Session 4: Money. Jean Imbs. November 2010

Transcription

1 Session 4: Jean November 2010

2 I So far, focused on real economy. Real quantities consumed, produced, invested. No money, no nominal in uences. I Now, introduce nominal dimension in the economy. First and foremost motivation is to be able to talk about long run in ation. I We talked a bit about in ation when we discussed time-consistent monetary policy. But these were uctuations of in ation, and how to trade them o against uctuations in output. I In reality, there is such a thing as long run in ation (2-3% now) - as well as de ations and hyperin ations. I is a puzzle in macro. Its roles are well known: helps in transactions (no double coincidence of needs), stores wealth, and a credit that can be borrowed and lent. I Surely however, other things exist that can ful ll all three roles. Why is it money that does? probably out of convenience.

3 Plan I The intertemporal budget constraint with money I Cash-in-advance: you must have cash to purchase goods and services I in the utility: money is, in e ect, a commodity, and holding it creates utility. I as an Intermediate Good: holding money economizes on shopping time I Transaction Cost: money economizes on the real transaction costs involved in consumption I Applications: Hyperin ations, The Friedman Rule and Superneutrality

4 The Budget Constraint I What we had previously could be written (in real terms, obviously): a t+1 + c t = x t + r t a t where a t denotes the stock of assets at time t, c t is consumption, x t is (exogenous) income and r t is the real rate of return on assets. Once again, everything is real. I Now think of the analogy with nominal variables. In particular, assume we have two types of assets that can be used to store wealth: money M t and bonds B t. In particular, B t denotes the total expenditures on one-period bonds, paying a nominal return R t. I The budget constraint becomes: B t+1 + M t+1 + P t c t = P t x t + R t B t where we wrote the constraint in nominal terms, with P t the aggregate price level.

5 Budget Constraint I How do we obtain the real budget constraint? simply de ate the previous expression, to obtain b t+1 (1 + π t+1 ) b t + m t+1 (1 + π t+1 ) m t + c t = x t + R t b t where b t = B t P t, m t = M t and we made use of the fact that B t+1 P t = B t+1 P t+1 P t+1 P t M t+1 P t = M t+1 P t+1 P t+1 P t I Rearrange P t B t P t = b t+1 (1 + π t+1 ) b t and M t P t = m t+1 (1 + π t+1 ) m t (1 + π t+1 ) ( b t+1 + m t+1 ) + c t = x t + (R t π t+1 ) b t π t+1 m t

6 Budget Constraint I By analogy with the budget constraint without money, we can set a t = b t + m t. The real return on bonds is 1+R t 1+π t+1 1 ' R t π t+1, and the real rate on money is π t+1 ; holding money has zero nominal return et hence money loses it purchasing power because of in ation. I The fall in the value of nominal money balances M t is in e ect a tax - an "in ation tax" - and it is "seignorage" income to the issuer of money, i.e. the government.

7 The Static Steady State I At the steady state, all real stocks are constant, including m and b. I So b = m = 0, but in ation π is not necessarily zero. I Suppose the growth rate of nominal money M M = µ, then m m = M M P P = µ π = 0 I So at the steady state µ = π. But no implication on the direction of causality - money growth equals in ation; but equally true that if in ation is targetted at π then that conditions money growth.

8 The Cash-in-Advance model I So far, money only entails costs (an in ation tax). So why hold it? Because it reduces transaction costs, provides store of value and a unit of account. I Cash-in-advance model (CIA) focuses on transactions demand. Simplest model of money: all goods and services must be paid for in full with cash. In particular, impose M D t = P t c t i.e. m D t = MD t P t = c t I supply assumed exogenous; equilibrium requires M D t = M S t = M t

9 Cash in Advance I Set up the dynamic problem: L t = s=0 fβ s U(c t+s ) + λ t+s [x t+s + (1 + R t+s ) b t+s + m t+s (1 + π t+s+1 ) (b t+s+1 + m t+s+1 ) c t+s ] +µ t+s [m t+s c t+s ] I First order conditions L t c t+s = β s U 0 (c t+s ) λ t+s µ t+s = 0 L t b t+s = λ t+s (1 + R t+s ) λ t+s 1 (1 + π t+s ) = 0 L t m t+s = λ t+s λ t+s 1 (1 + π t+s ) + µ t+s = 0

10 Cash in advance I Subtracting the second and third conditions: I Replace in the rst: λ t+s R t+s = µ t+s β s U 0 (c t+s ) = λ t+s (1 + R t+s ) I Now use the implied value for λ t+s in the second condition: I Rearrange and set s = 1: βu 0 (c t+1 ) U 0 (c t ) The standard Euler equation βu 0 (c t+s ) U 0 (c t+s 1 ) = 1 + π t+s 1 + R t+s R t = βu0 (c t+1 ) 1 + π t+1 U 0 (1 + r t+1 ) = 1 (c t )

11 Cash in advance I What about the dynamics of m and c? For sure, c t = m t at all points in time. So only need to solve for c t. I Assume for simplicity that R t and π t+1 are constant. Then the budget constraint implies (1 + π) (b t+1 + m t+1 ) + c t = x t + (1 + R) b t + m t I In terms of a = b + m: a t = R (c t x t + Rm t ) π a t R

12 Cash in advance I Which we can solve recursively to obtain a t = R + n 1 s=0 1 + π 1 + R 1 + π s (c t+s x t+s + Rm t+s ) 1 + R n a t+n I If r = R then and π > 0 (i.e. the real return on bonds is positive), a t = R lim n! s=0 1 + π n a t+n = R 1 + π s (c t+s x t+s + Rm t+s ) 1 + R

13 Cash in advance I For simplicity suppose c t and m t are at their long run equilibrium values, i.e. c t+s = c t and m t+s = m t. Then R a t = c t 1 + R R π R s=0 1 + π s x t+s + m t 1 + R R 1 + R 1 + R R π I or (R π) (b t + m t ) = c t R π 1 + R s=0 1 + π s x t+s + Rm t 1 + R

14 Cash in advance I Finally c t ' r 1 + r s=0 x t+s (1 + r) s + rb t πm t where we have used r ' R π. If in addition we assume x t+s = x t, then the consumption function becomes c t ' x t + rb t πm t Consumption falls with the stock of real money balances when in ation is positive. Having to pay for consumption using cash has introduced a non-neutrality into the economy. Why? because a nominal variable (in ation) a ects a real variable (consumption), through the in ation tax

15 Solving the CIA model using the Bellman equation I Now derive the same equilibrium using the value function. De ne ) V (b t 1, m t 1 ) = Max fct+s 1,m t+s,b t+s g 0 ( β s U(c t+s 1 ) s=0 subject to x t+s + (1 + R t+s ) b t+s + m t+s = and (1 + π t+s+1 ) (b t+s+1 + m t+s+1 ) + c t+s m t+s = c t+s

16 Solving the CIA model using the Bellman equation I Rewrite an unconstrained version: 8 >< V (b t 1 ) = Max fbt+s g 0 = Max bt ( >: β s 1 U( 1+π t+s x t+s 1 s=0 + 1+R t+s 1 1+π t+s b t+s 1 b t+s ) 1 U( 1+π t x t R t 1 1+π t b t 1 b t ) +βv (b t ) where we have made use of the fact that m t+s = c t+s in the budget constraint. In particular: π t+s+1 x t+s R t+s 1 + π t+s+1 b t+s b t+s+1 = c t+s+1 9 >= >; )

17 Solving the CIA model using the Bellman equation I The rst order condition writes: U 0 (c t ) = βv 0 (b t ) I The envelope theorem implies that I Update this once: V 0 (b t 1 ) = 1 + R t π t U 0 (c t ) V 0 (b t ) = 1 + R t 1 + π t+1 U 0 (c t+1 ) I And so back in the rst order condition: I QED. U 0 (c t ) = β 1 + R t 1 + π t+1 U 0 (c t+1 )

18 in the Utility I Model due to Sidrausky (1967). simply enters utility. I am happy to have cash in my pocket (though there s no reason why). I The transaction motive (comes next) o ers some justi cation for MIU. I Lagrangean becomes L t = s=0 fβ s U(c t+s, m t+s ) + λ t+s [x t+s + (1 + R t+s ) b t+s +m t+s (1 + π t+s+1 ) (b t+s+1 + m t+s+1 ) c t+s ]g

19 in the Utility I First order conditions L t c t+s = β s U c (c t+s, m t+s ) λ t+s = 0 L t b t+s = λ t+s (1 + R t+s ) λ t+s 1 (1 + π t+s ) = 0 L t m t+s = β s U m (c t+s, m t+s ) + λ t+s λ t+s 1 (1 + π t+s ) = 0

20 in the Utility I Subtracting the second and third conditions: I Replace in the rst: λ t+s R t+s = β s U m (c t+s, m t+s ) R t+s U c (c t+s ) = U m (c t+s, m t+s ) The right hand side re ects the marginal utility from holding an extra unit of real balances. The left hand side re ects the opportunity cost of doing so: namely not holding bonds, not earning interest on them, and not consuming the proceeds. I The second equation becomes βu c (c t+s, m t+s )(1 + R t+s ) = U c (c t+s 1, m t+s 1 ) (1 + π t+s ) I Rearrange and set s = 1: βu c (c t+1, m t+1 ) U c (c t, m t ) 1 + R t π t+1 = 1 The standard Euler equation - just slight timing di erence.

21 in the Utility I Condition on the marginal utilities of c and m has a natural interpretation. To see this, choose utility U(c t, m t ) = c1 t σ 1 m 1 σ 1 σ + η t 1 1 σ I Then the condition rewrites R t c σ t I Rewrite the demand for money: = ηm σ t 1 Rt σ m t = c t η demand increases in consumption (a transaction motive) and decreases in the interest rate (an opportunity cost). Reminiscent of the classic IS-LM: M P = L (Y, i).

22 MIU model using Bellman equation I As an exercise, derive the same equilibrium using the value function. De ne V (b t 1, m t 1 ) = Max fct+s 1,m t+s,b t+s g 0 ( β s U(c t+s 1, m t+s 1 ) s=0 ) subject to x t+s + (1 + R t+s ) b t+s + m t+s = (1 + π t+s+1 ) (b t+s+1 + m t+s+1 ) + c t+s I Do it as an exercise.

23 The Shopping Time Model I Model money as an intermediate good that is held to reduce shopping time: suppose total time (1) has to be allocated between work n t, leisure l t and shopping s t : n t + l t + s t = 1 I In addition, s t = S (c t, m t ) with S c > 0, S m < 0, S cc, S mm > 0 and S cm < 0. I Utility is derived from consumption and leisure (not money anymore) U = U (c t, l t ) = U [c t, 1 n t S (c t, m t )] which provides a direct motivation for the MIU framework.

24 The Shopping Time Model I Consider Lagrangean: L t = s=0 fβ s U(c t+s, l t+s ) + λ t+s [w t+s n t+s + (1 + R t+s ) b t+s +m t+s (1 + π t+s+1 ) (b t+s+1 + m t+s+1 ) c t+s ] +µ t+s [n t + l t + S (c t, m t ) 1]g

25 The Shopping Time Model I First order conditions L t c t+s = β s U c (c t+s ) λ t+s + µ t+s S c (c t, m t ) = 0 L t l t+s = β s U l (c t+s ) + µ t+s = 0 L t n t+s = λ t+s w t+s + µ t+s = 0 L t b t+s = λ t+s (1 + R t+s ) λ t+s 1 (1 + π t+s ) = 0 L t m t+s = λ t+s λ t+s 1 (1 + π t+s ) + µ t+s S m (c t, m t ) = 0

26 The Shopping Time Model I Subtracting the fourth and fth conditions: λ t+s R t+s = µ t+s S m (c t, m t ) I Now, use rst and second equations and substitute, for s = 0: [U c (c t ) U l (c t )S c (c t, m t )] R t = U l (c t )S m (c t, m t ) The right hand side re ect the leisure saved from shopping time thanks to an additional unit of money. Now holding that additional unit means forgone interet R t and thus some loss of consumption U c (c t )R t.but that lost consumption also saves on shopping time, which adds to utility. The left hand side therefore captures the net loss of utility from holding an additional unit of money. I We can get a money demand function from this condition - but it requires some speci c functional forms: U (c t, l t ) = ln c t + η ln l t, and s t = ψ c t m t

27 The Shopping Time model I Then we have: 1 c t ηψ 1 l t 1 1 c t m t 1 η s t l t R t = ηψ 1 l t c t m 2 t R t = η s t m t l t m t = c t " η s t l t 1 η s t l t # R 1 t Again we have a transaction demand for money, and a negative interest rate e ect. In addition to previous M d we now have a dependence on the shopping to leisure time ratio s t l t. An increase in the shopping time raises the demand for money. Online shopping, for instance, will reduce the demand for money in two ways: rst, use credit cards, and second, a fall in s t l t.

28 Transaction Costs I Last, consider the notion that money economizes on transaction costs incurred when consuming. Modify the budget constraint accordingly: b t+1 (1 + π t+1 ) b t + m t+1 (1 + π t+1 ) m t + c t + T (c t, m t ) = x t where T (0, m) = 0, T c, T cc, T mm > 0 and T m, T mc < 0 I Consider the Lagrangean: L t = s=0 fβ s U(c t+s ) + λ t+s [x t+s + (1 + R t+s ) b t+s +m t+s (1 + π t+s+1 ) (b t+s+1 + m t+s+1 ) c t+s T (c t+s, m t+s )]g

29 Transaction Costs I First order conditions L t c t+s = β s U 0 (c t+s ) λ t+s [1 + T c (c t+s, m t+s )] = 0 L t b t+s = λ t+s (1 + R t+s ) λ t+s 1 (1 + π t+s ) = 0 L t m t+s = λ t+s [1 T m (c t+s, m t+s )] λ t+s 1 (1 + π t+s ) = 0

30 Transaction Costs I The second condition implies λ t+s λ t+s = 1 + π t+s R t+s I Using the rst condition (and setting s = 1), this implies βu 0 (c t+1 ) U 0 (c t ) 1 + T c (c t, m t ) 1 + R t+1 = T c (c t+1, m t+1 ) 1 + π t+1 An Euler equation amended for the ratio of marginal transaction costs. I In addition R t+1 = T m (c t+1, m t+1 ) The forgone interest incurred by holding an additional unit of money is equal to the marginal cost saving this allows.

31 Demand and Hyperin ation - The Cagan model I Consider the nominal money demand function we derived: M t = P t c t Rt α = P t c t (r t + π t+1 ) α I Now we will suppose consumption and the real interest rate change little. Then money demand can be approximated as I Taking logarithms: M t = φp t π α t+1 ln M t p t = α (E t p t+1 p t ) where we ignore lnφ and introduce some uncertainty - hence the expectation operator. p t = ln P t.

32 Demand and Hyperin ation - The Cagan model I Now rearrange: I Substitute recursively: p t = α ln M t + α 1 + α E tp t+1 p t = α where we assumed that lim n! s=0 α 1 + α α 1 + α s E t ln M t+s n E t p t+n = 0 I Notice that the price level today responds to all future expected (discounted) realizations of monetary policy.

33 Hyperin ations I Now put some structure on what monetary policy does. Assume ln M t = µ + ε t grows at a constant rate, and there are some unexpected deviations from that rule, E t ε t+1 = 0. Then we have I That is I Which nally means ln M t+s = ln M t + µs + s i=1 E t ln M t+s = ln M t + µs p t = ln M t + µ α = ln M t + µα s=0 ε t+i α 1 + α s s

34 Hyperin ations I Now E t π t+1 = E t p t+1 p t = E t ln M t+1 ln M t = µ In a period of hyperin ation, consumption c t and the real rate r t change much less than the price level. Under these assumptions, this shows the astronomical rate of in ation requires a correspondingly high rate of monetary creation. This is what we observed in Germany 1923, in some South American countries in the 90s, and in Zimbabwe today. It happens when governments print money to pay for their expenditures.

35 The Friedman Rule I In principle, what was just discussed holds in the long term, since c t and r t are stabilised around their steady state values. Thus, in the long term, in ation is "always and everywhere a monetary phenomenon" I So choosing µ will pin down long run in ation. Which µ should governments choose? Friedman proposed an answer. I He argued the marginal private return to holding money is π, while the marginal social cost of producing money is virtually zero. Friedman proposed to eliminate the private cost of holding money by setting π = θ, the rate of private time preferences, which in the long run will equal the real interest rate r. I So Friedman basically suggested setting the nominal interest rate R = r + π to zero. Then the opportunity cost of holding money balances is e ectively zero.

36 Superneutrality I The classical dichotomy in macroeconomics is that nominal shocks have no long-run e ects on real variables. When these nominal shocks are money shocks, this is known as the superneutrality of money. Proportional changes in nominal money balances, and hence in ation, have no e ect on the real variables of the economy, such as consumption, investment, production and capital. I From the models we saw, it seems the presence of money in the household budget constraint imposes a real cost on the economy, and thus money is not super-neutral. But this was based only on the decisions of households, and thus on a partial equilibrium view of the economy. Now turn to a general equilibrium version. I In particular, we now need to account for the fact that in ation tax (which imposes a real cost on households) actually generates seignorage revenues for the government. And the government must also satisfy a budget constraint.

37 Superneutrality I So we will consider an economy where all the money is provided by the government, and seignorage revenues are returned to households in the form of transfers, which households are aware of. The household budget constraint writes (1 + π t+1 ) (b t+1 + m t+1 ) + c t + T t = x t + (1 + R t ) b t + m t where T t < 0 would imply a transfer to households. The government also has a budget constraint, which writes T t = m t (1 + π t+1 ) m t+1 I So we have the consolidated constraint (1 + π t+1 ) b t+1 + c t = x t + (1 + R t ) b t in which real money balances no longer appear!

38 Superneutrality I The Lagrangean will write: L t = s=0 fβ s U(c t+s ) + λ t+s [x t+s + (1 + R t+s ) b t+s (1 + π t+s+1 ) b t+s+1 c t+s ]g I with rst order conditions L t c t+s = β s U c (c t+s ) λ t+s = 0 L t b t+s = λ t+s (1 + R t+s ) λ t+s 1 (1 + π t+s ) = 0 I and the classic Euler condition: βu c (c t+1 ) U c (c t ) 1 + R t π t+1 = 1

39 Superneutrality I does not matter for consumption any more. Does it matter in the long run? Consolidated constraint implies c = x + (R π) b I What is the long run real rate of interest? The Euler condition requires that i.e: R π = 1 β β β 1 + R 1 + π = 1 I So long run consumption is given by (1 + π) ' 1 β β c = x + 1 β β b and money matters not at all. The economy is super-neutral.