Monetary Policy Announcements and Exchange Rates in Kenya Ferdinand Othieno1 & Ruth Wanjiru 2
Abstract The study examines how monetary policy announcements affect the level and volatility of exchange rates in Kenya for the period 2005-2014. The exchange rates under study are: the KES/USD, KES/GBP, KES/EUR and KES/UGX. The study uses the GARCH (1, 1) and EGARCH taking into account monetary policy announcements using dummy variables to analyze the relationship between monetary policy announcements and exchange rates. We find evidence that the monetary policy announcements do not significantly influence the level of exchange rates except where the announcement directly involves the exchange rate. Only announcements regarding foreign exchange intervention significantly influenced exchange rate volatility as was the case in the period 2011-2012. The long run variance indicated lack of persistence of shocks on the exchange rate volatility which provides further evidence of the relative insignificance of monetary policy announcements on exchange rate volatility. We conclude that in Kenya, monetary policy announcements are merely secondary determinants of exchange rate volatility. The current study differs from previous studies that have focused the impact of exchange rate volatility on international trade and economic growth and the Central Bank of Kenya response to changes in inflation, exchange rate and GDP growth literature regarding the relationship between monetary policy and exchange rate volatility in a frontier market. Key Words: Monetary policy announcements, Foreign Exchange, Volatility, GARCH, Kenya JEL Classification: E52, B23, F31
1
Correspondence to: Ferdinand Othieno, School of Finance and Applied Economics, Strathmore University, P.O. Box 59857, 00200-Nairobi, Kenya, Email:
[email protected] 2
Ruth Wanjiru, School of Finance and Applied Economics, Strathmore University, P.O. Box 59857, 00200-Nairobi, Kenya, Email:
[email protected]
1
1
Introduction
The relationship between monetary policy announcements and asset prices is examined by various policy transmission channels. (Mishkin, 1995) argues that under the interest rate channel, monetary tightening raises interest rates and with it the cost of capital. This reduces investment spending, therefore reducing output. Under the wealth effects channel, contractionary monetary policy leads to a decline in stock prices as investors offload their relatively liquid stock portfolios to meet other needs. The decline in asset prices leads to a fall in financial wealth which reduces consumption and ultimately output. With regard to the exchange rate channel, (Taylor, 1995) argues that monetary tightening leads to a subsequent rise in interest rates leading to an appreciation of the domestic currency. Net exports decline as they become more expensive, which ultimately reduces output. Research on the effect of monetary policy on asset prices has focused on the effect of monetary policy on stock market volatility as stock markets are more representative due to the presence of both institutional and individual investors with varying information and information-processing capacities. See for instance (Bomfim, 2001), (Cheung & Chinn, 2001), (Bernanke & Kuttner, 2004) and (Omrane, Bauwens, & Giot, 2005). However, studies on the impact of monetary policy on exchange rate volatility are also gaining momentum for two reasons: (1) increased trade integration both regionally and globally as evidenced by (Jansen & Haan, 2002) and (2) the justification or invalidation of central banks’ exchange rate intervention as was the case for (Hutchison & Fatum, 2002). The current study benefits from the latter strand and examines the impact of monetary policy announcements on the level and volatility of exchange rates in Kenya.
1.1
Monetary policy announcements in Kenya
Macroeconomic announcements in Kenya are made by the Monetary Policy Committee (MPC) established in 2008. These statements assess the progress of the implementation of monetary policy, specify the policies and instruments to achieve policy targets, and provide rationale for the policies and instruments mix. The communication of policy intentions through the MPC’s statements and reports is vital in the implementation of policy decisions and influencing financial markets. Before the 1990s, Kenya had a fixed exchange rate regime. The shilling was pegged on the Sterling pound upon independence and later on the US dollar. During this period, the CBK focused on liquidity ratios and reserve requirements implemented through moral suasion. (Kinyua, 2000) asserts that failure of these tools led to the introduction of open market operations and the liberalization of interest rates and exchange rates in the early 1990s. According to (Ndung'u, 1999), from the early 1990s, the conduct of monetary policy was focused on the behavior of the broad money aggregate, M2 (currency in circulation plus domestic currency deposits with banks and non-bank financial institutions (NBFIs)). Increased economic integration led to a shift from M2 to M3 (M2 plus foreign currency deposits held by residents). Based on the monetary policy statements, the CBK continues to focus on monetary aggregates to stabilize inflation. The CBK participates in the foreign exchange market in order to build foreign exchange reserves, service National debt denominated in foreign currency and to quell excess exchange rate volatility, according to (Chipili, 2013) and (Utsunomiya, 2013).
2
The level and volatility of the KES/USD exchange rate for the period 2005 – 2014 is captured in Figure 1 below: Figure 1: Level and volatility of the KES/USD exchange
The first panel in Figure 1 shows that the Kenya shilling appreciated against the dollar between 2005 and early 2007. According to the Monetary Policy Statements, this appreciation is attributable to foreign capital inflows informed by a conducive macroeconomic environment and weakening of the dollar at the onset of the global financial crisis. The Kenya shilling then depreciated against the dollar due to the post-election violence. The period immediately after the formation of the Monetary Policy Committee as shown by the line is characterized by relative calm. However, in 2011, the Kenya shilling weakened drastically against the dollar due to high international oil and food prices at a time when Kenya was experiencing unfavorable weather conditions. With regard to volatility, the period between 2005 and 2008 i.e. before the formation of the Monetary Policy Committee, is characterized by relatively high volatility in the KES/USD exchange rate. Excess volatility is observed early 2008 resulting from the post-election violence and the global financial crisis. (Utsunomiya, 2013) argues that the study of the response of exchange rate volatility to macroeconomic news is important in indicating the effectiveness of intervention in stabilizing the exchange rate. (Frenkel, 1981) uses a similar study to show the efficiency of the market by invoking the martingale hypothesis which posits that in an efficient market, asset prices should adjust rapidly to new information. Research in Kenya, linking macroeconomic news with exchange rate volatility is limited. While extensive studies have been carried out in this area for developed markets, researchers in Kenya such as (Musyoki, Pokhariyal, & Pundo, 2012) have focused on the impact of exchange rate volatility on international trade and economic growth. (Rotich, Maana, & Kathanje, 2008) tested whether the CBK responds to changes in inflation, exchange rate and GDP growth. The current study, therefore fills an important gap in literature regarding the relationship between monetary policy and exchange rate volatility in a frontier market. The study investigates the effect of monetary policy announcements on the level and volatility of exchange rates using daily data spanning the period 2005-2014. The period selected provides a wide scope for comparison of exchange rate volatility before and after the inception of the 3
Monetary Policy Committee in 2008. The period was also characterized by dynamic local and external macroeconomic shocks that affected the exchange rate volatility and triggered Central bank intervention.
2
Literature Review
One of the very early studies on exchange rates is the work of (Frenkel, 1981) who studied the volatility and predictability of exchange rates, specifically the US Dollar/Pound and US Dollar/Deutshe Mark. The author invoked the martingale hypothesis where the best prediction of the next period’s exchange rate is the current spot exchange rate comprising all available information and expectations regarding the future. Upon testing the predictability, he found that the forecasts only captured a few of the actual changes. He attributed the deviation of actual from predicted changes in exchange rates to new information in line with the Martingale hypothesis. Similarly, (Galati & Ho, 2003) investigated the extent to which the fluctuations in the daily euro/dollar exchange rate were influenced by news and found a positive correlation between good news and the appreciation of the exchange rate. Recently (Laakkonen, 2007) estimated the impact of macroeconomic announcements on the US Dollar/Euro exchange rate and found a significant rise in exchange rate volatility after the announcements. She argues that the volatility resulted from varying expectations regarding the announcements’ implication. (Evans & Lyons, 2003) found that a significant 30% of volatility in exchange rates is attributable to macroeconomic news. In contrast, the study by (Bollerslev & Andersen, 1998) found that the main cause of volatility in the German-dollar exchange rate was calendar effects. They argued that the effect of macroeconomic news on the volatility of exchange rate is merely secondary. Similarly, (Almeida, Goodhart, & Payne, 1998) who used the German mark/dollar exchange rate to distinguish the effects of macroeconomic announcements from those of other variables found that effects of macroeconomic news were extremely short-lived with a span of about three hours. There consensussus among researchers regarding the impact of scheduled announcements relative to unscheduled announcements on exchange rate volatility. (Tivegna, 2001) found that unscheduled announcements have a larger impact on exchange rates relative to scheduled announcements. Unscheduled news lead to a jump in the exchange rates, which soon dissipates with volatility reverting back to the normal level contrary to scheduled announcements whose effects the author finds to be more permanent. (Andersen, Bollerslev, Diebold, & Vega, 2003) who investigated the link between exchange rates and macroeconomic news found that while unscheduled announcements have a larger impact, expected announcements also boost volatility. 2.1
Empirical Review
Speed of adjustment of exchange rate volatility is at the core of successfully disaggregating the effects of macroeconomic news from those of other factors affecting exchange rates. (Cheung & Chinn, 2001) through their survey on currency traders found that news on variables such as inflation, unemployment, trade deficit and interest rates adjusted within minutes. They also highlight the difficulty in determining the surprise element of macroeconomic announcements due to the rapid adjustment of exchange rate volatility.
4
Similarly, (Jansen & Haan, 2002) found that volatility lasts a day before reverting back to normal. The quick adjustment of exchange rate volatility is consistent with the expected market hypothesis of equity prices which implies that current prices reflect all the available information and that prices adjust rapidly to new information in the market. (Patell & Wolfson, 1984) and (Ederington & Lee, 1993) found that volatility was highest during the first five minutes before adjusting to its normal level after forty minutes. Persistency is also aligned with the timing of announcements. (Andersen, Bollerslev, Diebold, & Vega, 2003) explored the significance of timing of macroeconomic announcements and found that scheduled announcements with uncertain release time led to gradual adjustment of the volatility as opposed to immediate jumps. They argued that the gradual adjustment was as a result of inadequate liquidity in the market. The high speed of adjustment of exchange rate volatility to macroeconomic announcements has informed the use of high frequency data by various researchers testing this relationship. (Jansen & Haan, 2002) argued that use of high frequency data is essential in separation of news effects on exchange rates from effects of other factors. (Ederington & Lee, 1993) and (Goodhart, Henry, Hall, & Pesaran, 1993) used tick by tick data. The latter tested the response of volatility with or without news while the former examined the speed with which markets adjust to news releases. (Andersen, Bollerslev, Diebold, & Vega, 2003) and (Laakkonen, 2007) used high frequency (5 minute) exchange rates. (Hakkio & Pearce, 1985) and (Tivegna, 2001) used intraday exchange rate data while (Jansen & Haan, 2002) and (Hutchison & Fatum, 2002) used daily Euro/US Dollar exchange rate data to determine the impact of statements made by ECB officials. Inspite of the use of high frequency data, volatility adjustment is still very rapid. Some researchers such as (Almeida, Goodhart, & Payne, 1998) have argued that the lack of volatility persistency of news effects indicates the relative insignificance of macroeconomic news on exchange rate volatility. They concluded that this short-lived effect signaled that macroeconomic news did not substantially affect exchange rate volatility.
2.2
Leverage effects
Presence of leverage effects in exchange rate volatility has given prominence to models that capture the leverage effects and volatility clustering properties. Earlier studies employed simple regressions and used daily return data which yielded relatively inaccurate results that failed to capture various stylized facts including leverage effects present in financial time series data. (Laakkonen, 2007) asserts that the availability of high frequency data and numerous variations of the GARCH models has improved the analysis of impact of news on exchange rate volatility. (Jansen & Haan, 2002) employed an EGARCH specification to investigate the effects of ECB Officials Statements to the level and volatility of the euro-dollar exchange rate. (Utsunomiya, 2013) used both GARCH and EGARCH models to investigate the effect of intervention frequency on the foreign exchange market in Japan. (Goodhart, Henry, Hall, & Pesaran, 1993) employed an OLS and GARCH-M framework on Dollar-Pound quotations to investigate how volatility responds with or without news. They found the OLS framework to be limited due to its assumption of constant variance violated by heteroscedasticity. (Kamau, Maana, Ngugi, & Tiriongo, 2013) used a GARCH specification to investigate the structure of the volatility of the Kenya shilling against other five currencies including the USD, GBP, EUR, TZS and UGS with the intention of capturing volatility clustering present in the exchange rates under study. 5
Literature finds that among the factors affecting exchange rate volatility are the degree of trade integration, foreign exchange intervention, the exchange rate regime, calendar effects, ARCH effects and macroeconomic announcements. There have been varying findings with regard to significance of each on exchange rate volatility. There evidently a gap in literature with regard to the significance of announcements on various monetary policy instruments on exchange rate volatility. Such a study is important in indicating the effectiveness of various monetary policy instruments in a given economy as shown by (Hakkio & Pearce, 1985) and (Tivegna, 2001). For instance, where monetary policy in either carried out through money targets as opposed to inflation targets, it is imperative to see how the market responds to information on monetary aggregates as well as inflation so as to gauge if Central banks’ monetary policy intentions are aligned with market expectations.
3
Analytical framework
3.1
Population and data issues
The study uses exchange rate data, particularly the KES/USD, KES/EUR, KES/GBP and KES/UGX exchange rates and macroeconomic announcements reported in the monetary policy statements by the MPC. The exchange rate selection criteria is informed by the level of trading activity involving a given exchange rate and the foreign country’s significance as Kenya’s trading partner. (Kamau, Maana, Ngugi, & Tiriongo, 2013) who investigated the structure of volatility of exchange rates in Kenya employed the same exchange rates arguing that the US dollar, the Sterling Pound and the Euro dominate the trading transactions and the foreign reserve basket in Kenya. Further, the USA, UK and Uganda are among Kenya’s key trading partners. The use of various exchange rates allows for comparative analysis vital in examining the consistency of findings. The MPC has published 33 monetary policy statements for the period under study. Both the exchange rate data and the monetary policy statements are available in the Central Bank of Kenya website.
3.2
Data Analysis
First the structure of exchange rate returns volatility is examined to establish the unit root characteristics. The test employed is the Augmented Dickey Fuller test which caters for the serial correlation in the error terms which is potentially present in an exchange rate series. It is represented as, ∆ 𝒓𝒕 = 𝜶𝟎 + 𝜶𝟏 𝒕 + 𝜷∆𝒓𝒕−𝟏 + ∑𝒏𝒊=𝟏 𝜶𝒊 ∆𝒀𝒕−𝒊 + 𝜺𝒕
(1)
The test’s null hypothesis is that the 𝛽 = 0 i.e. the series is stationary. The coefficient 𝛼1 shows the time trend for change in return and the last term 𝜀𝑡 is the white noise error term. Stationarity tests are carried out since non-stationary time series leads to spurious regressions. Further, a GARCH (1,1) model can be represented as: 𝒉𝒕 = 𝜶𝟎 + 𝜶𝟏 𝜺𝟐𝒕−𝟏 + 𝜷𝒉𝒕−𝟏
6
(2)
The coefficients imply non-stationarity in variance (𝛼1 + 𝛽 ≥ 1) is undesirable since the conditional variance does not converge to the unconditional variance in the long-run. As such, the series found to be non-stationary should be differenced to make it stationary. Using a GARCH (1,1) model, equation (2), the ARCH coefficient is represented by 𝛼1 and the GARCH coefficient by 𝛽. The ARCH-LM tests show that heteroscedasticity is present if both the ARCH and GARCH coefficients are significant. Leverage effects can be observed using an EGARCH or a TGARCH model represented by equation (3) and equation (4) respectively. ln ℎ𝑡 = 𝑤 + 𝛽 ln ℎ𝑡−1 + 𝛾
𝜀𝑡−1 √ℎ𝑡−1
+ 𝛼 [
|𝜀𝑡−1 |
√ℎ𝑡−1
2
− √𝜋]
2 2 ℎ𝑡 = 𝜃0 + 𝜃1 𝑢𝑡−1 + 𝛽ℎ𝑡−1 + 𝛾 𝑢𝑡−1 𝐼𝑡−1
(3) (4)
In both equations, ℎ𝑡 is the conditional variance dependent on lagged values of residuals and lagged values of conditional variances. The significance of the leverage effect is given by the significance of coefficient 𝛾 in both equations. The parsimonious nature of the GARCH (1,1) makes it popular in volatility modeling studies. However, it is unable to pick out leverage effects emphasized by (Nelson, 1990). Consequently, researchers opt for extensions of GARCH models that capture leverage effects. These include the threshold GARCH (TGARCH) model and the exponential GARCH (EGARCH). 2 2 ℎ𝑡 = 𝜃0 + 𝜃1 𝑢𝑡−1 + 𝛽ℎ𝑡−1 + 𝛾 𝑢𝑡−1 𝐼𝑡−1
(5)
Equation (5) is the basic TGARCH model where ℎ𝑡 is the conditional variance that is dependent 2 on lagged squared residuals 𝑢𝑡−1 and lagged conditional variance ℎ𝑡−1 . The final term captures leverage effects such that: 𝐼𝑡−1 = 1 if 𝑢𝑡−1 < 0, 𝐼𝑡−1 = 0 otherwise. For a leverage effect, 𝛾 > 0. Given these conditions, a positive return will lead to a lower conditional variance than a negative return of the same amount. 𝐥𝐧 𝝈𝟐𝒕 = 𝒘 + 𝜷 𝐥𝐧 𝝈𝟐𝒕−𝟏 + 𝜸
𝜺𝒕−𝟏 √𝝈𝟐𝒕−𝟏
+ 𝜶[
|𝜺𝒕−𝟏 | √𝝈𝟐𝒕−𝟏
𝟐
− √𝝅]
(6)
Equation (6) is an EGARCH (1,1) model. The coefficient 𝛾 would typically be negative which reflects that positive shocks generate less volatility than negative shocks. The analysis was carried out in cycles spanning from 1st April to 31st March for each year to capture the two monetary policy announcements every year. This allowed for comparative analysis between the effects of monetary policy announcements in each cycle. It also gave allowance for dissipation of effects from both the June and December announcements which is important to avoid overlapping of effects and for easier disaggregation of the determinants of exchange rate volatility. 7
Equation (7) is the mean equation that indicates the impact of macroeconomic announcements on the level of the exchange rates. 𝑅𝑡 = 𝜃0 + ∑𝑝𝑘=1 𝜃𝑘 𝑅𝑡−𝑘 + ∑𝑛𝑖=1 ∑𝑛𝑗=1 𝜃𝑖 𝐷𝑗𝑡 + 𝜀𝑡
(7)
𝑅𝑡 is the daily logarithmic exchange rate return. Log returns make it possible to compare different variables collected at different frequencies catering for the difference between exchange rates and macroeconomic news frequency. The returns are explained by previous values 𝑅𝑡−𝑘 . The order p for the lagged values was determined using the information criteria. 𝐷𝑗𝑡 indicates occurrence dummy variables representing monetary policy announcements while 𝜀𝑡 is a white noise error term. Dummy variables have been used by researchers such as (Ederington & Lee, 1993), (Goodhart, Henry, Hall, & Pesaran, 1993), (Bollerslev & Andersen, 1998) and (Jansen & Haan, 2002) to quantify news. Equation (8) is the variance equation that captures the impact of monetary policy announcements on exchange rate volatility. ln ℎ𝑡 = 𝑤 + 𝛽 ln ℎ𝑡−1 + 𝛾
𝜀𝑡−1 √ℎ𝑡−1
+ 𝛼 [
|𝜀𝑡−1 |
2
√ℎ𝑡−1
𝑛 − √𝜋] + ∑𝑚 𝑖=1 ∑𝑗=1 𝜃𝑖 𝐷𝑗𝑡
(8)
ℎ𝑡 represents the conditional variance dependent on its lagged values and dummy variables 𝐷𝑗𝑡 that indicate the occurrence of the monetary policy announcement. The leverage effects are captured by the coefficient 𝛾. Similar to (Ederington & Lee, 1993), this study provides
an analysis into the response of exchange rate volatility with and without news by defining a series of dummy variables 𝐷𝑗𝑡 where 𝐷𝑗𝑡 = 1 if an announcement is made and 0 otherwise.
4
Results and Findings
Using Augmented Dickey Fuller Test, the study found that the KES/EUR, KES/GBP, KES/UGX and KES/USD exchange rate series are all integrated of order one, I(1), which means that they have to be differenced once to make them stationary. Table 1: Stationarity test results Exchange rates T-statistic
log_EUR
log_GBP
log_UGX
log_USD
-36.8721
-36.5537
-12.1943
-36.8830
0.0000
0.0000
0.0000
0.0000
p-values
Based on the findings in table 1 as regards stationarity of exchange rates, the same were transformed into continuously compounding returns (logeur, loggbp, logusd & logugx) by taking the first log difference of the spot rates as shown in equation (9). 𝑝𝑡
𝑅 = 𝑙𝑜𝑔 (𝑝
𝑡−1
) ∗ 100
(9)
8
4.1
Level and Volatility of Exchange rates
Figure 2 shows the trends for the spot exchange rates i.e. the level of the Kenya shilling against the EUR, GBP, USD and UGX between January 2005 and June 2014. Figure 2: Level of exchange rates for the period 2005-2014
The Kenya shilling strengthened against the USD, GBP, EUR and UGX between 2005 and 2006. The shilling was buoyed by foreign exchange inflows from the export sector and by capital inflows informed by relatively attractive domestic interest rates as reported in the 18 th and 19th Monetary Policy Statement. The Kenya shilling strengthened against the dollar in the first half of 2007 owing the weakening of the USD in International currency markets, remittances, export sector earnings and capital inflows that lead to a surplus of balance of payments. In the last quarter of 2007, the Kenya shilling strengthened against the dollar due to a surge in the Foreign Direct Investment inflows from the CFC-Stanbic merger and the sale of shares to foreign investors by Equity Bank and Telkom Kenya as well as the weakening of the US dollar internationally. In early 2008, the Kenya shilling depreciated against all four currencies as shown by the spike just before the line due to the post-election violence. The establishment of the Monetary Policy Committee in 2008 is shown by the line in the graphs. The Kenya shilling depreciated against the GBP and the Euro in early 2009. These currencies gained against the dollar in the International currency markets. Kenya shilling strengthened against the Euro and GBP in 2010 due to the Greek Debt crisis that made the USD the safe haven currency. In 2011, the Kenya shilling weakened against the USD, EUR and GBP in 2011 due to commercial banks’ speculation activities and increased demand for foreign exchange to finance the widening current account deficit. The CBK quelled the currency depreciation by amending the foreign exchange guidelines and selling foreign currency to stabilize the exchange rate. This was followed by relative stability from 2012 to mid-2014. For the KES/UGX exchange rate, there is a general appreciation throughout the period. In the second half of 2012, the Kenya shilling strengthened against the Uganda Shilling. The weakening in the Uganda Shilling was attributed to temporary suspension of budget support by the country’s development partners, which reduced foreign exchange inflows into Uganda significantly. From Figure 2 we see that the KES/GBP, KES/EUR and KES/USD exhibit a similar trend. This implies that they react to similar shocks in a near uniform fashion which suggests that the size of 9
the impact does not differ considerably. The similarity between the three exchange rates has informed the use of the KES/USD (in the next section) in addition to its dominance in the foreign exchange market transactions in Kenya and in the foreign currency reserves. The KES/UGX is also used due to the significant trading activity between Kenya and Uganda and it captures the intra-regional exchange rate dynamics within the East African Community. Figure 4: Volatility of exchange rates for the period 2005-2014
Figure 3 shows a period of tranquility between 2005 and early 2007 followed by excess volatility for the period between 2007 and 2009 mainly due to the violence that followed the 2007 elections and partly due to the global financial crisis. The line in the graphs corresponds with the establishment of the Monetary Policy Committee in 2008. The period after the committee’s establishment is characterised by relative calm with a bout of excess volatility in 2011 which emanated from the high food import prices and international oil prices as well as exchange rate speculation. For all the exchange rate returns, spikes cluster in similar periods which shows that they respond to similar shocks though the size of the impact may differ. 4.2
Comparative Analysis
This subsection explored the structure of exchange rate volatility in relation to monetary policy announcements. a)
Without Monetary policy announcements
b)
With Monetary policy announcements
4.2.1 ARCH effects Without Monetary policy announcements This scenario captures the ARCH and GARCH effects on the volatility of the exchange rate series’ in the absence of monetary policy announcements. The same is defined by the GARCH (1, 1) model as shown below. 𝑅𝑡 = 𝜃0 + 𝛼1 𝑅𝑡−1 +𝜀𝑡 . 10
(10)
2 ℎ𝑡 = 𝛼0 + 𝛼1 𝜀𝑡−1 + 𝛽ℎ𝑡−1
(11)
Table 2: Coefficients and p-values showing the ARCH and GARCH effects on the KES/USD exchange rate Period
2008-2009 2009-2010 2010-2011 2011-2012 Mean Equation:
2012-2013 2013-2014
𝑅𝑡 = 𝜃0 + 𝛼1 𝑅𝑡−1 +𝜀𝑡
Coefficients 𝜃
0.0008 (0.0834)
-0.0000 (0.6615)
0.0000 (0.6353)
-0.0001 (0.8323)
0.0003* (0.0000)
0.0002 (0.2450)
𝛼1
0.1361* (0.0313)
0.0777 (0.2826)
0.1527 (0.0779)
0.1801* (0.0033)
-0.0536 (0.2297)
0.1686 (0.0564)
Variance Equation: ht = α0 + α1 ε2t−1 + βht−1 Coefficients α0
0.0000* (0.0000)
0.0000* (0.0000)
0.0000* (0.0284)
0.0000* (0.0000)
0.0000* (0.0001)
0.0000* (0.0031)
α1
0.5489* (0.0000)
0.1625* (0.0016)
0.1953* (0.0000)
0.6724* (0.0000)
1.6543* (0.0000)
0.2639* (0.0005)
𝛽
0.3652* (0.0000)
0.6901* (0.0000)
0.8073* (0.0000)
0.3254* (0.0000)
0.2516* (0.0000)
0.6700* (0.0000)
0.0004
0.0000
0.0000
Long run Variance: var(ut ) =
0.0000 0.0000 Long run Variance *indicates that the coefficient is significant at the 5% confidence level.
α0 1−(α1 +β)
-0.0001
Table 2 shows the coefficients over different periods in the absence of the monetary announcements using the KES/USD exchange rate returns. The KES/USD exchange rate returns exhibit volatility clustering as well as heteroscedasticity for all periods. This is implied by the significant ARCH and GARCH effects all through the periods providing justification for the use of Conditional heteroscedasticity models. For some periods such as 2010-2011 and 2011-2012, the sum of the GARCH and ARCH coefficients is 0.99 i.e. close to unity which according to (Brooks, 2008) is an indicator of the persistency of shocks to the conditional variance. The long run variance is little to nothing for all periods. This shows general lack of persistency of shocks to the exchange rate volatility in the long run. The long run variance is highest for the
11
period 2011-2012 where the CBK intervened to quell excess volatility occasioned by exchange rate speculation by commercial banks. Table 3: Coefficients and p-values showing the ARCH and GARCH effects on the KES/UGX exchange rate Period
Coefficients θ α1
Coefficients 𝛼0 𝛼1 𝛽
2008-2009 2009-2010 2010-2011 2011-2012 Mean Equation: 𝑅𝑡 = 𝜃0 + 𝛼1 𝑅𝑡−1 +𝜀𝑡 0.0009 (0.1036)
-0.0003 (0.4890)
-0.0166 -0.06420 (0.8499) (0.4025) Variance Equation ℎ𝑡 0.0006* (0.0099) 0.2684* (0.0000) 0.7366* (0.0000)
0.0000* (0.0009) 0.1346* (0.0031) 0.7601* (0.0000)
2012-2013 2013-2014
-0.0003 (0.4890)
-0.0002 (0.7176)
0.0001 (0.7470)
0.0004 (0.1445)
-0.0642 (0.4025)
-0.0135 (0.7655)
0.0050 (0.9447)
0.1551* (0.0187)
0.0000* (0.0007) 0.4677* (0.0000) 0.4026* (0.0000)
0.0000* (0.0000) 0.4519* (0.0002) -0.0542 (0.6577)
2 = 𝛼0 + 𝛼1 𝜀𝑡−1 + 𝛽ℎ𝑡−1
0.0000* (0.0009) 0.1346* (0.0031) 0.7601* (0.0000)
Long run variance : 𝑣𝑎𝑟(𝑢𝑡 ) =
0.0000* (0.0000) 1.0700* (0.0000) 0.1027 (0.0829)
𝛼0 1−(𝛼1 +𝛽)
-0.1200 0.0000 0.0001 0.0000 Long run Variance *indicates that the coefficient is significant at the 5% confidence level.
0.0002
0.0000
Using the KES/UGX log returns, the ARCH effects are highly significant all through the periods while the GARCH effects are insignificant for the period 2011-2012 and 2013-2014. The long run variance is also zero for most periods as was the case for the KES/USD which is evidence of lack of persistency of shocks on exchange rate volatility. 4.2.2 With Monetary policy announcements The model used is an extension of the GARCH (1,1) which incorporates the first (June) and second (December) announcement using dummy variables. The study records 0 for the period before the announcement and 1 upon occurrence of the Monetary Policy Announcement and for the period thereafter. 𝑅𝑡 = 𝜃0 + 𝛼1 𝑅𝑡−1 + ∑𝑛𝑖=1 𝛾𝑖 𝐷𝑖 + ∑𝑛𝑗 ∅𝑗 𝐷𝑗 +𝜀𝑡 2 ℎ𝑡 = 𝛼0 + 𝛼1 𝜀𝑡−1 + 𝛽ℎ𝑡−1 + ∑𝑛𝑖=1 𝛾𝑖 𝐷𝑖 + ∑𝑛𝑗 ∅𝑗 𝐷𝑗
(12) (13)
Equations (12) and (13) are inclusive of both the first and the second monetary policy announcement. Using KES/USD log returns, Table 4 shows the significance of the GARCH and 12
ARCH effects and the first and second announcement on the level and volatility of exchange rate returns. Table 4: Coefficients and p-values showing effects of monetary policy announcements on the KES/USD exchange rate. Period
2008-2009
2009-2010
2010-2011
Mean Equation: 𝑅𝑡 = 𝜃0 + 𝛼1 𝑅𝑡−1 +
2011-2012
∑𝑛𝑖=1 𝛾𝑖
𝐷𝑖 +
2012-2013
∑𝑛𝑗 ∅𝑗 𝐷𝑗
2013-2014
+𝜀𝑡
Coefficients 𝜃0
0.0002 (0.7984) 0.1086 (0.0903)
-0.0003 (0.7101) 0.0523 (0.5080)
0.0004 (0.3735) 0.1244 (0.1264)
0.0010 (0.0978) 0.1960* (0.0042)
0.0002* (0.0201) -0.0517 (0.2910)
0.0002 (0.7791) 0.1731* (0.0407)
0.0018 (0.1145)
-0.0000 (0.9569)
-0.0007 (0.1546)
0.0008 (0.3953)
0.0001 (0.4434)
-0.0001 (0.8968)
-0.0018 0.0007 0.0008 -0.0021* -0.0000 (0.0650) (0.0644) (0.0695) (0.0317) (0.9278) 2 Variance Equation: ℎ𝑡 = 𝛼0 + 𝛼1 𝜀𝑡−1 + 𝛽ℎ𝑡−1 + ∑𝑛𝑖=1 𝛾𝑖 𝐷𝑖 + ∑𝑛𝑗 ∅𝑗 𝐷𝑗
-0.0001 (0.8321)
𝛼1 𝛾 ∅
Coefficients 𝛼0 𝛼1 𝛽 𝛾 ∅
0.0000 (0.4056) 0.4329* (0.0000) 0.4433* (0.0001) 0.0000* (0.0013)
0.0000 (0.0659) 0.1450* (0.0242) 0.6510* (0.0000) -0.0000 (0.1230)
0.0000 (0.3014) 0.4233* (0.0006) 0.5385* (0.0000) 0.0000 (0.0983)
0.0000* (0.0003) 0.7648* (0.0000) -0.0372 (0.4847) 0.0000* (0.0004)
0.0000 (0.3705) 1.5953* (0.0000) 0.2574* (0.0000) 0.0000* (0.0263)
0.0000* (0.0042) 0.2962* (0.0002) 0.6963* (0.0000) -0.0000 (0.0762)
-0.0000* (0.0019)
-0.0000 (0.2329)
0.0000* (0.0127)
-0.0000* (0.0018)
0.0000 (0.8629)
0.0000 (0.9436)
0.0000
0.0000
Long run variance: 𝑣𝑎𝑟(𝑢𝑡 ) = Long run Variance
0.0000
0.0000
𝛼0 1−(𝛼1 +𝛽)
0.0000 0.0000
*indicates that the coefficient is significant at the 5% confidence level. In Table 4 above, the first and second announcements have an insignificant effect on the level of the exchange rate for all periods apart from 2011-2012 where the second announcement is significant at the 5% confidence level. In the variance equation, the ARCH effects are highly significant for all periods. The GARCH effect is highly significant for all periods with the exception of the period 2011-2012. Both announcements significantly affect the volatility of the exchange rate in the periods 2008-2009 and 2011-2012. Upon incorporating the monetary policy announcements, the long run variance is at zero for all periods. According to (Almeida, Goodhart, & Payne, 1998), lack of persistency indicates that macroeconomic news has no substantial effect on exchange rates. 13
The study then examined the KES/UGX exchange rate returns. The model shown below is not only inclusive of the ARCH and GARCH coefficients and the monetary policy announcement occurrence dummies but also the KES/USD exchange rate returns as a regressor to account for external shocks arising from outside the East African Community. 𝑅𝑡 = 𝜃0 + 𝛼1 𝑅𝑡−1 + ∑𝑛𝑖=1 𝛾𝑖 𝐷𝑖 + ∑𝑛𝑗 ∅𝑗 𝐷𝑗 +𝛿1 𝑅𝑡𝑠𝑢𝑑 + 𝜀𝑡 2 ℎ𝑡 = 𝛼0 + 𝛼1 𝜀𝑡−1 + 𝛽ℎ𝑡−1 + ∑𝑛𝑖=1 𝛾𝑖 𝐷𝑖 + ∑𝑛𝑗 ∅𝑗 𝐷𝑗 + 𝛿1 𝑅𝑡𝑠𝑢𝑑
(14) (15)
Table 5: Coefficients and p-values showing the effects of monetary policy announcements on the KES/UGX exchange rate Period
2008-2009
2009-2010
2010-2011
2011-2012
2012-2013
2013-2014
Mean Equation: 𝑅𝑡 = 𝜃0 + 𝛼1 𝑅𝑡−1 + ∑𝑛𝑖=1 𝛾𝑖 𝐷𝑖 + ∑𝑛𝑗 ∅𝑗 𝐷𝑗 +𝛿1 𝑅𝑡𝑠𝑢𝑑 + 𝜀𝑡 Coefficients 𝜃0
0.0010 (0.1257) 0.0296 (0.7383)
-0.0017 (0.1815) 0.0087 (0.8994)
-0.0013 (0.1791) -0.1120 (0.1634)
0.0003 (0.4935) 0.0171 (0.7821)
0.0001 (0.9045) 0.1701* (0.0409)
0.0005 (0.7248) -0.0323 (0.6776)
𝛾
-0.0012 (0.1918)
0.0016 (0.2355)
0.0011 (0.2789)
-0.0015* (0.0317)
-0.0004 (0.6728)
-0.0003 (0.8558)
∅
-0.0001 (0.8994)
-0.0011 (0.0825)
0.0000 (0.9981)
0.0024* (0.0097)
0.0009 (0.1971)
-0.0001 (0.8568)
𝛿1
0.9491* (0.0000)
0.7530* (0.0000)
0.3122* (0.0000)
0.7596* (0.0000)
0.7558* (0.0000)
0.5279* (0.0000)
𝛼1
2 Variance Equation: ℎ𝑡 = 𝛼0 + 𝛼1 𝜀𝑡−1 + 𝛽ℎ𝑡−1 + ∑𝑛𝑖=1 𝛾𝑖 𝐷𝑖 + ∑𝑛𝑗 ∅𝑗 𝐷𝑗 + 𝛿1 𝑅𝑡𝑠𝑢𝑑
Coefficients 𝛼0
0.0000 (0.1016) 0.3368* (0.0000)
0.0000* (0.0022) 0.0913* (0.0000)
0.0000 (0.0864) 0.1143 (0.1075)
0.0000* (0.0307) 0.9686* (0.0000)
0.0000* (0.0000) 0.2543* (0.0025)
0.0000* (0.0003) 0.2424* (0.0034)
𝛽
0.7074* (0.0000)
0.8945* (0.0000)
0.6643* (0.0000)
0.0554 (0.2176)
0.4315* (0.0000)
-0.1146 (0.4725)
𝛾
0.0000 (0.4928)
-0.0000* (0.0012)
-0.0000 (0.1286)
0.0000* (0.0000)
-0.0000 (0.2058)
-0.0000* (0.0034)
∅
0.0000 (0.6264) -0.0002 (0.1343)
0.0000* (0.0001) 0.0006* (0.0006)
0.0000* (0.0454) 0.0002 (0.3099)
0.0000 (0.4560) 0.0005 (0.1548)
0.0000 (0.0603) -0.0008* (0.0000)
-0.0000* (0.0144) 0.0001 (0.8221)
0.0000
0.0000
𝛼1
𝛿1
𝛼
0 Long run variance: 𝑣𝑎𝑟(𝑢𝑡 ) = 1−(𝛼1 +𝛽)
Long Variance
run
0.0000
0.0000
0.0000
0.0000
*indicates that the coefficient is significant at the 5% confidence level.
14
Table 5 shows that the announcements do not influence the level of the KES/UGX exchange rate substantially for all periods apart from 2011-2012 where both announcements are significant on the level. This can be attributed to the CBK explicit intervention in 2011 prompted by the foreign exchange speculation that led to the depreciation of the Kenya shilling. Additionally, the KES/USD exchange rate returns are highly significant on the level of the KES/UGX exchange rate for all periods. This alludes to the dominance of the US dollar in the international global market. In the variance equation, the ARCH effect is significant all through apart from the period 20102011. The GARCH effects are also significant with the exception of 2011-2012 and 2013-2014. The first and second announcements are both significant for the period 2009-2010 and 2013-2014. Only the first announcement is significant for the period 2011-2012. Surprisingly, KES/USD exchange rate affects the level and not the volatility of the KES/UGX exchange rate. The long run variance is at zero for all periods indicating the lack of persistency of shocks on the KES/UGX exchange rate volatility.
5
Discussion of Results
In the absence of monetary policy announcements, the ARCH and GARCH effects significantly explain the exchange rate volatility for all periods and for both exchange rates. This is consistent with (Bollerslev & Andersen, 1998) who found that the ARCH effects were among the main causes of volatility in the German-dollar exchange rate. Similarly, (Kamau, Maana, Ngugi, & Tiriongo, 2013) found that the ARCH effects were significant for the KES/UGX, KES/USD, KES/GBP, KES/EUR and KES/TZS exchange rate returns. The significance of the ARCH and GARCH effects is evidence of volatility clustering which justifies the use of an ARCH specification as a proxy for the structure of conditional variance. The same is able to capture the times series properties of daily returns. In the presence of monetary policy announcements, the ARCH effects remain significant for all periods for the KES/USD exchange rate returns. The study found that monetary policy announcements in Kenya have an insignificant relationship with the level of exchange rates. They however have a significant albeit period-specific effect on the volatility of exchange rates. This is consistent with the findings by (Jansen & Haan, 2002) and (Laakkonen, 2007) who found that the statements made by the ECB officials had a substantial though short-lived influence on volatility but not the level of the euro-dollar exchange rates. Similarly, (Chipili, 2013) argues that announcements regarding Central Bank foreign exchange intervention are intended to quell excess volatility as opposed to adjusting the level of the exchange rate to a desired value. (Kamau, Maana, Ngugi, & Tiriongo, 2013) assert that the CBK does not participate in the foreign exchange market to defend a particular value of the Kenya shilling but may intervene to stabilize excess volatility as was the case in the period 2011-2012. The study found that monetary policy announcements significantly affected exchange rate volatility only where the announcement had a direct link to the exchange rate e.g. exchange rate intervention efforts in 2011-2012. This is consistent with (Hutchison & Fatum, 2002) who found that announcements or statements on intervention had a significant effect on the Euro/dollar exchange rate volatility. From the same, the study deduced that the monetary policy announcements are not the main determinants of exchange rate volatility in Kenya consistent with (Bollerslev & Andersen, 1998) who found that the effect of macroeconomic announcements on the volatility of exchange rates was merely secondary. Further, the lack of persistency of shocks 15
on exchange rate volatility observed by the study is aligned with the findings of (Almeida, Goodhart, & Payne, 1998) where monetary policy announcements had a shortlived effect on the German/dollar exchange rate. The authors concluded that this observation signalled that macroeconomic announcements did not have a substanial effect on the volatility of exchange rates. In a similat strudy (Cheung & Chinn, 2001) found that the significance of monetary policy announcements was pegged on the content of the same i.e. the variables included in a given announcement. He argues that some variables carry more weight than others as regards the foreign exchange rate market over different time periods. In addition, the insignificance of monetary policy announcements on the volatility of exchange rates could be attributed to the market microstructure. According to the National Bureau of Economic Research (NBER), market microstructure involves the role of information in the price discovery process, control of liquidity, transaction costs and their implication for efficiency and regulation trading mechanisms and market structures. The same can inhibit reponse of the markets to monetary policy announcements. This is consistent with (Andersen, Bollerslev, Diebold, & Vega, 2003) who argued that slow or minimal response to announcements was as a result of inadequate liquidity.
6
Conclusion
The study found that monetary policy announcements have an insignificant effect on the level of exchange rates while they have a significant effect on the volatility of the same. This is aligned with the notion of a floating exchange rate regime where the Central Bank ought not to defend a particular value of the domestic currency against a foreign currency but instead only intervene to quell excess volatility as asserted by (Chipili, 2013) and (Kamau, Maana, Ngugi, & Tiriongo, 2013). In addition, the study found that the significance of monetary policy announcements on volatility of exchange rates in Kenya is pegged on the content of the announcement which must directly involve the exchange rate. The same is evident in 2011-2012, the only period where announcements, which included the CBK intervention on the prevailing exchange rate fluctuations, had a significant effect on both KES/USD and KES/UGX exchange rate. This is consistent with the findings of (Hakkio & Pearce, 1985) and (Cheung & Chinn, 2001). The general insignificance of monetary policy announcements on the volatility of exchange rates builds the case for the secondary nature of macroeconomic announcements as a determinant of exchange rate volatility as shown by (Bollerslev & Andersen, 1998). The exchange rate level and volatility in Kenya is subject to changes in the current account deficit, shocks from international markets, Diaspora remittances and tourism and export sector earnings as reported in the Monetary policy statements. Additionally, the one-off spikes in volatility e.g. in early 2007 are informed by events such as increased FDI cash flows due to cross country mergers and sale of equity to foreign investors. The actual occurrence of monetary policy announcements is therefore merely secondary to these more prominent and exogenous exchange rate volatility determinants.
16
References Almeida, A., Goodhart, C. A., & Payne, R. (1998). The Effects of Macroeconomic News on High Frequency Exchange Rate Behaviour. Journal of Financial and Quantitative Analysis, 33, 238-408. Andersen, T. G., Bollerslev, T., Diebold, F. X., & Vega, C. (2003). Micro effects of Macro Announcements Real Time discovery in Foreign Exchange. The American Economic Review, 93(1), 39-62. Bernanke, B., & Kuttner, K. (2004). What Explains the Stock Market's Reaction to Federal Reserve Policy. Federal Reserve System. Bollerslev, T., & Andersen, T. G. (1998). Answering Skeptics: Yes, Standard Volatility Models do Provide Accurate Forecasts. International Economic Review, 885-905. Bomfim, A. N. (2001). Pre-Announcement effects, News effects and Volatility: Monetary policy and the Stock Market. Journal of Banking and Finance. Brooks, C. (2008). Introductory Econometrics for Finance. Edinburgh: Cambridge University Press. Cheung, Y. -W., & Chinn, D. M. (2001). Currency traders and exchange rate dynamics: a survey of the US market. Journal of International Money and Finance, 20, 439-471. Chipili, J. (2013). Monetary Policy,Foreign Exchange Intervention and Exchange Rate Volatility in Zambia. The African Finance Journal, 115, 36-55. Ederington, L. H., & Lee, J. H. (1993). How Markets Process Information: News Releases and Volatility. Journal of Finance, 48, 1161-1191. Evans, M., & Lyons, R. K. (2003). How is Macro news transmitted to exchange rates? NBER Working Paper No 9433, Cambridge. Frenkel, J. A. (1981). Flexible Exchange Rates, Prices and the Role of News. Lessons from the 1970s. Journal of Political Economy, 89(4), 665-705. Galati, G., & Ho, C. (2003). Macroeconomic News and the Euro/Dollar Exchange Rate. Economic Notes, 32(3), 371 - 398.
17
Goodhart, C., Henry, S. G., Hall, S. G., & Pesaran, B. (1993). News Effects in a High Frequency Model of the Sterling-Dollar Exchange Rate. Journal of Applied Econometrics, 8(1), 1-13. Hakkio, C. S., & Pearce, D. K. (1985). The Reaction of exchange rates to economic news. Economic Inquiry, 23, 621-635. Hutchison, M. M., & Fatum, R. (2002). ECB Foreign Intervention and the Euro: Institutional Framework, News and Intervention. Institute of Economics. Economic Policy Research Unit. Jansen, D. J., & Haan, J. (2002). Statements of ECB Officials and their Effect on the Level and Volatility of the Euro-Dollar Exchange Rate. CESifo Conference on Macro, Money & International Finance,. Kamau, Maana, Ngugi, & Tiriongo. (2013). Understanding the exchange rate returns volatility in Kenya. Kinyua, J. (2000). Monetary Policy in Kenya: Evolution and Current Framework. Nairobi: Central Bank of Kenya. Kiptoo, C. (2007). Real Exchange Rate Volatility and misalignment in Kenya: 1993-2003; Assessment of its impact on International Trade and investments. University of Nairobi. Laakkonen, H. (2007). The Impact of Macroeconomic News on Exchange rate Volatility. Finish Economic Papers. Mishkin, F. (1995). Symposium on the Monetary Transmission Mechanism. Journal of Economic Perspectives, 9, 3-10. Musyoki, D., Pokhariyal, G. P., & Pundo, M. (2012). Real Exchange Rate Volatility in Kenya. Journal of Emerging Trends in Economics and Management Sciences, 3(2), 117-122. Ndung'u, N. S. (1999). Monetary and Exchange Rate policy in Kenya. Research Paper, Africa Economic Research Consortium. Nelson, D. (1990). Conditional Heteroscedasticity of asset returns: A new approach. Econometrica, 59, 347-370.
18
Omrane, W. B., Bauwens, L., & Giot, P. (2005). News Announcements, Market Activity and Volatility in the Euro-Dollar foreign exchange Market,. Journal of International Money and Finance, 24, 1108-1125. Patell, J. M., & Wolfson, M. A. (1984). The intraday speed of adjustment of stock prices to earnings and dividend announcements. Journal of Financial Economics, 13(2), 223- 252. Rotich, H., Maana, I., & Kathanje, M. (2008). A Monetary Policy Reaction Function for Kenya. Annual African Econometric Society Conference. Pretoria. Taylor, J. B. (1995). The Monetary Transmission Mechanism:An Empirical Framework. Journal of Empirical Perspectives, 9(4), 11-26. Tivegna, M. (2001). News and Dollar Exchange Rate Dynamics,. Universiy of Rome, Rome. Utsunomiya, T. (2013). A new approach to the effect of intervention frequency on the foreign exchange market:evidence from Japan. Applied Economics, 45, 3742-3759.
7
Appendices
7.1
Appendix A: Conditional volatility graphs 2008-2009
2009-2010
19
2010-2011
2011-2012
2012-2013
20
2013-2014
21
