The Announcement Effect: Evidence from Open Market Desk Data Selva Demiralp Board of Governors of the Federal Reserve System Oscar Jordá University of California, Davis April 5, 2001 Prepared for the Federal Reserve Bank of New York Conference “Financial Innovation and Monetary Transmission,” April 5-6, 2001. Announcement Effect The market’s reaction to the announcement of a new target for the overnight rate that causes the overnight rate to trade around the new target with few to no open market operations. Previous Literature Guthrie and Wright (2000), Woodford (2000), Taylor (2001), and Thornton (2000) Demiralp and Jordá (2001) “The Pavlovian Response to Fed Announcements” 1 Questions Has the post-1994 period of policy announcements changed the manner in which the Fed uses open market operations to control the federal funds market? What effect, if any, did this change in policy have on the behavior of the term structure? 2 What do we expect to find? Open market operations are no longer needed as a “signaling device” to modify the federal funds rate in the direction of the new target level. The Announcement effect regiments the formation of expectations in longermaturity term-rates. 3 Taylor (2001) A textbook model of the daily reserves market (i) Demand for Reserves Rtd ( f t Et f t 1 ) Speculative Demand where ft t , Precautionary Demand , , , 0 Changes in the effective funds rate lead to movements along the demand curve. Changes in the expected rate lead to shifts in the demand schedule, which eliminates any arbitrage opportunities. 4 (ii) Supply of Reserves Rts Rt 1 ( f t 1 f t *1 ) , 0 For a given level of the target, the Federal Reserve adjusts reserves to offset daily deviations of the effective rate from the target. For a given level of the funds rate, the Federal Reserve changes the reserve supply that is consistent with a new target level (i.e. whenever the target rate is changed). 5 Figure 1: A General Model of the Reserves Market with Anticipated Policy Actions S1old E1old S1new S0 E1new f1* E0A f0 E0 f 0* E0B D1 D0 O NBR1old NBR1new NBR0 NBR* 0 Reserves Rts Rt 1 E t 1 ( f t f t * ) 6 Open Market Operations and Target Announcements Two questions: Is there any difference in the size of operations needed to signal a new level for the target? Is there any difference in the portfolio of operations used in this signaling? Samples: April 25, 1984 to February 4, 1994 February 4, 1994 to August 18, 1998 August 18, 1998 to August 14, 2000 7 Types of Operation Adds Liquidity (Purchases) Detracts Liquidity (Sales) T-Bill domestic purchases Permanent (Outright) T-Bill Sales T- Bill foreign purchases Coupon domestic purchases PB Coupon Sales PS Coupon foreign purchases Temp. Term RP purchases Term Matched Sale TB Overnight RP Overnight Purchases Purchases TS Overnight Matched OB Sale Purchases OS Percent of the Time each Type of Operation is Executed Operation Pre-1994 Post-1994 PB 0.32 0.13 TB 0.13 0.27 OB 0.10 0.36 PS 0.03 0.02 TS 0.03 0.01 OS 0.05 0.04 Percent of the Time any Operation is Executed 0.55 0.67 8 What Day of the Maintenance Period Are Target Changes Most Often Executed? Pre-1994 Post-1994 Day 1 – Thurs. 0.37 0.05 Day 2 0.06 0.05 Day 3 0.02 0.05 Day 4 0.04 0.20 Day 5 0.07 0.05 Day 6 – Thurs. 0.22 0.05 Day 7 0.10 0.00 Day 8 0.03 0.00 Day 9 0.03 0.40 Day 10 0.05 0.15 Adjustments x jt ~ x jt TRs (t )1 9 Decomposing the gap: ft – ft* NEEDt f t f m*(t )1 wt* Em(t )1 f m*(t ) EXPECTt Em(t )1 f m*(t ) SURPRISE t f t* Em(t )1 f m*(t ) f * m( t )1 value of the target at the start of the maintenance period. wt* probability distribution of target changes within the maintenance period. * E f m ( t ) beginning of the maintenance m(t ) 1 period expectations of a target change. SURPRISEt = 0, except when ft* is non-zero. ACH Model Beginning of the maintenance period expectations of a target change. 10 Open Market Operations as a Function of the Effective-Target Federal Funds Rate Gap OMO OS xOB OB cOB cOS ft – ft* Tobit Specification: 0 x jt x * jt if x *jt c j if x *jt c j x *jt j ( f t f t * ) t 11 Specification 10 3 x k d i X t'i * jt 10 k 1 k 1 10 k N k t i 1 3 d NEEDt 1 d EXPECTt 1 Si SURPRISE t i jt k t k 1 k E k t i 0 k : modifies the threshold. Allows operations to be more likely certain days of the week. i : conditions on past and unobserved information. Complement vs. substitute operations. N,Ek : the responsiveness of the Desk varies according to the day of the m-period. Si : contemporaneous and up to three lag dynamic pattern of operations in the days surrounding the announcement. 12 Table 2 – Signs of the Coefficients in the TOBIT Regressions: Purchase Operations 1984-1994 Need Day 1994-1998 Expected Need 1998-2000 Expected Need Expected OB TB PB OB TB PB OB TB PB OB TB PB OB TB PB OB TB PB + . . . . . . ++ . -- ++ . . . . Friday ++ . . . ++ . ++ . . . . . -- . . Monday + . . . . . . . . . . . . + . . . . ++ . . . . . + _ . . . + . . . . . . . ++ . . . . ++ . . -- . . . . . . _ ++ . . . . . . . . - . + . . . . . . Friday ++ . . . . . . . . . . . ++ . . . . . Monday + . . . . . . . . . . . + . . - . . ++ . . . . . . . . . . . + . . . . . . . . . . . . . . . . . . . . . Thursday . ++ . ++ ++ . 1 2 3 Tuesday 4 Wednes. 5 Thursday 6 7 8 Tuesday 9 Wednes. + 10 Surprise – Lag 0 -- - . + . . . . . 1 . + . - . + . . . 2 -- . . . . . . . . 3 . . . . . . . . . 13 Results Frequency each operation is used Operation 84-94 94-98 PB OB 31.91% 10.04% 11.18% 26.50% 98-00 16.51% 58.16% 84-94: OB operations most responsive to variations in NEED but also period in which ft – ft* deviations are most persistent. 94-98: OB operations concentrate on first Friday and midway through the period – coincides with smallest ft – ft* deviations. 98-00: OB less responsive overall coincides with observed persistence in ft – ft* TB: Mostly associated with accommodation of expectations early in the m-period. PB: Operations of a technical nature. SURPRISE: 84-94 consistent with liquidity effect. 94-98 significant decrease in the response. 98-00 virtually zero response – consistent with announcement effect. Smoothing: Significant decline in the persistence of operations. 14 Response of Term Rates to the Announcement Kuttner (2000) Fed Funds Futures Rate Surprise ~ ms f s0, f s0, 1 ms E 1 (f* ) f* ~ ms : number of days in month s f s0, spot-month futures rate at date Agenda Has the predictive power of the futures market improved since 1994? What is the response of term rates to changes in the target? Tests of the timing hypothesis. 15 Predicting Changes in the Target Notice f t 1 {0.50,0.25,0,0.25,0.50} . Assume that the discrete changes in the target are related to a latent index, such that * f t*1 0.50 if 0.25 if if 0.5 st 1 ( , c1 ] st 1 (c1 , c2 ] st 1 ( c4 , ) (8) with * st 1 E t ( f t*1 ) m d tmonth E ( f 1 t t 1 ) * fomc * mp d tmp 1 E t ( f t 1 ) f d t 1 E t ( f t 1 ) u t 1 (9) wt 1 u t 1 So that, for example P( f t*1 0.25 | wt 1 ) P( c 3 wt 1 u t 1 c 4 ) ( c 4 wt 1 ) ( c 3 wt 1 ) (10) 16 Ordered Probit Estimates Coefficient Et (f t *1 ) At FOMC End of Month End of M. Period Avg. Log-Lik Pseudo R2 Observations 1989 – 1994 4.77** (0.76) 6.12 (4.30) -4.45** (0.94) 1.74 (1.24) -0.085 0.18 1229 1994 – 2002 7.19** (1.20) 9.35** (1.22) -7.75** (1.30) 0.26 (1.27) -0.029 0.70 2089 Note: Estimates of limit points omitted. **(*) indicates significant at the 99%(95%) confidence level. Predicted Frequencies Value -0.50 -0.25 0 0.25 0.50 1989 – 1994 1994 – 2002 Actual Pred. Error Value Actual Pred. Error 3 1 2 -0.50 8 10 -2 21 0 21 -0.25 9 0 9 1207 1230 -23 0 2058 2067 -9 0 0 0 0.25 9 7 2 0 0 0 0.50 5 5 0 17 Term Rate Responses to Target Changes Cook and Hahn (1989), Kuttner (2001) Rt 1 Et 1 (f t* ) 2~t u t Rt Maturity 3-month 6-month 2-year 5-year 10-year 1 Response to Anticipated 1989 – 1994 – 1994 2002 0.33** 0.07 (0.15) (0.04) 0.29** 0.04 (0.13) (0.04) 0.20 0.04 (0.17) (0.05) 0.16 0.01 (0.16) (0.06) 0.15 -0.03 (0.16) (0.05) (9) 2 Response to Unanticipated 1989 – 1994 – 1994 2002 0.90*** 0.82*** (0.13) (0.12) 0.88*** 0.63*** (0.11) (0.11) 0.74*** 0.51*** (0.14) (0.15) 0.60*** 0.39** (0.13) (0.17) 0.43*** 0.25 (0.13) (0.15) Notes: expression (12) is estimated only on days in which the target was changed. There are 24 observations for the pre-1994 (5/18/1989 – 2/3/1994) sample, and 31 observations for the post1994 (2/4/1994 – 2/1/2002) sample. Standard errors in parenthesis. ***/**/* indicates significant at the 99% /95%/90% confidence level. 18 The Timing Hypothesis R 0 f dFOMC s dSW ITCH 10 E 1 (f* ) 1f dFOMC E 1 (f* ) 1s dSW ITCH E 1 (f* )dSW ITCH 20~ 2f dFOMC~ 2s dSW ITCH~ u Timing At FOMC Outside FOMC At FOMC+ SWITCH Outside FOMC + SWITCH R2 3month 0.63** (0.18) 0.78** (0.07) 1.42** (0.50) 1.58** (0.53) 0.75 Maturity 62-year 5-year month 0.18 0.30 0.13 (0.15) (0.20) (0.21) 0.60** 0.64** 0.49** (0.07) (0.08) (0.09) 1.68** 1.73** 1.76** (0.44) (0.56) (0.60) 2.09** 2.06** 2.12** (0.47) (0.60) (0.64) 0.69 0.61 0.52 10year 0.02 (0.19) 0.34** (0.08) 1.42* (0.54) 1.74** (0.58) 0.48 Notes: Standard Errors in parenthesis. **(*) indicates significant at the 99%(95%) confidence level. Calculation of the Coefficients: 0 f At FOMC: 2 2 0 Outside FOMC: 2 At FOMC + SWITCH: 20 2f 2s Outside FOMC + SWITCH: 20 2s 19 Conclusions The Announcement effect has reduced the volume of open market operations necessary to signal changes in the target level of the federal funds rate. Public communication of policy has regimented the formation of expectations on future policy moves, improving the market’s ability to forecast future changes. The announcement of policy moves synchronizes policy-induced shifts in the term structure in a manner consistent with the expectations hypothesis. 20