CENTRAL BANK TRANSPARENCY AND SHORT TERM INTEREST RATE FORECASTING UNCERTAINTY

Marc Hayford and A.G. Malliaris

Loyola University Chicago

Prepared for the IBEFA/ASSA MEETINGSJanuary 3 – 5, 2009San Francisco, CA

Introduction

• Past 10 – 15 years central banks have moved to a more transparent monetary policy

• (Carpenter 2004) Types of monetary policy transparency:

goal transparencyexplicit inflation targets (20 countries)Fed’s “dual mandate”

instrument transparencyexplicit targets for short term interest rates more information on bias or tilt of future monetary policy

implementation transparency

Empirical literature on Monetary Policy transparency (MRT):

Development of indexes of MP transparency

Regress various measures of economic and financial activity on transparency index

Results:

Demertzis and Hallet (2002): ↑MPT → ↓ variance of inflation but not mean inflation

Cortareas et al (2003): ↑MPT → ↓ mean inflation but not variance

Ehrmann and Fratzcher (2007) ↑MPT of Fed → more predictable MP

Crowe and Meade (2007) ↑MPT → increase use of public information by private forecasters

Swanson (2004) ↑MPT of Fed → U.S. forecasters and financial markets have become better at forecasting short-term interest rates.

Our research is similar to Swanson (2004) however use a different data set and regression model.

Evolution of U.S. monetary policy towards more transparency:

In a series of steps since February 1994 the FED has revealed more information and revealed the information sooner about monetary policy.

• Prior to February 1994 market participants had to guess the Fed funds target: Minutes released with a lag of at least 30 days after the meeting.

From FOMC minutes December 1993, released 30 days after the meeting:

“In the implementation of policy for the immediate future, the Committee seeks to maintain the existing degree of pressure on reserve positions. In the context of the Committee's long-run objectives for price stability and sustainable economic growth, and giving careful consideration to economic, financial, and monetary developments, slightly greater reserve restraint or slightly lesser reserve restraint might be acceptable in the inter-meeting period. The contemplated reserve conditions are expected to be consistent with moderate growth in M2 and M3 over coming months.”

• Beginning February 1994: post FOMC meeting press releases signaled changes in Fed funds but not explicit target

FOMC Press Date: February 4, 1994  For immediate release

Chairman Alan Greenspan announced today that the Federal Open Market Committee decided to increase slightly the degree of pressure on reserve positions. The action is expected to be associated with a small increase in short-term money market interest rates.

The decision was taken to move toward a less accommodative stance in monetary policy in order to sustain and enhance the economic expansion.

Chairman Greenspan decided to announce this action immediately so as to avoid any misunderstanding of the Committee's purposes, given the fact that this is the first firming of reserve market conditions by the Committee since early 1989.

• Beginning July 1995: post FOMC meeting press releases explicitly state the Fed funds target

Release Date: July 6, 1995  For immediate release

Chairman Alan Greenspan announced today that the Federal Open Market Committee decided to decrease slightly the degree of pressure on bank reserve positions.

As a result of the monetary tightening initiated in early 1994, inflationary pressures have receded enough to accommodate a modest adjustment in monetary conditions.

Today's action will be reflected in a 25 basis point decline in the federal funds rate from about 6 percent to about 5-3/4 percent.

• May 1999: post FOMC meeting press releases reasoning for target level as well as expected future path of Fed funds target

Release Date: May 18, 1999

For immediate release

The Federal Reserve released the following statement after today's Federal Open Market Committee meeting:

While the FOMC did not take action today to alter the stance of monetary policy [FFR = 4.75%], the Committee was concerned about the potential for a buildup of inflationary imbalances that could undermine the favorable performance of the economy and therefore adopted a directive that is tilted toward the possibility of a firming in the stance of monetary policy. Trend increases in costs and core prices have generally remained quite subdued. But domestic financial markets have recovered and foreign economic prospects have improved since the easing of monetary policy last fall. Against the background of already-tight domestic labor markets and ongoing strength in demand in excess of productivity gains, the Committee recognizes the need to be alert to developments over coming months that might indicate that financial conditions may no longer be consistent with containing inflation.

Fed’s intention: Bernanke (2007):

Claim: increased instrument transparency , i.e. explicit Fed funds target + information about the bias or tilt of future monetary policy → ↑ effectiveness of monetary policy

Argument:

Fed only controls a short-term interest rate – Fed funds rate

Spending depending on long term interest rates

Expectations approach to the term structure of interest rates → long term interest rates = average of expected current and future short term interest rates

Increased MP transparency about current and future Fed funds rate

→ alignment of market expectations with future monetary policy

→ monetary policy more effective in moving long term interest rates and thus more effective in influencing economic activity and maintaining stable inflation.

Empirical question:

Has increased transparency about US monetary policy reduced the uncertainty market participants have about future monetary policy?

Attempt at an answer:

Use forecasts of future 3-month T-bill rates from the Survey of Professional Forecasters (SPF) as proxy for future expected Fed funds rate

- 3-mth T-bill rates closely tracks Fed funds rate

- Dispersion of future 3 mth T-bill rate forecasts can be used as a measure of uncertainty about the future monetary policy

- Ex post forecast errors can also be used as a measure of uncertainty about the future monetary policy

0

4

8

12

16

20

1980 1985 1990 1995 2000 2005

3 month T-bill Rate(secondary market)

Federal funds rate

Figure 1: Federal funds rate and the 3 month T-bill rate

Correlation = 0.99

Theoretical motivation

group of forecasters forecasting future monetary policy, F

individual forecaster i =1,2,…n has access to:

Public information on future monetary policy: (1) FF T

Private information on future monetary policy: (2) iP

i FF

where and i are white noise, independently distributed with zero means and variances 2

and 2 .

Actual realization of future monetary policy: (3) FF

is white noise with a zero mean and variance 2 and F is deterministic.

Forecaster i forecasts future monetary policy as a signal extraction problem (Morris and Shin (2002)):

(4) P

iT

i bFaFFE )(

where

22

2

a

22

2

b

a+b=1.

Note:

1) ↑ monetary policy transparency → TF less noisy signal of future MP → ↓2

2) ↓

2 → ↑ a relative to b → private forecasters place more weight on the public

information

3) As 2 → 0 , a → 1 and b → 0 → T

i FFE )( .

Note:

1) ↑MP transparency will ↓ dispersion of forecast errors

Dispersion of the forecast errors across forecasters:

n

i

n

iii

n

FEn

FE

STF1

2

1

)(1

)(

Or Using equation (4):

n

i

n

i

Pi

Pi

n

Fn

F

bSTF1

2

1

1

Which decreases as transparency increases that is, as 2 and

hence b decreases.

2) ↑MP transparency will ↓ex post forecast errors (holding constant future shocks to MP)

Mean squared ex post forecast error:

n

i

n

ii

n

FEn

F

EMSFE1

2

1

)(1

Or using equations (1), (2) and (3) and assuming that

0jiE for ji results in:

22222

1

2

1

ba

n

n

ba

EMSFEn

i

n

ii

Survey of Professional Forecasters:

Beginning of Feb, May, Aug and Nov, 30 to 50 professional forecasters are asked for their forecast of the 3 mth T-bill rate 0, 1, 2, 3 and 4 quarters ahead.

Forecast dispersions:

(5)

tn

i t

jtijt

jt n

TFETFSTF

1

2)(

ijtTF = j quarter ahead 3-mth T-bill forecast of forecaster i,

quarter t

)( jtTFE = mean forecast for quarter t, j = 0,1, 2, 3, 4 quarters ahead tn = number of forecasters each quarter is (30 to 50 people)

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

1985 1990 1995 2000 20050.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

1985 1990 1995 2000 2005

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

1985 1990 1995 2000 20050.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

1985 1990 1995 2000 2005

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

1985 1990 1995 2000 2005

Figure 2: Standard Dev iation of the three month T-bill rate

(Surv ey of Prof essional Forecasters)

Current quarterOne quarter ahead

Two quarters ahead Three quarters ahead

Four quarters ahead

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

1985 1990 1995 2000 20050.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

1985 1990 1995 2000 2005

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

1985 1990 1995 2000 20050.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

1985 1990 1995 2000 2005

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

1985 1990 1995 2000 2005

Figure 3: Three Month T-bi l l Rate Ex Post Forecast errors

(Survey of Professional Forecasters)

Current quarterOne quarter ahead

Two quarters ahead

Three quarters ahead

Four quarters ahead

Table 1: Mean of Standard deviation of 3 month T-bill survey forecast dispersions

0 to 4 quarters ahead Sample STF0 STF1 STF2 STF3 STF4

Pre-transparency 81:3 to 93:4 0.328 0.497 0.640 0.868 1.418 84:1 to 93:4 0.223 0.357 0.502 0.732 1.269

Post-transparency 94:1 to 08:1 0.162 0.248 0.375 0.493 0.841 95:4 to 08:1 0.161 0.247 0.364 0.478 0.817 99:3 to 08:1 0.160 0.267 0.389 0.497 0.800

Percent change in mean dispersion from pre (1981:3 to 1993:4) to post transparency sample STF0 STF1 STF2 STF3 STF4

94:1 to 08:1 -50.6% -50.1% -41.4% -43.2% -40.7% 95:4 to 08:1 -50.9% -50.3% -43.1% -44.9% -42.4% 99:3 to 08:1 -51.2% -46.3% -39.2% -42.7% -43.6%

Percent change in mean dispersion from pre (1984:1 to 1993:4) to post transparency sample STF0 STF1 STF2 STF3 STF4

94:1 to 08:1 -27.4% -30.5% -25.3% -32.7% -33.7% 95:4 to 08:1 -27.8% -30.8% -27.5% -34.7% -35.6% 99:3 to 08:1 -28.3% -25.2% -22.5% -32.1% -37.0%

• Analysis of spikes in forecast dispersions in excess of one standard deviation reveal:

In 91 quarters from 1984:1 to 2008:3 there are 35 “spikes” i.e. quarters of forecast dispersion in excess of at least one standard deviation.

Spikes correspond to:

financial crises

absolute size of the change in Fed funds rate

ttttt

tkktjt

NINERUSSIANASIANCRASH

FFRDDFFRSTF

1187 7654

3210

Basic regression model

Transparency dummies: Dk = 0 and then beginning in k = 94:1 95:4 99:3 Dk = 1

Financial crises dummies:87CRASH = 1 for 1987:4 and 0 otherwiseAISAN = 1 for 1997:3 and 0 otherwiseRUSSIAN = 1 for 1998:4 and 0 otherwiseNINE11 = 1 for 2001:4 and 0 otherwise

Hypothesis: Β2 and β3 < 0 if transparency has reduced uncertainty

Table 4a Sample period 1984:1 to 2008:1 (t-statistics in parentheses)* Dependent variable: tSTF0 (mean = 0.19)

Constant 0.19 (7.65)

0.19 (7.15)

0.18 (7.73)

0.17 (7.28)

0.17 (6.02)

0.14 (8.01)

tFFR 0.09 (2.075)

0.10 (2.89)

0.09 (3.00)

0.10 (3.17)

0.11 (2.71)

0.14 (4.46)

94D -0.08 (-2.09)

-0.10 (-3.56)

tFFRD *94 0.05 (0.96)

0.08 (1.96)

95D -0.06 (-1.69)

-0.08 (-3.30)

tFFRD *95 0.04 (0.86)

0.07 (1.84)

99D -0.07 (-2.41)

-0.04 (-2.15)

tFFRD *99 0.06 (1.27)

0.03 (0.67)

87CRASH 0.14 (5.95)

0.16 (7.08)

0.19 (11.60)

ASIAN 0.76 (120.1)

0.76 (115.9)

0.71 (41.27)

RUSSIAN 0.12 (11.05)

0.12 (9.78)

0.09 (5.85)

NINE11 -0.02 (-0.66)

-0.02 (-0.65)

-0.01 (-0.47)

2R 0.19 0.62 0.17 0.60 0.18 0.55 S.E. of reg 0.11 0.07 0.11 0.08 0.11 0.08 * t-statistics use White heteroskedasticity consistent standard errors

Table 4b Sample period 1984:1 to 2008:1 (t-statistics in parentheses)* Dependent variable: tSTF1 (mean =0.29)

Constant 0.28 (12.31)

0.27 (12.90)

0.27 (12.31)

0.25 (12.97)

0.23 (13.50)

0.22 (13.73)

tFFR 0.16 (4.89)

0.17 (5.30)

0.16 (5.31)

0.18 (5.83)

0.21 (6.46)

0.22 (6.41)

94D -0.09 (-3.16)

-0.08 (-2.91)

tFFRD *94 -0.01 (-0.24)

-0.02 (-0.35)

95D -0.06 (-2.50)

-0.06 (-2.21)

tFFRD *95 -0.02 (-0.47)

-0.03 (-0.58)

99D 0.00 (0.03)

0.01 (0.27)

tFFRD *99 -0.10 (-2.12)

-.10 (-1.91)

87CRASH 0.24 (11.98)

0.25 (13.77)

0.29 (19.70)

ASIAN 0.02 (1.32)

0.02 (1.07)

-.0.00 (-0.21)

RUSSIAN 0.05 (4.32)

0.06 (3.78)

-0.01 (-0.46)

NINE11 -0.06 (-1.71)

-0.05 (-1.44)

-0.03 (-0.69)

2R 0.46 0.48 0.41 0.44 0.36 0.41 S.E. of reg 0.09 0.08 0.09 0.09 0.09 0.09 * t-statistics use White heteroskedasticity consistent standard errors

Table 4c Sample period 1984:1 to 2008:1 (t-statistics in parentheses)* Dependent variable: tSTF2 (mean = 0.43)

Constant 0.43 (13.55)

0.42 (12.95)

0.43 (13.97)

0.42 (12.13)

0.36 (13.38)

0.36 (12.61)

tFFR 0.16 (3.53)

0.17 (3.62)

0.15 (2.99)

0.16 (3.43)

0.23 (4.82)

0.23 (4.66)

94D -0.08 (-1.89)

-0.07 (-1.61)

tFFRD *94 -0.08 (-1.34)

-0.08 (-1.26)

95D -0.10 (-2.47)

-0.08 (-1.74)

tFFRD *95 -0.04 (-0.46)

-0.07 (-1.10)

99D 0.02 (0.53)

0.03 (0.59)

tFFRD *99 -0.22 (-3.20)

-0.23 (-2.89)

87CRASH 0.18 (6.10)

0.19 (5.80)

0.24 (9.04)

ASIAN -0.06 (-2.07)

-0.05 (-1.88)

-0.08 (-2.66)

RUSSIAN 0.01 (0.40)

0.02 (0.88)

-0.10 (-3.96)

NINE11 -0.07 (-1.74)

-0.06 (-1.35)

0.01 (0.13)

2R 0.22 0.21 0.26 0.21 0.19 0.19 S.E. of reg 0.14 0.14 0.14 0.14 0.14 0.14 * t-statistics use White heteroskedasticity consistent standard errors

Table 4d Sample period 1984:1 to 2008:1 (t-statistics in parentheses)* Dependent variable: tSTF3 (mean = 0.59)

Constant 0.61 (10.05)

0.60 (9.34)

0.59 (10.30)

0.58 (9.61)

0.52 (10.87)

0.52 (10.11)

tFFR 0.26 (1.77)

0.27 (1.79)

0.26 (1.86)

0.27 (1.88)

0.32 (2.35)

0.34 (2.30)

94D -0.14 (-1.97)

-0.12 (-1.59)

tFFRD *94 -0.19 (-1.22)

-0.23 (-1.39)

95D -0.13 (-1.86)

-0.11 (-1.45)

tFFRD *95 -0.22 (-1.46)

-0.27 (-1.74)

99D -0.03 (-0.55)

-0.02 (-0.13)

tFFRD *99 -0.30 (-2.05)

-0.36 (-2.34)

87CRASH 0.21 (3.47)

0.23 (4.08)

0.28 (6.11)

ASIAN -0.12 (-3.04)

-0.11 (-2.94)

-0.16 (-3.18)

RUSSIAN -0.07 (-2.35)

-0.03 (-1.48)

-0.30 (-3.74)

NINE11 0.16 (2.35)

0.21 (3.85)

0.23 (4.42)

2R 0.14 0.12 0.15 0.12 0.11 0.09 S.E. of reg 0.31 0.31 0.31 0.31 0.31 0.32 * t-statistics use White heteroskedasticity consistent standard errors

Table 4e Sample period 1984:1 to 2008:1 (t-statistics in parentheses)* Dependent variable: tSTF4 (mean = 1.02)

Constant 1.12 (7.67)

1.09 (7.05)

1.08 (7.92)

1.05 (7.31)

0.98 (9.42)

0.96 (8.54)

tFFR 0.33 (1.20)

0.36 (1.26)

0.33 (1.29)

0.37 (1.35)

0.42 (1.73)

0.47 (1.79)

94D -0.32 (-1.89)

-0.28 (-1.63)

tFFRD *94 -0.24 (-0.81)

-0.24 (-0.73)

95D -0.27 (-1.74)

-0.24 (-1.48)

tFFRD *95 -0.31 (-1.07)

-0.32 (-1.03)

99D -0.21 (-1.63)

-0.19 (-1.37)

tFFRD *99 -0.34 (-1.27)

-0.41 (-1.34)

87CRASH 0.61 (4.32)

0.65 (5.01)

0.73 (7.24)

ASIAN 0.17 (2.35)

0.18 (2.30)

0.02 (0.14)

RUSSIAN -0.40 (-4.88)

-0.35 (-4.63)

-0.80 (-6.10)

NINE11 -0.04 (-0.22)

0.06 (0.36)

0.06 (0.40)

2R 0.13 0.11 0.12 0.11 0.10 0.1`0 S.E. of reg 0.55 0.56 0.55 0.56 0.55 0.56 * t-statistics use White heteroskedasticity consistent standard errors

Transparency and the ex post forecast errors

Ex post forecast error:

tn

i t

jjttjt n

TFETFE

1

2)(

where tT is realized value of the three month T-bill rate for quarter t.

ttt

ttjt

NINERUSSIAN

ASIANCRASHFE

11

87transindex

54

32t10

999594 *5.0*5.0*5.0transindex DDDt

Table 5 Sample period 1982:3 to 2008:1 (t-statistics in parentheses)* Dependent variable:

tFE0 tFE1 tFE2 tFE3 tFE4

Constant 0.13 (5.79)

0.53 (7.45)

0.96 (8.73)

1.26 (9.36)

1.51 (9.35)

Transindex -0.01 (-0.53)

-0.16 (-2.28)

-0.28 (-2.63)

-0.33 (-2.55)

-0.36 (-2.38)

87CRASH -0.04 (-1.57)

-0.38 (-5.34)

-0.80 (-7.26)

-1.13 (-8.40)

-1.07 (-6.06)

ASIAN -0.03 (-1.77)

0.08 (1.79)

-0.57 (-8.05)

-0.78 (-9.31)

-0.97 (-9.99)

RUSSIAN -0.07 (-4.86)

0.42 (8.97)

0.26 (3.74)

-0.04 (-0.78)

-0.09 (-0.91)

NINE11 0.03 (1.21)

1.12 (16.54)

1.21 (12.19)

2.06 (17.58)

3.02 (22.69)

2R -0.04 0.07 0.06 0.09 0.10

S.E. of reg 0.14 0.44 0.67 0.82 0.97 Mean of dependent variable

0.12 0.43 0.77 1.04 1.28

* t-statistics use White heteroskedasticity consistent standard errors

Conclusions: Regression results suggest that:

1) The 1994 and 1995 change in monetary policy transparency had the biggest impact of the level of forecast dispersions:

The standard deviation of forecasts decreased (relative to their sample means) by about: -30% to -50% for the 1 quarter ahead T bill forecasts -20% to -30% for the 2 quarter ahead T bill forecasts -20% for the 3 quarter ahead T bill forecasts

2) The range of the decrease in forecast dispersions is similar for the 4 and 5 quarter ahead forecast dispersions, although the estimates are not statistically significant.

3) The responsiveness of forecast dispersions to changes in the Federal funds rate decreases due to increased transparency but the coefficients are not precisely estimated.

4) Controlling for financial crises, the ex post T-bill forecast errors have decreased over time as monetary policy has become more transparent.

In sum: Controlling for financial/political crisis and the absolute value of the change in the Federal funds rate, the changes in monetary policy transparency introduced in 1994 and 19995 resulted in a large reduction in T-bill forecast dispersions and in our interpretation uncertainty about future monetary policy, 1 to 3 quarters into the future.