MACROECONOMICS LECTURE 7 Alexander Rubin HSE University in Saint-Petersburg September 30, 2022 Alexander Rubin (HSE) Macroeconomics September 30, 2022 1 / 64 Consumer Choice Consumers do not supply only labour. They also decide how much to consume and how much to save. Suppose that income of consumer is given {Yt }∞ t=0 . A part of this income comes from supply of labour, but it also can include other categories including financial wealth. Then we want to know how this income is divided on consumption and savings. Savings is a supply of loanable funds in the economy. It’s a source of investments which forms capital of the firms. Alexander Rubin (HSE) Macroeconomics September 30, 2022 2 / 64 Keynesian Consumption Function The most simple way to model consumer choice is to use Keynesian consumption function. C = Ca + M P C(Y − T ), where C – consumption, Ca – autonomous consumption, M P C ∈ (0, 1) – marginal propensity of consume, Y – income, T – lump-sum tax. Here consumption (and savings) depends only on current incomes. Alexander Rubin (HSE) Macroeconomics September 30, 2022 3 / 64 Keynesian Consumption Function The idea had been expressed by John Meynard Keynes in General Theory of Employment, Interest and Money (1936). “the fundamental psychological law, upon which we are entitled to depend with great confidence both a priori from our knowledge of human nature and from the detailed facts of experience, is that men are disposed, as a rule and on the average, to increase their consumption as their income increases but not by as much as the increase in the income”. Alexander Rubin (HSE) Macroeconomics September 30, 2022 4 / 64 Keynesian Function: Graph C C = Ca + M P C(Y − T ) 1 The graph for Keynesian function starts above zero because of the autonomous consumption. 2 The slope is less than one because a part of consumer’s income is saved. 3 But what the data says? 45o Y Alexander Rubin (HSE) Macroeconomics September 30, 2022 5 / 64 Keynesian Function: Over Time C 1 If use use data within a country over time, then aggregate consumption is proportional to aggregate income. 2 It differs from the estimation with cross-sectional data. Long Run 45o Y Alexander Rubin (HSE) Macroeconomics September 30, 2022 6 / 64 Keynesian Function: Across Groups C Whites 1 Estimations of consumption function give different results for different groups: e.g. blacks and whites. 2 It depends on the average income which is more or less constant within a group. 3 Permanent income varies substantially across groups. Blacks 45o Y Alexander Rubin (HSE) Macroeconomics September 30, 2022 7 / 64 Permanent Income Hypothesis Friedman, M. (1957). The permanent income hypothesis. In A theory of the consumption function (pp. 20-37). Princeton University Press. See also Romer’s textbook, Chapter 8 “Consumption”: Romer, D. (2012). Advanced Macroeconomics, 4e. New York: McGraw-Hill. Alexander Rubin (HSE) Macroeconomics September 30, 2022 8 / 64 Permanent Income Hypothesis: Model Total consumer’s income is divided into permanent Y P and transitory Y T part. Transitory income is zero-mean and uncorrelated with permanent part: Yit = YiP + YitT Consumption depends mostly on the permanent part: Cit = YiP + eit Transitory consumption eit is also zero-mean and uncorrelated with permanent income. Both transitory income and consumption are unpredictable. Alexander Rubin (HSE) Macroeconomics September 30, 2022 9 / 64 Permanent Income Hypothesis: OLS Now let’s estimate linear regression model using OLS: Ci = a + bYi + ui After minimizing the sum of squared errors we get: b̂ = cov(C, Y ) cov(Y P + e, Y P + Y T ) var(Y P ) = = var(Y ) var(Y P + Y T ) var(Y P ) + var(Y T ) â = C̄ − b̂Ȳ = (1 − b̂)Y P We see that b̂, aka M P C, depends on the proportion of variance permanent income in total income variance. Alexander Rubin (HSE) Macroeconomics September 30, 2022 10 / 64 Permanent Income Hypothesis: OLS If variation in transitory income YitT is small enough, then: b̂ → 1 â → 0 Over long period of time variation in permanent income plays the major role. Therefore, change in come trnslates into change in consumption almost one-to-one. The relative variances of permanent and transitory income are similar for the groups but average incomes are different. Therefore, consumption also different. Alexander Rubin (HSE) Macroeconomics September 30, 2022 11 / 64 Intertemporal Choice Marginal propensity of consume is not a structural parameter. It depends on interest rates and other macreconomic conditions including future anticipated income. We need to endogenize M P C. Economic agents live in time and they have to decide how much to consume, save, invest and work taking the long-run perspective into the account, i.e. make intertemporal choice. Therefore, they decide how much to consume today and tomorrow. As in case of labour market consumer sovles optimization problem. This problem is now dynamic. Alexander Rubin (HSE) Macroeconomics September 30, 2022 12 / 64 Discounting To study intertemporal choice we need to use a concept of discounting. We need to compare two types of values: monetary values, e.g. income and consumption utilities from consumption. Therefore, we need two distinct ways to compare them. Alexander Rubin (HSE) Macroeconomics September 30, 2022 13 / 64 Discounting of Monetary Flows: One Period Assume that agent has a right for some monetary value CF which will arrive at the end of period. The alternative is to get it right now as S and invest at some interest rate r. It means that opportunity costs of waiting are rS and consumer will be indifferent between these two options if: S(1 + r) = CF This is no-arbitrage condition again. Then discounted value, or presetn value of CF is equal to: S = DV = Alexander Rubin (HSE) CF 1+r Macroeconomics September 30, 2022 14 / 64 Discounting of Monetary Flows: Many Periods Now assume that agent waits for a flow of payments CFt , t = 0, 1, 2, ...T during time T . By the same argument we can write for some t: S(1 + r)t = CFt , t = 0, 1, 2, ...T Then we discount the future value of CFt in the following way: S = DV = CFt , t = 0, 1, 2, ...T (1 + r)t Again, consumer is indifferent between these two options if the equality above holds. Alexander Rubin (HSE) Macroeconomics September 30, 2022 15 / 64 Discounting of Monetary Flows: Cash Flow The present, or discounted, value of the whole flow CFt , t = 0, 1, 2, ...T is equal to: T X CFt DV = (1 + r)t+1 t=0 If payments arrive at the beginning of the period, then the formula is slightly different: T X CFt DV = (1 + r)t t=0 Alexander Rubin (HSE) Macroeconomics September 30, 2022 16 / 64 Compounding Monetary Values The inverse operation to discounting is called compounding. If helps to compute future value of present monetary holdings at the end of the period t: CF = F V = S(1 + r)t Alexander Rubin (HSE) Macroeconomics September 30, 2022 17 / 64 Comparing Monetary Flows Now we can make comparisons of monetary flow over time. The rule is that one should discount all the payments taking their values and timing into the account and them sum them up and compare. Alexander Rubin (HSE) Macroeconomics September 30, 2022 18 / 64 Console Bond Consider an infinite flow of equal payments CF = CF0 = CF1 = CF2 = ... Then we can obtain simple formula for present value. Use the fact the we have geometric series with factor q ∈ (0, 1): ! ∞ X CF 1 1 1 = CF = DV = CF 1 t+1 (1 + r) 1 + r 1 − 1+r r t=0 If payments arrive at the beginning of every period, then: ! ∞ X 1 1 1+r DV = CF = CF = CF 1 t (1 + r) r 1 − 1+r t=0 Alexander Rubin (HSE) Macroeconomics September 30, 2022 19 / 64 Subjective Discount Factor Subjective discount factor is a number between zero and one: β ∈ (0, 1). It characterizes relative utility of tomorrow and today consumption: β= u(Ct+1 ) , u(Ct ) where u(·) – utility function at time t and Ct – aggregate consumption at time t. Discount factor is a deep, or structural, parameter that characterizes consumer’s behavior. It doesn’t depend on government policy interventions while M P C does. Alexander Rubin (HSE) Macroeconomics September 30, 2022 20 / 64 Subjective Discount Rate We also can imagine (just imagine!) that discount factor can be computed using some discount rate ρ: β= 1 1+ρ It follows that “discount rate” ρ can be computed as follows: ρ= 1 −1 β It’s easy to see that ρ > 0. Alexander Rubin (HSE) Macroeconomics September 30, 2022 21 / 64 Budget Constraint: First Period Now assume that agent lives for two periods: t = 0, 1 and receives incomes Y0 and Y1 . Then budget constraint in the first period is: C0 + B1 ≤ Y0 + B0 where B0 – initial financial wealth, C0 is consumption in the first period; B0 are net assets in the first period. B0 < 0 corresponds to net borrowing and B0 > 0 net lending (saving). Alexander Rubin (HSE) Macroeconomics September 30, 2022 22 / 64 Budget Constraint: Second Period Assume that agents borrow and save at the same interest rate r. Then budget constraint in the second period is: C1 + B2 ≤ Y1 + (1 + r)B1 Assuming monotonicity of preferences we conclude that both budget constraints hold as equalities. Alexander Rubin (HSE) Macroeconomics September 30, 2022 23 / 64 Transversality Condition B2 ≤ 0 because it’s optimal to consume everything at the end (agent doesn’t derive any utility from savings or leaving bequest). This is transversality condition. B2 ≥ 0 because we prohibit Ponzi scheme, when consumer borrows to consume more and then repay principal and interest with new borrowings. This is No-Ponzi-game condition. Transvesality condition prevents overaccumulation of wealth, while No-Ponzi-game condition prevents overaccumulation of wealth. Two ineqaulities together give B2 = 0. Alexander Rubin (HSE) Macroeconomics September 30, 2022 24 / 64 Intertemporal Budget Constraint Express B1 from the first budget constraint: B 1 = B 0 + Y0 − C 0 Now substitute it into the second budget constraint: C1 = Y1 + (1 + r)(B0 + Y0 − C0 ) Divide by (1 + r) and rearrange terms: C0 + Alexander Rubin (HSE) Y1 C1 = B0 + Y0 + 1+r 1+r Macroeconomics September 30, 2022 25 / 64 Intertemporal Budget Constraint: General Case Here we see that discounted money value of consumption is no more than discounted value of income. C0 + C1 Y1 = B0 + Y0 + 1+r 1+r This budget constraint can be generalized for any number of periods: T X t=0 T X Yt Ct = B + 0 (1 + r)t (1 + r)t t=0 We can also consider the infinite horizon (T → ∞) if series on the leftand right-hand side converge. In other words, consumption and income do not grow fast. Alexander Rubin (HSE) Macroeconomics September 30, 2022 26 / 64 Intertemporal Budget Constraint: Intuition Here we see that discounted money value of consumption is no more than discounted value of income. C0 + Y1 C1 = B0 + Y0 + 1+r 1+r It means that current consumption is not necessary equal to current income. Rather consumer compares present values and may allocate his lifetime income over time. Gross interest rate 1 + r is a “price” of current consumption. Consumption of one unit today costs 1 + r units of tomorrow consumption. Alexander Rubin (HSE) Macroeconomics September 30, 2022 27 / 64 Intertemporal Budget Constraint: Graph C1 1 Let B0 = 0 for simplicity. 2 The ratio of “prices” is (1 + r). 3 The point (Y0 , Y1 ) is always on the budget line. 4 Budget line rotates around (Y0 , Y1 ) when r changes. C0 + Y1 Y0 Alexander Rubin (HSE) C1 Y1 = Y0 + 1+r 1+r C0 Macroeconomics September 30, 2022 28 / 64 Intertemporal Preferences Agent has to choose between today and tomorrow consumption given the subjective discount factor. Preferences are represented by utility function: U (C0 , C1 ) = u(C0 ) + βu(C1 ) Here u(·) is an instantaneous utility function, i.e. utility function for only one period of time. We assume that preferences are stable over time and only β makes difference. Alexander Rubin (HSE) Macroeconomics September 30, 2022 29 / 64 Intertemporal Preferences: General Case Consumer’s preferences can be generalized for any periods of time: U (C0 , C1 , ...CT ) = T X β t u(Ct ) t=0 We can also consider the infinite horizon (T → ∞) if series β t u(Ct ), t = 0, 1, 2, .. converge. Alexander Rubin (HSE) Macroeconomics September 30, 2022 30 / 64 Consumer’s Problem: Two Periods Consider two-period model of consumption: max U (C0 , C1 ) = U (C0 ) + βU (C1 ) s.t. C0 ,C1 C0 + Y1 C1 = B 0 + Y0 + 1+r 1+r Lagrange function for this problem: C1 Y1 max L = U (C0 ) + βU (C1 ) + λ C0 + − B0 − Y0 − C0 ,C1 ,λ 1+r 1+r Alexander Rubin (HSE) Macroeconomics September 30, 2022 ! 31 / 64 Consumer’s Problem: Solution First-order conditions for the problem are: U ′ (C0 ) + λ = 0 λ =0 βU ′ (C1 ) + 1+r Exclude λ and write: U ′ (C0 ) = β(1 + r) U ′ (C1 ) This is known as Keynes-Ramsey rule, or Euler equation. Alexander Rubin (HSE) Macroeconomics September 30, 2022 32 / 64 Consumer’s Problem: Logarithmic Utility Consider two-period model of consumption: max U (C0 , C1 ) = log C0 + β log C1 C0 ,C1 C0 + s.t. Y1 C1 = B 0 + Y0 + 1+r 1+r Lagrange function for this problem: C1 Y1 max L = log C0 + β log C1 + λ C0 + − B 0 − Y0 − C0 ,C1 ,λ 1+r 1+r Alexander Rubin (HSE) Macroeconomics September 30, 2022 ! 33 / 64 Consumer’s Problem: Lagrangian Fist-order conditions are: 1 +λ=0 C0 β λ + =0 C1 1 + r Exclude λ and write C1 1+r = β(1 + r) = C0 1+ρ Again, this is Keynes-Ramsey rule for the logarithmic utility function. Alexander Rubin (HSE) Macroeconomics September 30, 2022 34 / 64 Keynes-Ramsey Rule Keynes-Ramsey Rule relates consumption values in two adjacent periods of time. Therefore, it brings dynamics into models. We can generalize it to the situation when consumer lives more than two periods t = 1, 2, ...T . Even infinite number of them t = 1, 2, ... 1+r Ct+1 = β(1 + r) = Ct 1+ρ Intuitively, this means that consumption/savings choice depends on the preferences, interest rates and this choice is made over the whole horizon, i.e. agents are forward-looking. Alexander Rubin (HSE) Macroeconomics September 30, 2022 35 / 64 Consumer’s Problem: Graphical Solution ‘ Net borrower (C0∗ ≥ Y0 ) Net lender (C0∗ ≤ Y0 ) C1 C1 Y1 C1∗ C1∗ Y1 C0∗ Alexander Rubin (HSE) Y0 C0 Macroeconomics Y0 C0∗ September 30, 2022 C0 36 / 64 Increase in Interest Rate C1 1 Rational consumer obeys WARP. 2 If agent is willing to save under lower interest rate he will save under higher one. 3 Net lender remains net lender and may lend more. 4 Net borrower may become net lender or borrow less. Y1 Y0 Alexander Rubin (HSE) C0 Macroeconomics September 30, 2022 37 / 64 Decrease in Interest Rate C1 Y1 Y0 Alexander Rubin (HSE) 1 Rational consumer obeys WARP. 2 If agent is willing to borrow under higher interest rate he will borrow under lower. 3 Net borrower remains net borrower and may borrow more. 4 Net lender may become net borrower or lend less. C0 Macroeconomics September 30, 2022 38 / 64 Current Consumption Consider Keynes-Ramsey Rule for logarithmic utility and combine it with budget constraint. C1 = β(1 + r) C0 C1 Y1 C0 + = B0 + Y0 + 1+r 1+r Substitute C2 from the rule into budget constraint: C0 + C0 β(1 + r) Y1 = C0 (1 + β) = B0 + Y0 + 1+r 1+r Or finally we have: 1 Y1 C0 = B0 + Y0 + 1+β 1+r Alexander Rubin (HSE) Macroeconomics ! September 30, 2022 39 / 64 Marginal Propensity to Consume It’s easy to see that 0 < 1/(1 + β) < 1. Therefore, only a part of future income is consumed today. ! Y1 1 B0 + Y0 + C0 = 1+β 1+r Now assume that income Y1 increases by some ∆Y1 . Then only a part of this increase will be consumed: ∆C0 = 1 ∆Y0 1+β Factor 1/(1 + β) is marginal propensity to consume. It’s a share of current income spent to consumption. Alexander Rubin (HSE) Macroeconomics September 30, 2022 40 / 64 Solow Revisited Compare intertemporal budget constraint with the one from the Solow model: Ct + Bt+1 = Bt (1 + rt ) + wt Lt | {z } Yt In case of Solow model income is not purely exogenous. Agent takes wage rate wt as given and supplies labour inelastically, but at the end wage rate is determined at the equilibrium inside the model. The only difference is that we allowed variable interest rate. Alexander Rubin (HSE) Macroeconomics September 30, 2022 41 / 64 Intertemporal Budget Constraint: First Step Take first two constraints: C0 + B1 ≤ B0 + Y1 C1 + B2 ≤ B1 (1 + r) + Y2 Divide the second constraint by 1 + r and sum them up. B1 cancels out and we get: B2 Y1 C1 + ≤ B 0 + Y0 + C0 + 1+r 1+r 1+r Alexander Rubin (HSE) Macroeconomics September 30, 2022 42 / 64 Intertemporal Budget Constraint: Second Step Now add the third constraint: C0 + B2 Y1 C1 + ≤ B 0 + Y0 + 1+r 1+r 1+r C2 + B3 ≤ B2 (1 + r) + Y2 Divide the third constraint by (1 + r)2 and sum them up. B2 /(1 + r) cancels out and we get: C0 + C1 C1 B2 Y1 Y2 + + ≤ B0 + Y0 + + 2 2 1 + r (1 + r) (1 + r) 1 + r (1 + r)2 Alexander Rubin (HSE) Macroeconomics September 30, 2022 43 / 64 Intertemporal Budget Constraint: Other Steps We can continue in the same way and get: ∞ X t=0 ∞ X Yt Bt+1 Ct + lim ≤ B0 + t t t→∞ (1 + r) (1 + r) (1 + r)t t=0 Again we have to restrict Bt+1 to prevent overaccumulation of debt and call it No-Ponzi-game condition: lim t→∞ Bt+1 ≥0 (1 + r)t Intuitively, it means that debt should not grow too fast (exponentially). What it it does? Alexander Rubin (HSE) Macroeconomics September 30, 2022 44 / 64 Infinite Horizon Model: Maximization Problem Consumer maximizes: max ∞ {Ct }t=0 U (C) = ∞ X β t u(ct ) s.t. t=0 Ct + Bt+1 ≤ Bt (1 + r) + Yt , t = 0, 1, 2, ... Lagrange function for this problem: max Ct ,Bt ,λ L= Alexander Rubin (HSE) ∞ X t=0 β t u(ct ) + ∞ X λt (Bt (1 + r) + Yt − Ct − Bt+1 ) t=0 Macroeconomics September 30, 2022 45 / 64 Infinite Horizon Model: First-Order Condition First-order condition for consumption: β t u′ (Ct )λt = 0 β t+1 u′ (Ct+1 )λt+1 = 0 First-order condition for savings: λt (1 + r) − λt+1 = 0 If follows that: u′ (Ct ) λt β t u′ (Ct ) = = =1+r t+1 ′ ′ β u (Ct+1 ) βu (Ct+1 ) λt+1 Alexander Rubin (HSE) Macroeconomics September 30, 2022 46 / 64 Infinite Horizon Model: Solution Again we have Euler equation: u′ (Ct ) = β(1 + r) u′ (Ct+1 ) For example, for linear-quadratic utility u(Ct ) = Ct − 2b Ct2 : 1 − bCt = β(1 + r) 1 − bCt+1 We use this result later to discuss Hall hypothesis. Alexander Rubin (HSE) Macroeconomics September 30, 2022 47 / 64 Consumption Smoothing Let’s assume for a moment that β(1 + r) = 1. Then C0 = C1 Or in general case: C0 = C1 = ... = Ct = ... Consumption smoothing happens even if β(1 + r) ̸= 1. Consumers doesn’t like sharp fluctuations in consumption. Consumption smoothing is positive if we expect such behavior from rational consumer. Consumption smoothing is normative because it’s rational to do so. Alexander Rubin (HSE) Macroeconomics September 30, 2022 48 / 64 Consumption Smoothing: Graph u(C) 1 Special case β(1 + r) = 1. 2 Consumption of average C̄ brings more utility than average of utilities. 3 The result follows from the concavity of instantaneous utility function u(·). 4 See also Jensen inequality. u(C2 ) u(C̄) ū(C) u(C1 ) C1 Alexander Rubin (HSE) C̄ C2 C Macroeconomics September 30, 2022 49 / 64 Linear-Quadratic Preferences Assume linear-quadratic preferences: b U (Ct ) = Ct − Ct2 2 Then utility maximization gives us: 1 − bC0 = β(1 + r) 1 − bC1 Assuming that β(1 + r) = 1 we get: C0 = C1 This is a version of Friedman (1957) and Hall (1978) Permanent Income Hypothesis. Alexander Rubin (HSE) Macroeconomics September 30, 2022 50 / 64 Linear-Quadratic Preferences: Uncertainty Now assume that future income is uncertain and consumer maximizes expected utility: " !# b 2 b 2 max U = E0 C0 − C0 + β C1 − C1 s.t. C0 ,C1 2 2 C0 + C1 Y1 = Y0 + 1+r 1+r Income is arbitrary stationary process. On can show that in this case: C0 = E0 [C1 ] Or generally speaking: Ct = Et [Ct+1 ] Alexander Rubin (HSE) Macroeconomics September 30, 2022 51 / 64 Linear-Quadratic Preferences: Consumption Now take budget constraint and substitute Keynes-Ramsey rule: " # " # C0 Y1 E0 C0 + = E0 Y0 + 1+r 1+r Use properties of mathematical expectation: 1+r E[Y1 ] C= Y0 + 2+r 1+r ! In case of uncertainty consumption depends on expected income. Alexander Rubin (HSE) Macroeconomics September 30, 2022 52 / 64 PIH: Intertemporal Choice Model Assume that consumer lives infinitely long t = 1, 2, ... then: C ∞ X t=0 ∞ X Yt 1 = (1 + r)t (1 + r)t t=0 The sum of geometric series on the left-hand side is (1 + r)/r. C= ∞ r X Yt 1+r (1 + r)t t=0 This is permanent income and share of this income spent on consumption is equal to r/(1 + r) ∈ (0, 1). Alexander Rubin (HSE) Macroeconomics September 30, 2022 53 / 64 Permanent and Average Lifetime Income It’s easy to see that permanent income is a weighted average of the whole life-time income with weights 1/(1 + r)t , t = 0, 1, 2.... ∞ r X Yt C= 1+r (1 + r)t t=0 If case of uncertainty, permanent income is a weighted average of the expected income: # " ∞ X Yt r E0 C= 1+r (1 + r)t t=0 Alexander Rubin (HSE) Macroeconomics September 30, 2022 54 / 64 Permanent Shock Assume that shock affects permanent income changing it once and forever. This is permanent shock. Then new incomes are Yt + ∆Y : ∞ r 1+r r X ∆Y ∆Y = ∆Y = ∆C = 1+r (1 + r)t 1+r r t=0 Change in permanent income translates into change in income one-to-one. Alexander Rubin (HSE) Macroeconomics September 30, 2022 55 / 64 Transitory Shock Assume that shock affects transitory income in the first period, i.e. this is transitory shock. Then new income is Y1 + ∆Y : ∆C = r ∆Y 1+r Therefore, only a part of increase in income will be spent on consumption. Assume that r = 0.01, then (r/(1 + r) = 0.01 and only 1% of transitory income will be spent on consumption. The other part is saved. Alexander Rubin (HSE) Macroeconomics September 30, 2022 56 / 64 Anticipated Future Shock Assume that a consumer anticipates a permanent shock in the next period t = 1. For example, an income tax cut which increases permanent income. What will happen to consumption right now? ∆C = ∞ r 1 ∆Y r X ∆Y ∆Y = = t 1+r (1 + r) 1+rr 1+r t=1 Consumption is affected immediately before actual increase in income, but this increase is slightly smaller than before 1/(1 + r) ≈ 0.99. In case of unanticipated shock consumption will be affected next period only. Alexander Rubin (HSE) Macroeconomics September 30, 2022 57 / 64 Expected and Unexpected Shocks It follows that consumers react differently on different types of shocks. Anticipated shocks affect consumption and savings before the shock happens while unanticipated shocks affect only when it occurs. Permanent shocks mostly affect consumption while transitory shocks mostly affect savings. Alexander Rubin (HSE) Macroeconomics September 30, 2022 58 / 64 The Role of Expectations We saw that reaction of an agent depends on his expectations, i.e. the process of thinking about the future and ability to do it correctly. Different types of expectations may be embedded into a model: constant expectations, i.e. agent thinks something about future and these thoughts may be correct but he doesn’t change his mind with time in response to any shocks adaptive expectations, i.e. agent corrects his perception of future when new information arrives rational expectations, i.e. agent uses all the available information and builds correct prediction. In case of rational expectations an agent is forward-looking. In case of adaptive – backward-looking. Alexander Rubin (HSE) Macroeconomics September 30, 2022 59 / 64 Why PIH May not Work? In reality PIH may be violated due to different reasons. 1 Bounded rationality, i.e. departure from rational model and/or from rational expectations. 2 Precautionary savings, i.e. additional savings due to risk-aversion. 3 Liquidity constraints, i.e. inability to lend or borrow. 4 Presence of durable goods in consumption bundle, i.e. we need to take into the account depreciation of durable goods instead on expenditures on them. 5 Habit formation, i.e. proclivity to consume the same amount. Alexander Rubin (HSE) Macroeconomics September 30, 2022 60 / 64 Precautionary Savings Take Euler equation for general (non-quadratic) utility fiunction: u′ (C0 ) = β(1 + r) E0 [u′ (C1 )] Assume again that β(1 + r) = 1. Then: u′ (C0 ) = E0 [u′ (C1 )] It’s easy to see that if expectation increases, then consumption decreases, because marginal utility is a decreasing function: ↑ E0 [u′ (C1 )] =⇒ ↓ C0 Let u′′′ > 0 and variance in consumption increases while expected value remains the same. Alexander Rubin (HSE) Macroeconomics September 30, 2022 61 / 64 Precautionary Savings: Graph u′ (C1 ) 1 Mean-preserving spread, i.e. expectation is the same, but variance is larger. 2 Marginal utility is concave, i.e. u′′′ > 0. 3 E0 [u′ (C1 )|σ1′ ] 4 E0 [u′ (C1 )|σ1 ] Today consumption decreases, i.e. agent makes precautionary savings. C1 E0 [C1 ] Alexander Rubin (HSE) E0 [u′ (C1′ )|σ1′ ] > E0 [u′ (C1 )|σ1 ]. Macroeconomics September 30, 2022 62 / 64 Liquidity Constraints: Not Bounding C1 1 Due to liquidity constraint consumer cannot borrow in the first period. 2 Red part of budget set is now unavailable for him. 3 Due to parameters and preferences liquidity constraint is not active and bounding. C1∗ Y1 C0∗ Alexander Rubin (HSE) Y0 C0 Macroeconomics September 30, 2022 63 / 64 Liquidity Constraints: Bounding C1 C1∗ = Y1 C0∗ = Y0 Alexander Rubin (HSE) 1 Due to liquidity constraint consumer cannot borrow in the first period. 2 Red part of budget set is now unavailable for him. 3 Due to parameters and preferences liquidity constraint is active and bounding consumer choice. 4 In the absence of liquidity constraint optimal consumption may be larger C0∗ > Y0 . C0 Macroeconomics September 30, 2022 64 / 64