1
Is the Taylor rule useful as a benchmark to the euro
zone during the sovereign debt crisis?
Name: Stefan van den Berg Student number: 10367004
2 Abstract
This thesis investigates the usefulness of the simple Taylor and the forward-looking Taylor rule adjusted for interest rate smoothing, a benchmark model which predicts the optimal short-term interest rate for the euro zone. The main aim of this paper is to find out which model is better at following the short-term interest rate set by the ECB from October 2008 up to the end of 2014. Results suggest that the Taylor rule is a good benchmark model for the monetary policy setting in the euro zone during the sovereign debt crisis, however a forward-looking Taylor rule adjusted for interest rate smoothing is better as a benchmark to the short-term interest rate setting for the euro zone during financial stable as well as instable times.
Statement of Originality
This document is written by Student Stefan van den Berg who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3
Contents
Abstract 2
I. Introduction 4
II. Literature review 5
III. Methodology and Data 9
Data 10 Model 12 Method 13 IV. Results 14 VI. Conclusion 20 References 23 Appendix 25
4
I. Introduction
Stage three of the Economic and Monetary Union (EMU) started on the 1th of January 1999. This stage contained the introduction of a single currency for all European Union (EU) countries, the Euro. Since then there is only one short-term interest rate for all central banks in the EU. Since 1999 the interest rate is set by the ECB and it is adopted by al central banks in the EMU. The primary objective of the ECB as an independent central bank is to maintain price stability. According to the ECB price stability contributes to achieving high levels of economic activity and employment (www.ecb.europa.eu). This can be done through monetary policy.
However the recent sovereign debt crisis highlighted the weaknesses of the EMU, causing an enormous wave of uncertainty and economic distress. For the last decade, because of close-to-zero lending rates, many countries lent money in order to stimulate economic growth. Besides lending a government could raise money by selling government bonds. These bonds (the government lending money from investors) were seen as save investments, such that interest rates on these were low. In 2009 there was uncertainty about Greece paying back its loans. Even in 2014 Greece still had the biggest debt presented in % GDP, with 177.1% this is more than 80% points above the 91.9% EU-19’s average (www.ec.europa.eu/eurostat). Interest rates rose with almost 8% on their 10-Y government bonds. This showed the weak side of the Euro. The ECB had to intervene in order to keep the Euro system functional (Lane, P.R., 2012).
A key discussion in monetary economics is the evaluation of a central bank policy, specifically the interest rate setting of the European Central Bank. The ECB takes many variables into account when setting interest rates, such that it is difficult to understand why interest rates may have changed. Therefore a simplified benchmark model would be useful in times of financial distress. Taylor determined the first policy rule model in 1993. Using graphical evidence, he showed that the inflation and output gap tend to drive the federal funds rate set by the Fed from 1987 to 1992 (Taylor, J.B., 1993). In Europe this model also has extensively been applied. There have been written many papers about the usefulness of the Taylor rule. According to Sauer and Sturm (2007) the Taylor rule has been a reliable model for predicting interest rates in Europe. However there has not been much research on whether the Taylor rule could be useful during a period of financial distress. Therefore this
5
paper will contribute to the existing literature, examining the usefulness of the Taylor rule in a turbulent period. In more detail, this paper will address the issue whether the Taylor rule is a good benchmark for setting short term interest rates from October 2008 up to the end of 2014 (sovereign debt crisis). This period will be compared with a period of financial stability, namely from 1999 up to 2007.
In order to keep financial stability, central banks will deviate from their normal policy in times of economic distress. Therefore the hypothesis is that the Taylor rule is not useful as a short term interest rate benchmark during the sovereign debt crisis. Using besides a simple Taylor rule, a Taylor rule with forward-looking expectations which is also adjusted for interest rate smoothing, this paper will examine short term interest rates in the euro zone. By comparing actual interest rates with predicted interest rates of these models, conclusions can be given on which Taylor rule is an accurate predictor for both periods. The next section will introduce the Taylor rule and provide relevant literature. Section 3 will provide data and the methodology of this research. Where section 4 will present the output gained out of the regression. At last in Section 5 concluding remarks will be provided.
II. Literature review
Since the Maastricht Treaty (1992) the ECB is seen as an independent central bank. According to Sauer and Sturm (2007) a highly independent central bank is believed to keep inflation low and therefore to be better at targeting price stability. The goals for monetary policy in the EMU are explained in this Treaty. According to article 105,
‘’The primary objective of the European System of Central Banks (ESCB) shall be to maintain price stability. Without prejudice to the objective of price stability, the ESCB shall support the general economic policies in the Community with a view to contributing to the achievement of the objectives of the Community as laid down in Article 2…’’
6
‘’To promote throughout the Community a harmonious and balanced development of economic activities, sustainable and non-inflationary growth respecting the environment, a high degree of convergence of economic performance, a high level of employment and of social protection, the raising of the standard of living and quality of life, and economic and social cohesion and solidarity among Member States.’’
(Treaty on European Union, 1992)
In countries or areas with a more dependent bank, other considerations could have effect on the objective of price stability. For political reasons a dependent bank could for example decide to target unemployment instead of solely focusing on inflation. The Governing Council of the ECB states that price stability is measured by the harmonized index of consumer prices (HICP index), which measures the year-to-year increase in price. The primary objective is to maintain price stability; a positive inflation target close, but below 2%. In order to reach these objectives, the operational framework of the Euro system consists of three instruments. The first is open-market operations, such as refinancing operations which tend to drive interest rates in the EMU. The second is standing facilities, which aim to provide and absorb overnight liquidity and provides a ceiling for the overnight interest rate. The third is the minimum reserve requirements, which intent to pursue the aims of stabilising money market interest rates.
There are many papers available about the usefulness of the Taylor rule on monetary policy in the EMU. Therefore the current section relates to several strands of the literature. Besides these papers the original Taylor rule will be introduced.
Taylor (1993) shows that the inflation gap and output gap tend to follow the observed path of the US federal funds rate between the late 1980s and early 1990s. The Taylor rule recommends an optimal interest rate for a central bank, depending on four variables. The first variable is the equilibrium real interest rate. Second variable is the current inflation. The third is the inflation gap, which is the difference between the current inflation and the target inflation rate. When the current inflation is above target inflation, this rule recommends the central bank to raise interest rates. The fourth and last is the output gap, this is the difference between the actual output and the potential output. This factor tends to raise interest rates when the gap is positive. Taylor (1993) proposed the following equation (1) as what is now known as the Taylor rule.
7
– – (1)
Where is nominal fed funds rate, is the equilibrium real interest rate, is rate of inflation at time t, is the target inflation rate, is logarithm of real output and is the logarithm of potential output.
During the last decade, the Taylor rule (1993) has become a popular tool for evaluating the monetary policy of central banks. As benchmark for predicting interest rates in Europe, the Taylor rule is used by investors, traders, and numerous other ‘policy’ watchers. Besides numerous papers on reaction functions on the Fed, the usefulness of the Taylor rule (Judd and Rudebusch, 1998) and the Taylor rule applied on Europe before the introduction of the Euro (Gerlach and Schnabel, 2000), some papers introduced the Taylor rule with forward-looking expectations (Bernanke and Gertler, 2000; Clarida et al. 1998).
The latter Clarida et al. (1998) introduced such Taylor rule with forward-looking expectations. They estimated reaction functions on the G3 countries (Germany, Japan and the US) and on the European E3 countries (Italy, France and the UK). Since 1979 major central banks (Bundesbank, Fed) adopted to control inflation, therefore their papers main focus is on what type of monetary policy central banks used since 1979. Their findings support the claim that G3 countries tend to pursue some sort of inflation targeting. Clarida et al. (1998) also find evidence on these G3 countries anticipating at expected inflation opposed to anticipating at lagged inflation. The E3’s commitment to the European Monetary System (EMS) made their regressions more difficult. Their research on the E3 countries shows less supported results than their research on the G3 countries did. While using the Bundesbank policy as a benchmark, they find sufficient evidence on the E3 countries setting interest rates being at a higher level than macroeconomic conditions warranted. Clarida et al. (1998) concluded from the latter that inflation targeting could be superior to fixing exchange rates.
This forward-looking Taylor rule is supported by Gert Peersman and Frank Smets (1999). By introducing a closed economy model based on a Rudebusch and Svensson (1999) study, their paper examines whether the Taylor rule is useful as a benchmark for the euro zone. With a regression on quarterly data on period 1975-1997, their results show that the Taylor rule model stabilizes output and inflation for the euro zone when macroeconomic
8
shocks appear. Such that they conclude that the forward-looking Taylor rule which also accounts for interest rate smoothing predicts short-term interest rates well. According to Gert Peersman and Frank Smets (1999) their model robustness shows that the model is robust against structural economic changes. If it behaves well under one policy, it could be useless after changes in policy or economy. The latter is particularly of interest for this paper, as the period before and during the sovereign debt crisis is examined. An estimation error of the output gap does not tend to affect the Taylor rule’s reliability significantly, therefore it is concluded that the model is robust against small changes in the parameters of the model. These favourable results are in compliance with the results gathered from the regression which they also applied on the US. However the estimations of the equilibrium real interest rate are uncertain. There are more drawbacks to this rule model. The rule is restrictive for the fact there are limited variables in the model. Besides inflation and output there are several variables which have impact on the overall performance of the economy. Variables such as unemployment, exchange rates and asset prices have to be taken into consideration, while they affect consumption and therefore welfare. Second, instrument rules may not be robust to changes in the structure of the economy as noted by Peersman and Smets (1999). An independent central bank such as the ECB would never want to commit to such a restrictive rule.
The Taylor rule can however still be a useful tool for evaluating the policy setting of a central bank. The rule is for example used as a communication device to explain the interest rate setting to the general public (Peersman and Smets, 1999). In addition Fernandez and Rzhevskyy (2007) still find the Taylor rule relevant for today’s estimations of optimal interest rates.
More recent research such as that of Sauer and Sturm (2007), also evaluated the monetary system of the ECB. They wondered whether the ECB has been following a stabilizing or destabilizing policy under presidency of Wim Duisenberg. In this particular period (1998-2003), their results show that the ECB was allowing changes in inflation and hence follows a destabilizing policy. However this result seemed to be largely due to non-forward-looking expectations in their model. When using forward- looking expectation measures of inflation, results indeed showed the appliance of a stabilizing policy of the ECB.
One thing has to be mentioned, these studies were mostly conducted in a period without the ECB setting monetary policy. However Belke and Polleit (2007), took a sample
9
from 1999-2005 for the euro zone and a sample from 1987-2005 for the Fed. They found that the Taylor rule is better in predicting interest rates for the Fed than for the ECB. The Fed set interest rates in line with Taylor rule’s predictions, while the ECB deviated from predictions during the pre-crisis period. Their results also show that the ECB puts a larger weight on the output gap relative to the inflation gap. However this may be due to the low-inflation rates in this particular period. These findings have to be considered in this paper, however their results could be biased because of the small sample they used on the euro zone, therefore this paper will also contribute to this research, explaining if the ECB deviated from the Taylor rule’s predictions.
Literature above agrees on some views. The Taylor rule is applicable on Europe when smoothing interest rates and following a forward-looking model. The ECB was focussing on inflation and output gap since 1999 and stabilizes them. However it is still unclear whether the Taylor rule is still applicable and whether the ECB still focuses on output and inflation. In de next section data and methodology will be discussed.
III. Methodology and Data
The main objective of this paper is to investigate whether the Taylor rule is applicable in times of financial distress, more specifically during the sovereign debt crisis. Following from the literature review, it can be concluded that a forward-looking Taylor rule adjusted for interest rate smoothing provides better estimates than the simple Taylor rule. However literature about the performance of the Taylor rule during the sovereign debt crisis is still in its infancy. Therefore the performance of the extended Taylor rule will be compared with that of the simple Taylor rule during the sovereign debt crisis. To compare results, both models are also tested on a period of relative financial stability. The first period that will be looked at is from January 1999 to October 2008, the relatively stable era during which the ECB was responsible for the conduct of monetary policy in the euro area. In the third quarter the EONIA dropped with more than 1%, therefore the third quarter of 2008 will be considered as the start of the sovereign debt crisis (see table 1). The second period that will be looked at is from October 2008 up to January 2015, during this period of financial distress the ECB was responsible for monetary policy in the euro area. The comparison between
10
these two periods is needed in order to give concluding remarks about the performance of the Taylor rule. In addition various real equilibrium interest rates will be used to see how the forward-looking Taylor rule adjusted for interest rate smoothing behaves under different assumed equilibrium rates. In both periods these three real equilibrium interest values will be evaluated used in order to see which fits best. The interest rate smoothing parameter, which is also unknown, will be estimated by OLS. At last a Chow test will indicate whether the coefficients of both models changed significant over time. Although interesting, the comparison of the Taylor rule’s performance in the euro area with its performance in the US will not be discussed, this is left for later study. This section will now further discuss which models will be used and the data needed for regression.
Data
For estimating the weights of all factors of the Taylor rule, data is needed of various variables for two periods, namely from 1999 up to 2008Q3 and from 2008Q4 up to the end of 2014. Quarterly data will be used, while the data of some variables (GDP) is only released quarterly. Therefore the first period will provide data for a 9 ¾ year time period, which conducts 39 observations. The second period will provide data for a smaller time span, meaning 25 observations. With these amounts of observations per time span, results out of regression have to be interpreted with caution. This section will now further discuss which data is used and where it can be found.
For examining the performance of the two rules, actual nominal short-term interest rates for the euro zone are needed as a dependent variable. The Euro OverNight Index Average (EONIA) is the 1-day interbank interest rate for the Euro zone. In other words, it is the rate at which banks provide loans to each other with a maturity of 1 day. Therefore EONIA can be considered as the 1 day Euribor rate. Graph 1 presents this short-term interest rate for the euro zone from 1999 up to the end of 2014.
11
Graph 1: Nominal short-term interest rate of euro area (quarterly)
This graph shows the nominal short-term interest rate set by the ECB (EONIA).
The first independent variable needed is the inflation rate. Inflation rates of the euro area are published monthly by the ECB. These monthly inflation rates will be converted to quarterly rates in order to fit in the regression. The inflation is measured as the overall inflation in the euro area (HICP). It is the weighted average of price indices of member states who have adopted the euro. The ECB has an inflation target rate of below but close to 2%, therefore a 2% target inflation rate will be used in this paper. Expected inflation rates are also published by the ECB survey of professional forecasts (SPF) quarterly. Actual and expected inflation gaps from 1999 up to 2015 are presented in graph 2.
Graph 2: inflation gap of euro area (quarterly)
This graph shows the inflation gap of the euro area per quarter.
The last two variables are difficult to measure. The equilibrium real interest rate could be interpreted as the average real interest rate of a particular period, however it is not possible to estimate the exact value of this parameter. Therefore three equilibrium real
12
interest rates will be used in order to see which fits best in the regressions. The second variable which is hard to measure is the output gap. This variable consists of actual output minus potential output. Actual output data will be retrieved from the ECB website. Potential output will be estimated by adding the Hodrick-Prescott filter in excel. For the forward-looking Taylor rule expected output gaps are also needed. Therefore quarterly real GDP growth forecasts published by the SPF will be used. The Hodrick-Prescott filter is applied to quarterly data on real GDP, as described by Giorno, Claude, et al (1995). Actual and expected output gaps are presented in graph 3.
Graph 3: Output gap of euro area (quarterly)
This graph shows the output published by the ECB minus the HP filter calculated potential output of the euro area in quarterly data.
Model
Simple Taylor rule:
(1)
Forward-looking Taylor rule adjusted for interest rate smoothing:
(2)
For using equation (1) and (2) in STATA, the equations need to be rewritten as follows. By subtracting and on both sides of equation (1), equation (3) which can be used to calculate the short-term interest rate appears. In equation (2) and need to be
13
subtracted on both sides, then the independent variable will be rewritten, in order to compose equation (4). This results in the following two functions:
Simple Taylor rule:
(3)
Forward-looking Taylor rule adjusted for interest rate smoothing:
–
(4)
Method
Previous mentioned variables will now be regressed against the short-term interest rate of the euro zone without a constant. As well as the simple Taylor rule model, the forward-looking Taylor rule adjusted for interest rate smoothing (extended model) will be examined with different real equilibrium interest rate parameters. From these regressions it will become clear which model follows the short-term interest rate setting by the ECB best. First all needed weights of the two equations (3) and (4) will be calculated, for all variables and for both periods. Then predicted interest rates for the euro zone can be calculated by inserting quarterly data for both periods in both estimated equations (5) and (6). Graphs of the differences in predicted and actual interest rates per quarter will be provided in the results section. With these estimates, both models can be evaluated. If one of the models better follows the actual interest rate setting by the ECB, it will be concluded that the latter is a better model for predicting short-term interest rates for the euro zone. The evaluation of the two models will discuss various parameters. These parameters will be discussed in the following section. At last a Chow-test will provide results from which can be concluded whether the ECB conducted different monetary policies in the two periods looked at.
14
IV. Results
In this section results out of regression will be discussed. The simple and extended Taylor rule will be evaluated for both periods separately. Although different real equilibrium interest rates are used, this section will only discuss estimates with and for period 1, for period 2 and for the whole dataset, while these were the best fit in the regression for their particular period. From graphs 4 and 5, it can be seen that the forward-looking Taylor rule adjusted for interest rate smoothing follows the ECB short-term interest rate better than the simple Taylor rule between 1999 and 2008Q3. Regressions for the period of financial instability, illustrate as well that the extended Taylor rule is a better predictor of the short-term interest for the euro zone. This can also be seen from higher adjusted estimated value for the extended model in both periods. Regression results can be found in tables 1 and 2. The estimated Taylor rules will be presented next.
The estimated simple Taylor rule in period 1 will look like the following equation (5): (5)
The inflation gap coefficient is negative and the coefficient of the output gap is positive. The constant which includes the equilibrium real interest rate is also positive and set to one. A negative inflation gap coefficient is doubtful, why this is not likely will be further discussed in the next section. In the same period the forward-looking Taylor rule adjusted for interest rate smoothing is estimated to look like the following equation (6).
Forward-looking Taylor rule adjusted for interest rate smoothing period 1:
– (6)
This model estimates that the equilibrium interest rate is lower and equal to zero. The inflation gap coefficient is negative, while the output gap coefficient is positive. The coefficients of both gaps are estimated to be smaller, such that the ECB focuses less on these gaps in period 1 following the extended model. Besides that the ECB focuses less on these gaps, interest rate smoothing does have significant positive effect on the interest setting. Both models are shown in graph (4).
15
Graph 4: Short-term interest rate estimates, period 1 (quarterly)
From this graph it can be seen that the extended model is better at following the short-term interest setting by the ECB. The maximum deviation of the extended Taylor rule from actual rates is about 1.5%, where this is about 3.5% for the simple Taylor rule. From table (1) and (2) the same can be concluded, while the extended model had higher R squared values.
For the period from 2008Q3 up to the end of 2014, the estimated simple Taylor rule will look like the following equation (7).
Simple Taylor rule period 2:
(7)
In this period of financial distress, regression on the simple Taylor rule estimates the equilibrium real interest rate to be around 3% lower than in period 1, such that it is negative, therefore a real equilibrium interest rate of -2% is assumed, which is not unthinkable in such a financial unstable period. Besides this negative real interest rate, the inflation gap coefficient is estimated to be about equally negative as in the financial stable period. The output gap coefficient is estimated to be negative and small. For the period of financial distress the forward-looking Taylor rule adjusted for interest rate smoothing is estimated as the following equation (8).
Forward-looking Taylor rule adjusted for interest rate smoothing period 2:
– (8)
16
From this equation it can also be seen that the equilibrium real interest rate coefficient is estimated to be negative. The inflation and output gap coefficients are positive and negative. The inflation gap tends to have more negative impact on the interest setting by the ECB, when estimated by the extended model. Same as the simple model, the extended model estimates a small output gap coefficient. Both models are shown in graph (5).
Graph 5: Short-term interest rate estimates, period 2 (quarterly)
From this graph it can be seen that the extended model is better at following the short-term interest setting by the ECB. The maximum deviation of the extended Taylor rule from actual rates is about 1.5%, where this is about 3% for the simple Taylor rule. From table (1) and (2) the same can be concluded, while the extended model had higher R squared values.
17 Table 1: Regression simple Taylor rule
Variables 1999Q1-2008Q3 2008Q4-2014Q4 1999Q1-2014Q4 Inflation gap -1.2228*** (0.1795) -1.1252*** (0.1761) 0.5367*** (0.1168) Output gap 0.6027*** -0.0786 -0.9956*** (0.0759) (0.1718) (0.1896) Crisis dummy X X -1.2064*** (0.2252) R squared Adj. R squared 0.6502 0.6313 0.7551 0.7338 0.4568 0.4301 Observations 39 25 64
This table shows the regressions of the simple Taylor rule for the periods 1999Q1-2008Q3, 2008Q4-2014Q4 and for the whole period 1999-2014.The ***, ** and * stand for respectively 1%, 5% or 10% significance. The numbers between the brackets are the standard errors of the coefficients.
18
Table 2: Regression forward-looking Taylor rule adjusted for interest rate smoothing
Variables 1999Q1-2008Q3 2008Q4-2014Q4 1999Q1-2014Q4 Inflation gap -0.3659*** (0.1788) -0.6014*** (0.1396) -0.5622*** (0.1265) Output gap 0.1876*** -0.0847* 0.0144 (0.0542) (0.0419) (0.0341) Smoothing parameter Crisis dummy R squared Adj. R squared 0.2399*** (0.0469) X 0.8604 0.8487 0.2829*** (0.0419) X 0.8957 0.8814 0.3961*** (0.0343) -1.5253*** (0.1353) 0.8281 0.8167 Observations 39 25 64
This table shows the regressions of forward-looking Taylor rule adjusted for interest rate smoothing for the periods 1999-2006, 2007-2014 and for the whole period 1999-2014, with rho=0.5. The ***, ** and * stand for respectively 1%, 5% or 10% significance. The numbers between the brackets are the standard errors of the coefficients.
With a Chow test, it can be checked whether the coefficients of the two models have changes significantly over the two periods. When the test suggests that coefficients have changed significantly, it will be concluded that the ECB changed its monetary policy during the sovereign debt crisis. Data is used from regression results of both models. Chow test results are given in table 5. The Chow test statistic has the following form:
(9)
Where is the sum of squared residuals (SSR) of the total period, is the SSR from period 1 and the SSR of period 2. and are the number of observations in each group and is the number of parameters used. The test statistic follows the F distribution with and degrees of freedom.
19 Table 3: Models model r* Simple period 1 = + 1+ -1.2228+ 0.6027+ X Simple period 2 = + -2+ -1.1252+ -0.0786+ X Extended period 1 = + 0+ -0.3689*+ 0.1876*+ 0.2399 Extended period 2 = + -2+ -0.6014*+ -0.0847*+ 0.2829
This table shows the estimated models, the * stands for the expected inflation and output gap coefficients. Where short-term interest rate, actual inflation, real
equilibrium interest rate, the coefficient of the inflation gap, the coefficient of the output gap and the interest rate smoothing parameter.
Table 4: Hypotheses
Hypothesis Simple Taylor rule Extended Taylor rule
H0 H1
20 Table 5: Chow test F-statistic results
Components Simple Taylor rule Extended Taylor rule
63.4718901 19.8095592 9.81040102 9.64054865 15.8827094 5.53723105 2 3 39 39 25 25 60 58 44.11 5.89
This table shows the Chow test statistics. The critical F value for the Simple Taylor rule is , for the extended rule the critical F value is . Both F values indicate that the coefficients are significantly different between periods. Such that H0 can be rejected for both models.
V. Conclusion
The aim of this thesis is to investigate if the Taylor rule is a useful benchmark model for the monetary policy set in the euro zone during the sovereign debt crisis. In times of economic distress, central banks will deviate from their normal policy in order to keep financial stability. Therefore the hypothesis was that the Taylor rule would not be useful as a benchmark for short term interest rates for the euro zone.
Clarida et al. (1998) introduced such Taylor rule with forward-looking expectations. This forward-looking Taylor rule is supported by Peersman and Smets (1999). They conclude that the forward-looking Taylor rule adjusted for interest rate, is better at predicting short-term interest rates for the euro zone than a simple Taylor rule. More recent research from Sauer and Sturm (2007), also showed that the extended Taylor rule is useful as a benchmark model for the euro zone.
From regressions of the simple and extended Taylor rule on the euro area, it can be concluded that the extended Taylor rule follows the short-term interest rate set by the ECB better in a relative stable and instable period. Whereas the extended Taylor rule performs
21
even better during the second period (the sovereign debt crisis). According to both models the inflation gap coefficient is negative during the sovereign debt crisis. A negative coefficient on the inflation gap is not likely to be true. When inflation rises with one percent, the ECB would reduce the short-term interest rate, which is not of use when stabilizing inflation. According to the results from both models, the coefficient on the output gap has decreased, meaning that the ECB focuses less on the output gap during a financial instable period. Both models agree on a negative constant during the sovereign debt crisis (a negative equilibrium real interest rate), this is not common, although not unthinkable in times of economic distress. These findings are in compliance with conclusions out of the literature. However the adjusted R squared of the simple Taylor rule model is below 70 percent for the first period and just above 70 percent for the second. Therefore the simple model does not give such accurate estimates on the short-term interest rate setting of the ECB as the extended model does for both periods. The extended model has a adjusted R squared of almost 90% for both periods, meaning that the variance of the independent variables explain the variance of the interest rate set by the ECB for 90%. To decide whether the Taylor rule is of use as a benchmark model to the euro zone during the sovereign debt crisis, a Chow test was constructed. From this Chow-test it became clear that both models indicate a significant change in monetary policy over the two periods. This indicates in compliance with the results out of regression, that the ECB puts less focus on the output gap during the unstable period. Although one variable does not seem to have a big affect on the interest setting by the ECB, the extended model still comes with more accurate estimates. Therefore findings suggest that the extended Taylor rule is a good benchmark model for the short-term interest rate setting in the euro zone during financial instable periods.
These results have to be interpreted with caution, more research is needed in order to conduct more reliable conclusions. There are some limitations to the results of this paper. For period 1, 39 observations were used and for period 2, 25 observations were used. The use of more data would deliver more accurate and reliable results. Besides that limited data is used, some variables had to be estimated. For example, the equilibrium real interest rate is unknown for the euro area, this had to be set as a constant in all regressions. Potential output is also unknown for the euro zone, with the HP filter, potential output was estimated. The last limitation is about whether the coefficients gained out of the regressions indeed have impact on the monetary policy setting of the ECB. Correlation is not causation.
22
However due to policy statements of the ECB, this is assumed to be true. The research question was if the Taylor rule can be used as a benchmark model for the euro zone during the sovereign debt crisis. This paper concludes that the forward-looking Taylor rule adjusted for interest rate smoothing is good at following the short-term interest rate in the euro zone during the sovereign debt crisis, which is not in accordance with the hypothesis set in section 1. The hypothesis was that a central bank would deviate from its normal policy setting during financial instable times. Evidence suggests that the ECB indeed deviated from its normal policy during a crisis. However, when estimated coefficients are adjusted for that particular period, the extended model still gives accurate estimates.
When a future crisis will appear, these results may not hold. Therefore more research has to be done, on more financial instable periods, in order to compose more accurate conclusions on if the Taylor rule is a good benchmark to the euro zone in times of financial instability.
23
References
Belke, A. and Polleit, T. (2007). How the ECB and the US Fed set interest rates. Applied Economics, 39(17): 2197-2209.
Clarida, R., Gali, J. and Gertler, M. (1998). Monetary policy rules in practice. Some international evidence. European Economic Review, 42(6): 1033-1067.
Peersman, G. and Smets, F. (1999). The Taylor rule: A useful monetary policy benchmark for the euro area?. International Finance, 2(1): 85-116.
Rudebusch, G. and Svensson, L. (1999). Policy Rules for inflation targeting. In J. Taylor, ed., Monetary Policy Rules, Chicago: Chicago Universtity Press.
Taylor, J.B. (1993). Discretion versus policy rules in practice. Carnegie-Rochester Conference Series on Public Policy, 39: 195-214.
Clarida, R., Gali, J. and Gertler, M. (2000). Monetary policy rules and macroeconomic stability: evidence and some theory. The Quarterly Journal of Economics, 115(1): 147-180.
Gerlach, S., and Schnabel, G. (2000). The Taylor rule and interest rates in the EMU area. Economics Letters 67, no. 2 165-171.
Judd, J.P., and Rudebusch, G.D. (1998). Taylor's Rule and the Fed: 1970-1997. Economic Review - Federal Reserve Bank of San Francisco 3-16.
Lane, P.R., and Milesi-Ferretti, G.M. (2012). External adjustment and the global crisis. Journal of International Economics 252-265
24
Bernanke, B., and Gertler, M. (1999). Monetary Policy and Asset Price Volatility. Economic Review, 17
Fernandez, A.Z., and Nikolsko-Rzhevskyy, A. (2007). Measuring the Taylor rule’s performance. Economic Letter
Sauer, S., Sturm, J.E. (2007). Using Taylor Rules to Understand European Central Bank Monetary Policy. German Economic Review, 2007, Vol.8(3), 375-398
Benefits of price stability. (n.d.). Retrieved May 5, 2015, from
https://www.ecb.europa.eu/mopo/intro/benefits/html/index.en.html
General government gross debt - annual data. (2015, March 2). Retrieved May 6, 2015, from
http://ec.europa.eu/eurostat/tgm/table.do?tab=table&init=1&language=en&pcode= teina225&plugin=1
25
Appendix
Graph 4: Short-term interest rate estimates, period 1 (quarterly)
26 Simple Taylor rule:
Period 1 r*=0 r*=1 r*=2 outputgap .6708607 .1556085 4.31 0.000 .3555679 .9861534 inflationgap -.7098839 .3677512 -1.93 0.061 -1.455019 .0352507 itpir0 Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 62.859054 39 1.61177062 Root MSE = 1.0549 Adj R-squared = 0.3096 Residual 41.1725434 37 1.11277144 R-squared = 0.3450 Model 21.6865106 2 10.8432553 Prob > F = 0.0004 F( 2, 37) = 9.74 Source SS df MS Number of obs = 39
outputgap .6027424 .0759579 7.94 0.000 .4488371 .7566476 inflationgap -1.222844 .1795121 -6.81 0.000 -1.58657 -.8591184 itpir1 Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 28.043054 39 .719052667 Root MSE = .51492 Adj R-squared = 0.6313 Residual 9.81040102 37 .265145974 R-squared = 0.6502 Model 18.232653 2 9.11632649 Prob > F = 0.0000 F( 2, 37) = 34.38 Source SS df MS Number of obs = 39
outputgap .5346241 .1644263 3.25 0.002 .2014648 .8677833 inflationgap -1.735805 .3885904 -4.47 0.000 -2.523164 -.948446 itpir2 Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 71.227054 39 1.82633472 Root MSE = 1.1147 Adj R-squared = 0.3197 Residual 45.970958 37 1.24245833 R-squared = 0.3546 Model 25.256096 2 12.628048 Prob > F = 0.0003 F( 2, 37) = 10.16 Source SS df MS Number of obs = 39
27 period 2 r*=-2 r*=-1 r*=0 outputgap -.0785805 .1717605 -0.46 0.652 -.4338941 .2767331 inflationgap -1.125238 .1761444 -6.39 0.000 -1.489621 -.7608557 O Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 64.846632 25 2.59386528 Root MSE = .83099 Adj R-squared = 0.7338 Residual 15.8827094 23 .690552581 R-squared = 0.7551 Model 48.9639226 2 24.4819613 Prob > F = 0.0000 F( 2, 23) = 35.45 Source SS df MS Number of obs = 25
outputgap .0723485 .1615452 0.45 0.658 -.2618332 .4065302 inflationgap -.801472 .1656684 -4.84 0.000 -1.144183 -.4587607 N Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 34.418632 25 1.37674528 Root MSE = .78157 Adj R-squared = 0.5563 Residual 14.0496728 23 .610855341 R-squared = 0.5918 Model 20.3689592 2 10.1844796 Prob > F = 0.0000 F( 2, 23) = 16.67 Source SS df MS Number of obs = 25
outputgap .2232775 .3009954 0.74 0.466 -.399379 .845934 inflationgap -.4777057 .308678 -1.55 0.135 -1.116255 .1608434 itpir0 Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 53.990632 25 2.15962528 Root MSE = 1.4562 Adj R-squared = 0.0180 Residual 48.7750812 23 2.1206557 R-squared = 0.0966 Model 5.2155508 2 2.6077754 Prob > F = 0.3109 F( 2, 23) = 1.23 Source SS df MS Number of obs = 25
28 r*=1
r*=2
whole period with r*=0 + dummy
outputgap .3742065 .472235 0.79 0.436 -.602686 1.351099 inflationgap -.1539394 .4842882 -0.32 0.753 -1.155766 .8478871 itpir1 Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 123.562632 25 4.94250528 Root MSE = 2.2847 Adj R-squared = -0.0561 Residual 120.058934 23 5.21995367 R-squared = 0.0284 Model 3.50369756 2 1.75184878 Prob > F = 0.7183 F( 2, 23) = 0.34 Source SS df MS Number of obs = 25
outputgap .5251356 .6506303 0.81 0.428 -.8207957 1.871067 inflationgap .1698269 .6672368 0.25 0.801 -1.210458 1.550111 itpir2 Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 243.134632 25 9.72538528 Root MSE = 3.1478 Adj R-squared = -0.0189 Residual 227.901233 23 9.90874924 R-squared = 0.0627 Model 15.2333994 2 7.61669972 Prob > F = 0.4752 F( 2, 23) = 0.77 Source SS df MS Number of obs = 25
Crisisdummy -1.206449 .2252417 -5.36 0.000 -1.656847 -.7560502 outputgap .536696 .1168247 4.59 0.000 .3030905 .7703015 inflationgap -.9955929 .189618 -5.25 0.000 -1.374758 -.6164282 itpir0 Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 116.849686 64 1.82577634 Root MSE = 1.0201 Adj R-squared = 0.4301 Residual 63.4718901 61 1.04052279 R-squared = 0.4568 Model 53.3777959 3 17.7925986 Prob > F = 0.0000 F( 3, 61) = 17.10 Source SS df MS Number of obs = 64
29 Extended Taylor rule:
period 1 r*=0 r*=1 Eoutputgap .1876456 .0541961 3.46 0.001 .0777308 .2975604 Einflationgap -.3689334 .1787899 -2.06 0.046 -.7315362 -.0063307 it1r0 .2399385 .0469096 5.11 0.000 .1448014 .3350755 itEpit1r0 Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 69.034254 39 1.77010908 Root MSE = .51749 Adj R-squared = 0.8487 Residual 9.64054865 36 .267793018 R-squared = 0.8604 Model 59.3937053 3 19.7979018 Prob > F = 0.0000 F( 3, 36) = 73.93 Source SS df MS Number of obs = 39
Eoutputgap .1214688 .0772 1.57 0.124 -.0351001 .2780377 Einflationgap -.1969248 .2555017 -0.77 0.446 -.7151063 .3212567 it1r1 .0130968 .0949718 0.14 0.891 -.1795149 .2057085 itEpit1r1 Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 25.018254 39 .641493692 Root MSE = .74513 Adj R-squared = 0.1345 Residual 19.9880761 36 .555224336 R-squared = 0.2011 Model 5.03017792 3 1.67672597 Prob > F = 0.0422 F( 3, 36) = 3.02 Source SS df MS Number of obs = 39
30 r*=2 period 2 r*=0 r*=-1 Eoutputgap -.1616713 .1096398 -1.47 0.149 -.3840312 .0606885 Einflationgap .1270383 .4060727 0.31 0.756 -.6965153 .9505918 it1r2 .0129384 .2202383 0.06 0.953 -.4337255 .4596023 itEpit1r2 Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 59.002254 39 1.51287831 Root MSE = 1.1939 Adj R-squared = 0.0578 Residual 51.3130409 36 1.42536225 R-squared = 0.1303 Model 7.68921305 3 2.56307102 Prob > F = 0.1650 F( 3, 36) = 1.80 Source SS df MS Number of obs = 39
Eoutputgap -.2368171 .1240343 -1.91 0.069 -.4940486 .0204144 Einflationgapt1 .1979567 .3276086 0.60 0.552 -.481462 .8773753 it1r0 -.1669345 .2909965 -0.57 0.572 -.7704242 .4365552 itEpit1r0 Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 46.214432 25 1.84857728 Root MSE = 1.3401 Adj R-squared = 0.0285 Residual 39.5082693 22 1.79583042 R-squared = 0.1451 Model 6.7061627 3 2.23538757 Prob > F = 0.3175 F( 3, 22) = 1.24 Source SS df MS Number of obs = 25
Eoutputgap -.1587622 .0629102 -2.52 0.019 -.2892301 -.0282944 Einflationgapt1 -.4175316 .1992228 -2.10 0.048 -.8306945 -.0043688 AE -.0033339 .0986741 -0.03 0.973 -.2079715 .2013036 AJ Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 24.642432 25 .98569728 Root MSE = .73851 Adj R-squared = 0.4467 Residual 11.9985924 22 .545390565 R-squared = 0.5131 Model 12.6438396 3 4.21461319 Prob > F = 0.0010 F( 3, 22) = 7.73 Source SS df MS Number of obs = 25
31 r*=-2
whole period with r*=0
Eoutputgap -.084671 .0418992 -2.02 0.056 -.1715647 .0022226 Einflationgapt1 -.601433 .1396141 -4.31 0.000 -.8909749 -.3118912 it1r2 .2829828 .0453745 6.24 0.000 .1888819 .3770837 AK Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 53.070432 25 2.12281728 Root MSE = .50169 Adj R-squared = 0.8814 Residual 5.53723105 22 .25169232 R-squared = 0.8957 Model 47.533201 3 15.8444003 Prob > F = 0.0000 F( 3, 22) = 62.95 Source SS df MS Number of obs = 25
dummy -1.525254 .1353007 -11.27 0.000 -1.795896 -1.254613 Eoutputgap .0143595 .034104 0.42 0.675 -.0538587 .0825777 Einflationgap -.5622461 .1264809 -4.45 0.000 -.8152456 -.3092465 it1r0 .3960821 .0343384 11.53 0.000 .3273951 .4647691 itEpit1r0 Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 115.248686 64 1.80076072 Root MSE = .57459 Adj R-squared = 0.8167 Residual 19.8095592 60 .330159319 R-squared = 0.8281 Model 95.4391268 4 23.8597817 Prob > F = 0.0000 F( 4, 60) = 72.27 Source SS df MS Number of obs = 64
