Monetary Policy Shocks: What Have We Learned and to What End? Christiano, Eichenbaum and Evans Handbook of Macroeconomics

Transcription

1 Moneary Policy Shocks: Wha Have We Learned and o Wha End? Chrisiano, Eichenbaum and Evans Handbook of Macroeconomics

2 Moneary Policy Shocks Idea: Assessing moneary policy canno be done hrough saisical experimenaion due o he Lucas criique. Wha does he Lucas criique say? Quesion: Hence, we have o use a srucural heoreical model o evaluae moneary policy. How do we choose which model? 2

3 Moneary Policy Shocks R.E. Lucas (1980) need o es hem [models] as useful imiaion of realiy by subjecing hem o shocks for which we are fairly cerain how acual economies or par of economies would reac. The more dimensions on which he model mimics he answers acual economies give o simple quesions, he mode we rus is answers o harder quesions. How o carry his program in pracice? 3

4 Moneary Policy Shocks 1. one isolaes moneary policy shocks in acual economies and characerizes he naure of corresponding moneary experimens (idenified VAR) 2. characerize he response of he acual economy o hese moneary experimens (impulse responses). 3. Perform he same experimens in he model economies o be evaluaed and compare wih he oucomes of acual economies. 4

5 Moneary Policy Shocks General idenificaion sraegies: 1. Shor-run idenificaion: Cholesky, recursiveness assumpion. 2. Even sudies: a la Romer and Romer (1989) however, he Fed s behavior is always endogenous (Hoover and Perez, 1994a, b). 3. Long-run idenificaion: Blanchard and Quah (1989), King, Plosser, Sock and Wason (1991). 4. Demiralp and Hoover (2003): Graph analysis 5

6 Moneary Policy Shocks Overview of he resuls: Alhough here is no agreemen on he correc way of idenifying exogenous shocks, alernaive idenificaion assumpions seem o deliver very similar conclusions: EM P PCOM FF NBRX DM2 EM P PCOM FF NBRX DM

7 Moneary Policy Shocks In response o a conracionary moneary policy shock: shor-erm ineres raes rise.8 Response of Fed Funds Rae o Cholesky One S.D. Fed Funds Rae Innovaion

8 Moneary Policy Shocks oupu, employmen, profis, and money aggregaes decline Response o Cholesky One S.D. Innovaions ± 2 S.E..1 Response of Employmen o Fed Funds Response of Nonborrowed Reserves o Fed Funds Response of M2 growh o Fed Funds

9 Moneary Policy Shocks prices respond very slowly.10 Response of Prices o Cholesky One S.D. Fed Funds Rae Innovaion wages fall bu lile 9

10 Moneary Policy Shocks Moneary policy shocks accoun for a modes porion of oupu and price volailiy. % FEV of EM % FEV of P

11 Moneary Policy Shocks Economic inerpreaions of exogenous moneary policy shocks 1. As a shock o he preferences of he moneary auhoriy 2. As a way for he Fed o accommodae privae secor expecaions 3. Measuremen error in he preliminary daa available o he FOMC 11

12 Moneary Policy Shocks Remarks Impulse responses are elasiciies or slopes esimaed from he daa residuals should no be inerpreed. YY YY vs. X Variable Coefficien Sd. Error - Saisic Prob. C X X Variable Coefficien Sd. Error - Saisic Prob. C X

13 Moneary Policy Shocks Puzzles: 1. Price Puzzle: prices rise in response o a posiive shock o FF 2. Liquidiy Puzzle: FF increases in response o an increase in narrow money. 3. UIP Puzzle: posiive shocks o FF resul in persisen deviaions in UIP (in favor of he US). 13

14 Moneary Policy Shocks We conclude his subsecion by noing ha even if moneary policy shocks have played only a very small role in business cycle flucuaions, i does no follow ha he sysemaic componen of moneary policy has played a small role 14

15 Moneary Policy Shocks Imposing non-recursive assumpions on a VAR Consider a sysem wih y, p, i } where we achieve idenificaion by assuming: C 0 { = * 0 * Noice ha he sysem is over-idenified: here is an exra 0 resricion. Hence, his is esable. 0 * * 0 0 * 15

16 Moneary Policy Shocks Esimaing he model ha imposes he exra resricion: y i p = β = β = α iy yy py y y y α + β + β ip p yp pp p p β iy + β + β y 1 yi pi i 1 i 1 + β + ε ip + ε p y p 1 + β i ii 1 + ε i The likelihood from his model and he likelihood from he reduced form VAR can hen be compared using a LR es. However, rejecion of he es does no inform us abou which zero coefficien resricion is incorrec. 16

17 Moneary Policy Shocks Narraive Approach Treamen Effecs Lieraure If he was less annoying in regard o his classmaes, he was more so in his classrooms. He had learned from Golieb he rick of using he word "conrol" in reference o he person or animal or chemical lef unreaed during an experimen, as a sandard for comparison: and here is no rick more infuriaing. When a physician boased of his success wih his drug or ha elecric cabine, Golieb always snored, "Where was your conrol? How many cases did you have under idenical condiions, and how many of hem did no ge he reamen?" Now Marin began o mouh i -- conrol, conrol, conrol, where's your conrol? ill mos of his fellows and a few of his insrucors desired o lynch him. Arrowsmih (1925) by Sinclair Lewis, firs American o win he Nobel Prize for lieraure. 17

18 Moneary Policy Shocks Definiions: y oucome variable X = Z, W ) observable variables; W ( U unobservable variables D y 0 y 1 p insrumens binary indicaor of reamen (policy inervenion) oucome variable under non-reamen oucome variable under reamen = 1 X ] p( X ) [ D reamen given he observables. propensiy score, he probabiliy of 18

19 Moneary Policy Shocks Remark: we observe reamen on he reaed, non-reamen on he unreaed bu we do no observe non-reamen on he reaed and reamen on he unreaed. In oher words, we do no direcly observe y 1 y0 Common measures of reamen effecs: Average Treamen Effec (ATE): y y ) E ( 1 0 Average Treamen Effec on he Treaed (ATET): E ( 0 y1 y D = 1) Remark: mos of he work usually consiss in esimaing he counerfacual y ) E or E ( y 0 = 1) ( 0 D 19

20 Moneary Policy Shocks An imporan assumpion: selecion on observables, ignorabiliy or unconfoundedness : condiional on he observables, he oucomes are independen of he reamen (condiion 1 in he paper). Remark: if here is selecion on unobservables hen reamen paricipaion is endogenous and we usually require some ype of maching esimaor o re-randomize reamen paricipaion. I does no seem ha he paper permis his scenario. 20

21 Moneary Policy Shocks A causal map: Z observables insrumens W y D reamen U unobservables 21

22 Moneary Policy Shocks Condiional Independence Given Predicive Variables for Unobservable Causes,Whie (2005): The insrumens W are predicors (bu no causes) of he unobserved variables U so ha D X ( Z, W ), i.e., D does no Granger-cause U given X. U = Remark: his is no your usual insrumenal variable! In linear regression, we insrumen he regressors o avoid correlaion wih he residuals. Here we insrumen he residuals. 22

23 Moneary Policy Shocks An Example: Dummy Variable Regression Noice: = + (. Inuiion would sugges ha o y D y1 1 D ) y0 measure he effec of he policy inervenion we could esimae: y = X ' β + D α + ε Why is αˆ LS usually a biased esimae of ATET? he densiy of he unobservables given he observables is no he same for he reaed and he unreaed, i.e., here is a selecion effec. 23

24 Moneary Policy Shocks 24 Essence of he Angris-Kuerseiner es: If he reamen has no effec, hen oucomes and reamens should be independen. This resuls in equaion (5): ( )( ) [ ] 0 ) ( ) ( ) ( ) 1, ( = = = z z p D y y E z p z y y P z D y y P 1 so he es is a sor of es of uncorrelaedness beween policy innovaions and oucomes.

25 Moneary Policy Shocks Suggesions: Selecion on unobservables seems a real problem in general: Is condiion 1 a reasonable assumpion in pracice? The elephan in he room: researchers acually care abou he reamen effec David Dickey: he power of a es is he power of he es imes he probabiliy ha i will be used. Applicaion: why no use fuures daa on he fed funds rae arge o obain direc measures on policy innovaions? (see Kuner, 2001; Demiralp and Jordà, 2004; Cochrane and Piazzessi, 2005) 25

26 Moneary Policy Shocks Conclusions Wha happens afer a moneary policy shock? he answer o his quesion is used o evaluae compeing macro models, NOT o measure he effecs of moneary policy (since i is difficul o grapple wih he effecs of sysemaic moneary policy). We will see saisical mehods o measure when he responses from a macro model and from he daa are 26

27 Moneary Policy Shocks close in he saisical sense Minimum Disance Mehods. Once a macro model is validaed by he daa (by comparing impulse responses) hen we can do conrolled experimenaion in he model (no in he daa). 27