Monetary Policy Shocks and Stock Returns: Evidence ... - CiteSeerX

260 downloads 266 Views 48KB Size Report
i∆ , is proxied by the change in the rate of the three-month sterling. LIBOR futures contract as traded on the Euronext/LIFFE market, relative to the day before the.
Monetary Policy Shocks and Stock Returns: Evidence from the British Market

A. Gregoriou a, A. Kontonikas b*, R. MacDonald b, A. Montagnoli c

a

Brunel Business School, Economics and Finance Section, Brunel University, Uxbridge, UB8 3PH, UK b

Department of Economics, University of Glasgow, Adam Smith Building, Glasgow, G12 8RT, UK c

Department of Economics, University of Stirling, Stirling, FK9 8BR, UK

September 2006

Abstract This paper examines the impact of anticipated and unanticipated monetary policy announcements, of the Bank of England’s Monetary Policy Committee on UK sectoral stock returns. The monetary policy shock is generated from the change in the three-month sterling LIBOR futures contract. Using a panel GMM estimator we find that both the expected and unexpected components of monetary changes are significant, but that only the surprise term is significant when we control for the impact of the sectors financial position.

JEL classification: C33; E44; E52; G13. Keywords: Asset Prices; Monetary Policy; Panel Data.



Corresponding author: A. Kontonikas, Department of Economics, University of Glasgow, Adam Smith Building, Glasgow, G12 8RT, UK. Email: [email protected]

1

1.

Introduction In this paper we examine the impact of anticipated and unanticipated monetary policy

announcements, of the Bank of England’s Monetary Policy Committee (MPC), on UK sectoral stock returns. The monetary policy shock is generated form the change in the three-month sterling LIBOR futures contract at the time of the MPC meeting and the sample period runs from June 1999 to November 2005. Using a panel GMM estimator we find that both the expected and unexpected components of monetary changes are significant, but that only the surprise term is significant when we control for the impact of the sectors financial position. This latter finding is consistent with the efficient markets hypothesis. The rest of the paper is structured as follows. The next section describes the stock market data and the calculation of the monetary policy shock. Section 3 presents the empirical model and results. Section 4 concludes.

2.

Data Our stock returns dataset covers seventy two FTSE Industrial sub-sectors which form the ten

basic UK industries: oil & gas, basic materials, industrials, consumer goods, healthcare, consumer services, telecommunications, utilities, financials, and technology. We measure stock returns for subsector i at day t that the Monetary Policy Committee (MPC) meets, yit,, as the first difference of the natural log of the daily closing stock price (Sit): yit = 100*(ln Sit − ln Si ,t −1 ) . The data was collected

from Datastream. Following Kuttner (2001), we use data from interest rate futures contracts in order to derive the monetary policy shock. In the UK, there is no futures contract tracking the MPC-controlled policy instrument (such as the two-week repo rate used in US-based studies). The closest substitute that

2

exists, and which we use, is a futures contract based upon the three-month LIBOR rate,1 and this rate is widely accepted as a very good indicator of market expectations of future policy changes2. Thus, the monetary policy shock, ∆i u , is proxied by the change in the rate of the three-month sterling LIBOR futures contract as traded on the Euronext/LIFFE market, relative to the day before the (monthly) MPC meeting: ∆i u = f m ,t − f m ,t −1

(1)

where f m ,t is the implied futures rate (100 minus the contract price) associated with the contract that expires on the month that the MPC meets3. Finally, we measure the expected change in interest rates,

∆i e , as the actual change in the three-month LIBOR rate minus the surprise: ∆i e = ∆ i − ∆ i u

(2)

The sample period under investigation is June 1999 – November 2005, providing us with 79 MPC meetings (which is the time-series dimension of the panel).

3.

Econometric model and results

The OLS (fixed effects model) is the most common panel estimator. However a key problem with this OLS model is that it does not deal with the likely presence of endogeneity in our data. Although an Instrumental Variable (IV) estimator would address such endogeneity it, in turn, fails to

1

The futures contract is based on the British Bankers’ Association London Interbank Offered Rate (BBA LIBOR) for three month sterling deposits at 11:00 on the last trading day. The settlement price will be 100.00 minus the BBA LIBOR rounded to three decimal places. 2 Lindholdt and Wetherilt (2004) employ LIBOR rates at various maturities and find that there has been a clear improvement in the ability of the market to forecast policy rate changes by the Bank of England. 3 For example, to calculate the monetary policy shock associated with the 9/06/2005 MPC meeting, we use the rate implied by the three-month LIBOR futures contract that expires on 13/06/2005. Kuttner (2001) and Bernanke and Kuttner (2005), in their analysis of US monetary announcements adjust the federal funds rate for the number of days remaining in the month. This is necessary in their case because they use the federal funds futures rate, the settlement of of which is based upon the average fed funds rate of the last month in the futures life. We do not adjust our LIBOR rates for the number of remaining days in month because in the UK the 3 month LIBOR futures settles at the 3m-libor of the last trading day.

3

capture the cross-sectoral heterogeneity that exists in our data set. Therefore, in order to tackle both the endogeneity and cross-sectoral heterogeneity in the data we use a GMM estimator. Specifically we use the GMM estimators developed by Arellano and Bond (1991) which makes use of internal instruments for each time period to tackle endogeneity. The econometric model is given by:

yit = α i + γ t + X it + eit

(3)

where yit is defined as above, α i is the time-invariant unobserved sector-specific fixed effect (e.g. differences in the initial level returns), γ t captures the unobservable individual –invariant time effects (e.g. shocks that are common to all sectors), X is a vector of the explanatory variables. We examine four alternative cases regarding the explanatory variables vector in Eq. (3). In models (3.1) and (3.3)

X includes only the expected or unexpected interest rate change. In order to examine the robustness of our results with respect to the presence of additional control variables related to the financial position of each sub-sector, models (3.2), and (3.4) include the dividend yield (dy) and the price to cash flow ratio (pc) as additional variables in X 4. If E ( eit eiz ) = 0 holds for z ≠ t across all the sectors then it represents the following moment conditions:

E ( yi ,t − z ∆eit ) = 0 for z ≥ 2; t = 3,......., T .

(4)

If X it are weakly exogenous then we also have the following additional moment conditions: E ( X i ,t − z ∆eit ) = 0 for z ≥ 2; t = 3,......., T .

(5)

4

We experimented also with the price-earnings ratio, the market value, and the price-to-book value. These variables turned out insignificant and subsequently dropped from the analysis.

4

The single equation GMM panel estimator generally specifies a dynamic panel model in first differences and exploits the above moment conditions5. Therefore, the lagged (two time periods or more) levels of endogenous and weakly endogenous variables of the model become appropriate instruments for addressing endogeneity. The single GMM panel estimator provides consistent coefficient estimates. [INSERT TABLE 1 HERE]

The estimation results are presented in Table 1. The fixed and time effects of the panels appear significant, implying that the firm and time specific shocks differ significantly across the firms in our dataset. In addition, a test for first order residual serial correlation is insignificant, which suggests that the panel results do not suffer from serial correlation. The Sargan tests confirm the validity of the instruments in the GMM model6. Estimates of models (3.1) and (3.3) indicate that the expected and unexpected interest rate change is significantly positive and negative, respectively, in line with previous evidence for the US market by Bernanke and Kuttner (2005). However, once we control for the impact of a sector’s financial position, only the surprise component of interest rate changes remains statistically significant. This is in line with the efficient markets hypothesis which predicts that it is only the unanticipated effects of economic policy which influence asset prices Variables pc and dy are positive and significant which is in conformity with the accounting literature (Akbar and Stark, 2003) where higher dividends and higher cash flows are associated with higher financial returns.

5

The Model is transformed into first differences in order to eliminate the fixed effects. It should be noted that the serial correlation test for the GMM is done on the first difference of the residuals because of the transformations involved.

6

5

4.

Conclusions This paper investigates the impact of anticipated and unanticipated monetary policy

announcements, of the Bank of England’s Monetary Policy Committee (MPC), on UK sectoral stock returns. The monetary policy shock is generated from the change in the three-month sterling LIBOR futures contract for a sample period from June 1999 to November 2005. Using a panel GMM estimator we show that both the expected and unexpected components of monetary changes impact significantly on stock returns, but that only the surprise term is significant when we control for the impact of the sectors financial position. This latter finding is consistent with the efficient markets hypothesis.

6

References

Akbar, S. and A. Stark. “Deflators, Net Shareholder Cash Flows, Dividends, Capital Contributions and Estimated Models of Corporate Valuation”, Journal of Business Finance and Accounting, 2003, 30, 1211-1233. Arellano, M. and S. Bond. “Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations”, Review of Economic Studies, 1991, 58(2), 277-297. Bernanke, B. and K. Kuttner. “What Explains the Stock Market's Reaction to Federal Reserve Policy?”, Journal of Finance, 2005, 60, 1221-1257. Kuttner, K. “Monetary Policy Surprises and Interest Rates: Evidence from the Fed Funds Futures Market”, Journal of Monetary Economics, 2001, 47, 523-544. Lildholdt, P. and A. Wetherilt. “Anticipation of Monetary Policy in UK Financial Markets”, Bank of England Working Paper 241, 2004.

7

Table 1: GMM panel estimates, 1999-2005. Variable

Model 3.1

Model 3.2

Model 3.3

Model 3.4

constant

8.57 (5.14)***

9.38 (8.86)***

8.55 (4.49)***

9.35 (4.71)***

pc dy

---

0.01 (6.84)***

---

0.01 (6.81)***

---

0.28 (25.68)***

---

0.29 (27.22)***

∆i e ∆i u αi γt σ

0.475 (1.88)*

0.102 (1.48)

---

---

---

---

-0.591 (-1.90)*

-0.186 (-2.22)***

[0.00]

[0.00]

[0.00]

[0.00]

[0.00]

[0.00]

[0.00]

[0.00]

0.141

0.04

0.141

0.04

AR(1)

(0.296)

(0.383)

(0.273)

(0.383)

242.3[461]

248.6[472]

242.7[462]

248.6[472]

Sargan

χ 2 (r )

NOTES: For the GMM panel estimator AR(1) is the first order Lagrange Multiplier test performed on the first difference of the residuals because of the transformations involved. α i and γ t are the fixed and time effects. Sargan tests follow a

χ 2 distribution with r degrees of freedom under the null hypothesis of valid instruments. The endogenous explanatory variables in the panel are GMM instrumented setting z ≥ 3. (.) are p values, (.) are t statistic. * , **, and *** indicate significance at the 10%, 5%, and 1% level, respectively.

8