Extracting the monetary policy surprises

6. We extract monetary policy surprises by calculating the change in the forward rate agreements on the days of the policy rate announcement by the MPC. Since the market may react to the lack of change in the policy rate, if a change had been anticipated, the calculation includes all the MPC decisions, as follows:

where ${\mathrm{\Delta }i}_t^u$ is the unexpected change in the policy rate, and t denotes the day of the MPC announcement. As in Gregoriou et al. (2009), the forward rate for three months (FRA1X4) is used as an indicator of market expectations of future policy change. Once the unexpected change in the policy rate is computed, the expected change can be derived as the difference between the actual and the unexpected change in the policy rate, as follows:

where ${\mathrm{\Delta }i}_t^e$ and Δi_t denote the expected and the actual changes in the policy rate on the day of the MPC announcement, respectively.

7. The expected and unexpected changes in the policy rate are calculated from July 2004 until December 2014.⁴ During this period, there were only two occasions in which the MPC deviated from this practice: the first, on October 22, 2008, when the MPC held an extraordinary meeting and decided to increase the interest rate by 300bp to 11½ percent to curb the pressures on the exchange rate. The second occasion was on December 8, 2008, at the first scheduled MPC meeting of that month, by reducing the interest rate by 50bp to 10½ percent.

8. Overall, the sample includes 128 meetings, in which the MPC considered to change the policy rate (Table 1). Data on the timing of MPC meetings and the policy rate decisions was taken from the MNB website and data on forward rate agreements was taken from Bloomberg. In slightly more than half of the meetings (52 percent), the MPC decided to change the policy rate. Interestingly, the data also shows that the frequency of policy rate cuts is significantly higher than that of policy rate hikes, reflecting in part the MNB’s cautious and gradual approach in easing the monetary policy stance, while monetary tightening was applied rather abruptly, often in response to adverse shocks.

^{Table 1}.

Monetary Policy Council Meetings, July 2004–December 2014

^{Table 1}.

Monetary Policy Council Meetings, July 2004–December 2014

# of MPC meetings to discuss the policy rate			128
# of changes in the policy rate			66
	Of which
		Policy rate cuts	53
		Policy rate hikes	13
# of no change in the policy rate			62

View Table

^{Table 1}.

Monetary Policy Council Meetings, July 2004–December 2014

# of MPC meetings to discuss the policy rate			128
# of changes in the policy rate			66
	Of which
		Policy rate cuts	53
		Policy rate hikes	13
# of no change in the policy rate			62

Results

9. The results show that about 38 percent of the monetary policy decisions were fully anticipated, i.e., did not trigger a change of 5bps or more in the forward rate agreements. The majority of these cases occurred when the MNB decided to keep the policy rate on hold (Table 2). Additionally, the results show that, while revisions to market expectations were to both sides when the MNB decided to keep the policy rate on hold or to tighten, in episodes of policy rate cuts, the actual change in the policy rate was larger than expected by the market (chart). In this regard, in 72 percent of the decisions involving policy rate cuts, expectations were for a smaller policy rate cut and—in two occasions—expectations were for a policy rate hike. The “undershooting” of expectations in episodes of monetary policy easing is also prominent in magnitude: the actual policy rate cuts were 32bps, on average, while the expected policy rate cuts were 23bps on average.

^{Table 2}.

Policy Change and Market Expectations, July 2004–December 2014

^1/Unexpected change was below 5bps.

^{Table 2}.

Policy Change and Market Expectations, July 2004–December 2014

	Policy rate cuts	Policy rate hikes	No change
# of episodes which resulted in a looser-than-expected stance	38	5	11
# of episodes which resulted in a tighter-than-expected stance	9	3	13
# of episodes in which the actual change was fully anticipated 1/	6	5	38
Total	53	13	62

^1/Unexpected change was below 5bps.

View Table

^{Table 2}.

Policy Change and Market Expectations, July 2004–December 2014

	Policy rate cuts	Policy rate hikes	No change
# of episodes which resulted in a looser-than-expected stance	38	5	11
# of episodes which resulted in a tighter-than-expected stance	9	3	13
# of episodes in which the actual change was fully anticipated 1/	6	5	38
Total	53	13	62

^1/Unexpected change was below 5bps.

Hungary. Actual and Expected Change in the Policy Rate, July 2004-December 2014

Citation: IMF Staff Country Reports 2015, 093; 10.5089/9781484314555.002.A003

Source: IMF staff calculations.* Chart excludes the decision of October 22, 2008.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Hungary. Actual and Expected Change in the Policy Rate, July 2004-December 2014

Citation: IMF Staff Country Reports 2015, 093; 10.5089/9781484314555.002.A003

Source: IMF staff calculations.* Chart excludes the decision of October 22, 2008.

• Download Figure
• Download figure as PowerPoint slide

Hungary. Actual and Expected Change in the Policy Rate, July 2004-December 2014

Citation: IMF Staff Country Reports 2015, 093; 10.5089/9781484314555.002.A003

Source: IMF staff calculations.* Chart excludes the decision of October 22, 2008.

• Download Figure
• Download figure as PowerPoint slide

10. The “undershooting” of expectations was also prominent in the two monetary policy easing cycles since the onset of the crisis. In these two episodes (December 2008–October 2010, and August 2012–July 2014), the markets’ expectations were almost persistently below the actual policy rate cuts and, accordingly, expectations were revised downwards following the policy rate announcements (chart).

Hungary. The Unexpected and Expected Change in the Policy Rate

Citation: IMF Staff Country Reports 2015, 093; 10.5089/9781484314555.002.A003

Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Hungary. The Unexpected and Expected Change in the Policy Rate

Citation: IMF Staff Country Reports 2015, 093; 10.5089/9781484314555.002.A003

Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

Hungary. The Unexpected and Expected Change in the Policy Rate

Citation: IMF Staff Country Reports 2015, 093; 10.5089/9781484314555.002.A003

Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

11. The uncertainty surrounding the monetary policy actions has recently declined significantly. The unexpected change in the policy rate has moderated (in absolute values), and the standard deviation of the unexpected change in the policy rate (adjusted to the actual change in the policy rate) has reached the lowest level in nine years. This perhaps suggests that the MNB’s transparency with regard to the direction and magnitude of future policy rate changes has somewhat increased (chart).

Hungary. The Unexpected Change in the Policy Rate

Citation: IMF Staff Country Reports 2015, 093; 10.5089/9781484314555.002.A003

* S.D adjusted to the average change in the policy rate (in absolute values, 18-month MA).Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Hungary. The Unexpected Change in the Policy Rate

Citation: IMF Staff Country Reports 2015, 093; 10.5089/9781484314555.002.A003

* S.D adjusted to the average change in the policy rate (in absolute values, 18-month MA).Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

Hungary. The Unexpected Change in the Policy Rate

Citation: IMF Staff Country Reports 2015, 093; 10.5089/9781484314555.002.A003

* S.D adjusted to the average change in the policy rate (in absolute values, 18-month MA).Source: IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

Comparison with peers

12. We compare the monetary policy surprises in Hungary to those in Poland and the Czech Republic. While the comparison needs to be taken with caution given the differences in the economic environment and magnitude of the shocks faced by the three economies, it shows that the distribution of the unexpected changes in the policy rate differed significantly (Table 3). In particular, while in all three countries the mean of the monetary policy surprises is not significantly different from zero, Hungary had the highest variability of monetary policy surprises (adjusted to the changes in the policy rate). This in part reflects the sizable shock that the economy faced in 2008 and the high volatility of inflation. In the last four years, however, the volatility of the unexpected change in policy rate in Hungary declined significantly below that of its regional peers.⁵

^{Table 3}.

The Unexpected Change in the Policy Rate (in percentage points)

¹For Hungary the sample period is 7/2004–7/2014; for Czech Rep. it is 1/1999–7/2014; and for Poland it is 11/2001–7/2014.

²Adjusted to the absolute value of the average change in the policy rate.

^{Table 3}.

The Unexpected Change in the Policy Rate (in percentage points)

	Hungary	Czech Rep.	Poland
	Full sample1
Mean	−0.022	−0.021	−0.030
Maximum	1.903	0.127	0.588
Minimum	−0.846	−0.660	−1.677
Adjusted standard deviation2	2.882	1.730	2.685
	2008–09
Mean	−0.013	0.002	−0.086
Maximum	1.903	0.127	0.041
Minimum	−0.714	−0.100	−1.677
Adjusted standard deviation 2	9.203	0.418	5.466
	2010–2014
Mean	−0.026	−0.002	0.002
Maximum	0.128	0.082	0.185
Minimum	−0.240	−0.120	−0.265
Adjusted standard deviation2	0.966	1.106	2.113

¹For Hungary the sample period is 7/2004–7/2014; for Czech Rep. it is 1/1999–7/2014; and for Poland it is 11/2001–7/2014.

²Adjusted to the absolute value of the average change in the policy rate.

View Table

^{Table 3}.

The Unexpected Change in the Policy Rate (in percentage points)

	Hungary	Czech Rep.	Poland
	Full sample1
Mean	−0.022	−0.021	−0.030
Maximum	1.903	0.127	0.588
Minimum	−0.846	−0.660	−1.677
Adjusted standard deviation2	2.882	1.730	2.685
	2008–09
Mean	−0.013	0.002	−0.086
Maximum	1.903	0.127	0.041
Minimum	−0.714	−0.100	−1.677
Adjusted standard deviation 2	9.203	0.418	5.466
	2010–2014
Mean	−0.026	−0.002	0.002
Maximum	0.128	0.082	0.185
Minimum	−0.240	−0.120	−0.265
Adjusted standard deviation2	0.966	1.106	2.113

¹For Hungary the sample period is 7/2004–7/2014; for Czech Rep. it is 1/1999–7/2014; and for Poland it is 11/2001–7/2014.

²Adjusted to the absolute value of the average change in the policy rate.

13. The undershooting of expectations was also prevalent in the Czech Republic. The analysis shows that the expected cut in the policy rate was below (5bps or more) the actual cut in the policy rate in 64 percent of the times (compared to 30 percent in Poland), while, in episodes of policy rate hikes, the expected increase in the policy rate was below the actual increase in 91 percent of the times (compared to 26 percent in Poland).

The baseline model

14. To explore whether monetary policy surprises have an impact on financial market developments, we examine the change in key financial market indicators on the day of the policy rate announcement (t). The financial market indicators that we consider are: the daily change in the yields of 5-year and 10-year government bonds (ΔGB5_t and ΔGB10_t, respectively), and the daily change of the stock market price index (ΔBUX_t). In addition, we look at the impact of the monetary policy surprise on sovereign risk premia (measured by Hungary’s 5-year CDS spread, ΔCDS_t) given that monetary policy surprises may change investors’ perceptions regarding financial stability risks. The relationship between monetary policy surprises and the financial variables can be expressed as follows:

where ΔY_t is a vector of the aforementioned financial market indicators. The coefficient β₁ measures the response of the financial market indicators to the monetary policy surprise, ${\mathrm{\Delta }i}_t^{un}$. The coefficient β₂ measures the response of the financial market indicators to the expected change in the policy rate, though the prior assumption is that there should be no response since financial indicators are likely to internalize market expectations before the monetary policy announcement takes place. Additionally, the specification contains a couple of additional variables (X_j): the change in the Standard and Poor’s 500 stock price index (ΔS&P500_t) to control for the impact of external factors, and a dummy for monetary easing cycles (MP_Easing), which is also interacted with both the expected and unexpected change in the policy rate. This dummy is included to evaluate whether the response of the market to these two components depends on the direction of the policy rate change. The tendency of market expectations to undershoot the change in the policy rate in episodes of monetary policy easing may suggest that there is an asymmetric effect. A constant (α) is included to capture any trends in the dependent variables, and ε_t is a white noise.

15. Table 1A in the Appendix presents a selection of descriptive statistics for the variables included in the estimation. The statistics are reported for 128 observations for the examined period (July 2004–December 2014). The correlation matrix, which is presented in Table 2A in the Appendix, shows that the correlation of the unexpected change in the policy rate with the financial market indicators has broadly the expected sign: There is a positive correlation with the yields on 5-year and 10-year government bonds as well as with the CDS spread. The latter may suggest that larger-than-expected increase (reduction) in the policy rate is associated with episodes of increased (lower) uncertainty in the financial markets. In addition, there is a relatively low correlation between the unexpected and the expected change in the policy rate, and between the unexpected change in the policy rate and the change in the S&P500, thus dismissing concerns about possible co-linearity.

16. Estimation results are presented in Table 4. They indicate that monetary policy surprises had a significant impact on all four financial market indicators and, for some indicators, the unexpected changes in the policy rate have a non-trivial effect. More specifically, while the results suggest that the impact was less pronounced during episodes of monetary policy easing, a surprise of 100bps contributed on average to a change of nearly 40bps on the yield on the 5-year government bond during other episodes.⁶ The impact on the 10-year yield is somewhat lower—as was also found in Kuttner (2001) and Rezessy (2005)—suggesting that a surprise of 100bps contributed on average to a change of nearly 20bps. The latter result appears significant in episodes of monetary easing and tightening as well as after decisions to keep the policy rate on hold.

^{Table 4}.

The Effect of Monetary Policy Surprises on Selected Financial Market Indicators¹

¹The model is estimated using Ordinary Least Squares (OLS) with Newey-West standard errors.

²Expressed in percentage points.

* Indicates significance level of 5 percent, ** indicates significance level of 10 percent.

^{Table 4}.

The Effect of Monetary Policy Surprises on Selected Financial Market Indicators¹

	ΔGB52		ΔGB102		ΔBUX		ΔCDS2
c	1.292	1.064	1.748	1.512	−0.305**	−0.299	1.273	1.667
Δiun,2	0.396*	0.376*	0.197*	0.163*	−0.017*	−0.011*	0.167*	0.169*
Δiun * MP_Easing	−0.328*	−0.304*	−0.082	−0.036	0.012	−0.008	−0.141*	−0.185*
Δie,2		0.026		0.005		0.007		−0.097
Δie * MP_Easing		−0.012		0.033		−0.038*		0.052
ΔS&P500		−0.332		−1.17**		0.330*		−1.364*
MP_Easing	−5.596*	−4.977**	−6.52*	−4.85**	0.919*	−0.055	−2.326	−3.635**
Adjusted R2	0.342	0.328	0.179	0.186	0.104	0.247	0.128	0.205
D.W statistic	2.082	2.111	1.878	1.981	2.027	2.117	1.967	1.897
# of Obs.	128	128	128	128	128	128	128	128

¹The model is estimated using Ordinary Least Squares (OLS) with Newey-West standard errors.

²Expressed in percentage points.

* Indicates significance level of 5 percent, ** indicates significance level of 10 percent.

View Table

^{Table 4}.

The Effect of Monetary Policy Surprises on Selected Financial Market Indicators¹

	ΔGB52		ΔGB102		ΔBUX		ΔCDS2
c	1.292	1.064	1.748	1.512	−0.305**	−0.299	1.273	1.667
Δiun,2	0.396*	0.376*	0.197*	0.163*	−0.017*	−0.011*	0.167*	0.169*
Δiun * MP_Easing	−0.328*	−0.304*	−0.082	−0.036	0.012	−0.008	−0.141*	−0.185*
Δie,2		0.026		0.005		0.007		−0.097
Δie * MP_Easing		−0.012		0.033		−0.038*		0.052
ΔS&P500		−0.332		−1.17**		0.330*		−1.364*
MP_Easing	−5.596*	−4.977**	−6.52*	−4.85**	0.919*	−0.055	−2.326	−3.635**
Adjusted R2	0.342	0.328	0.179	0.186	0.104	0.247	0.128	0.205
D.W statistic	2.082	2.111	1.878	1.981	2.027	2.117	1.967	1.897
# of Obs.	128	128	128	128	128	128	128	128

¹The model is estimated using Ordinary Least Squares (OLS) with Newey-West standard errors.

²Expressed in percentage points.

* Indicates significance level of 5 percent, ** indicates significance level of 10 percent.

17. Similar to previous studies (Bernanke and Kuttner, 2005, and Gregoriou et al. 2009), the results confirm a negative effect of monetary policy surprises on stock market returns. In this regard, a positive (negative) monetary policy surprise of 100bps contributed to a decline (increase) of nearly 1.1–1.7 percent in stock market prices. The response of the latter may reflect a rebalancing of investors’ portfolio as the new information about the trajectory of the policy rate can potentially affect the firms’ future liquidity, cost of capital, and overall profitability.⁷

18. The results also suggest that positive monetary policy surprises had a perverse impact on the sovereign risk premia. The estimations show that positive (negative) monetary policy surprises during episodes of monetary tightening or unchanged policy stance were associated with an increase (decline) in the CDS spread. One plausible explanation is that episodes of monetary tightening or unchanged policy rate were associated with episodes of increased uncertainty and pressure on reserves, and positive monetary policy surprises tended to exacerbate markets’ concerns about financial stability risks.⁸

19. As expected, the coefficients of Δi^e turned out to be insignificant, thus implying that, by and large, market expectations are already internalized in financial variables. The only exception is the impact on the stock market prices during the monetary policy easing cycles, which suggests that that expectation for a lower policy rate contributed to an increase in the stock prices.

Robustness

20. We complement the baseline model with three alternative specifications to ensure the robustness of the results. First, we test whether the results remain valid for the recent period, which includes the global financial crisis and the subsequent economic recovery. Second, we examine whether the magnitude of the effect of the monetary policy surprises are dependent on sign of the surprise. And third, we re-evaluate the results by controlling for regional shocks rather than global financial shocks.

• Sample stability: Eq. (1) was estimated for the pre-crisis period and for the subsequent period. The collapse of Lehman Brothers (LB) in September 2008 was chosen as the starting point of the second period (“post-LB”). The results, which are presented in Table 3A in the Appendix, show that the effect of an unexpected change in the policy rate on the yields of government bonds is significantly higher in the pre-LB collapse, but only during monetary tightening and unchanged monetary policy stance. The impact on the risk premia is evident only in the post LB collapse, and only during monetary tightening or unchanged monetary policy stance. The effect on the stock market prices is significant and negative only during the post LB collapse.

• Direction of the surprise: As the magnitude of the markets’ response may depends on the sign of the surprise (e.g., Wang and Mayes, 2012), a dummy (Neg_surprise), which obtains a value of one for 72 episodes in which the monetary policy surprise was negative, was included with interaction with the unexpected change in the policy rate.⁹ The results, which are presented in Table 4A in the Appendix, show that the interaction term Δi^un * Neg_surprtse has a negative and significant impact on AGB5 and ADM, suggesting an asymmetry in the magnitude of the effect. More specifically, they show that a positive monetary surprise of 100bps increases the yield on 5-year government bonds by about 35bps while a similar size negative surprise reduces it by about 16bps.¹⁰ In the case of ΔCDS, the results show that only a positive monetary surprise contributed to an increase in the sovereign spread, though the magnitude is relatively small.

• Regional shocks: As Hungary’s financial variables may be driven more by regional events rather than global developments, we replaced ΔS&P500 with the change Poland’s sovereign CDS spread (ΔPOL_CDS) as an alternative specification.¹¹ The results, which are presented in Table 4A in the Appendix, indicate that, even controlling for the regional effects, the unexpected change in the policy rate had a significant effect on the financial market indicators. In this regard, the effect on the yield of 10-year government bonds was found significant regardless of the policy rate’s direction; however, the effect of the monetary policy surprise on 5-year yields and the change in the sovereign CDS spread is significant, but mainly during episodes of monetary tightening and unchanged policy rate. Interestingly, stock market prices reacted to both monetary surprises and expected change in the policy rate during monetary policy easing cycles.