ (z; r, w (r)) = K intersect with r = Fk (K, N) − δ • AGE (r) — continuous — not necessarily monotonically increasing in r (a) income vs. substitution; (b) w(r) effect typically: monotonically increasing — A (r) → ∞ as R → β −1 Figures removed due to copyright restrictions. See Figures IIa and IIb on p. 668 in Aiyagari, S. Rao. "Uninsured Idiosyncratic Risk and Aggregate Savings." Quarterly Journal of Economics 109, no. 3 (1994): 659-684. • comparative statics — ∂ A (0, b) ∂b >0 ∂ A (r, b) > 0 typically: ∂b 20 — ↑ σ 2y ⇒ ↑ A Table removed due to copyright restrictions. See Table II on p. 678 in Aiyagari, S. Rao. "Uninsured Idiosyncratic Risk and Aggregate Savings." Quarterly Journal of Economics 109, no. 3 (1994): 659-684. • wealth distribution: not as skewed • transition? monotonic? 9 Inequality • CEQ-PIH and CARA inequality increases linearly unbound inequality • CRRA inequality increases initially bounded inequality 21 Figure removed due to copyright restrictions. See Figure 2 on p. 444 in Deaton, Angus, and Christina Paxson. "Intertemporal Choice and Inequality." Journal of Political Economy 102, no. 3 (1994): 437-467. 22 Figure removed due to copyright restrictions. See Figure 4 on p. 445 in Deaton, Angus, and Christina Paxson. "Intertemporal Choice and Inequality." Journal of Political Economy 102, no. 3 (1994): 437-467. 23 Figure removed due to copyright restrictions. See Figure 6 on p. 450 in Deaton, Angus, and Christina Paxson. "Intertemporal Choice and Inequality." Journal of Political Economy 102, no. 3 (1994): 437-467. 24 Figure removed due to copyright restrictions. See Figure 1d) on p. 769 in Heathcote, Jonathan, Kjetil Storesletten, and Giovanni L. Violante. "Two Views of Inequality Over the Life-Cycle." Journal of the European Economic Association 3, nos. 2-3 (2005): 765-775. Deaton and Paxson Revisionisist (Heathcoate, Storesletten, Violante) Guvenen Storesletten, Telmer and Yaron: 10 Life Cycle: Consumption tracks Income Carroll and Summers: 11 Other Features and Extensions • Social Security: Hubbard-Skinnner-Zeldes (1995): “Precautionary Savings and Social Security” Scholz, Seshadri, and Khitatrakun (2006): “Are Americans Saving "Optimally" for Retirement?” • Medical Shocks: Palumbo (1999) 25 Figure removed due to copyright restrictions. See Figure 1 in Guvenen, Fatih. "Learning Your Earning: Are Labor Income Shocks Really Very Persistent?" American Economic Review. (Forthcoming) http://www.econ.umn.edu/~econdept/learning_your_earning.pdf Figure 9 Figure removed due to copyright restrictions. See Figure 1 on p. 613 in Storesletten, Kjetil, Chris Telmer, and Amir Yaron. "Consumption and Risk Sharing over the Life Cycle." Journal of Monetary Economics 51, no. 3 (2004): 609-663. Figure 10 26 Figure removed due to copyright restrictions. See Figure 5 on p. 624 in Storesletten, Kjetil, Chris Telmer, and Amir Yaron. "Consumption and Risk Sharing over the Life Cycle." Journal of Monetary Economics 51, no. 3 (2004): 609-663. Figure 11 Figure removed due to copyright restrictions. Figure 12 27 Figure removed due to copyright restrictions. Figure 13 Figure removed due to copyright restrictions. Figure 14 28 Figure removed due to copyright restrictions. Figure 15 • Learning Income Growth: Guvenen (2006) • Hyperbolic preferences: Harris-Laibson • Leisure Complementarity Aguiar-Hurst (2006): “Consumption vs. Expenditure” • Attanasio-Weber: Demographics and Taste Shocks 29