1
Quantitative easing and its effect on the
expected inflation in the United Kingdom
BSc Thesis Economics
University of Amsterdam
Vincent Waterman
Student number: 6066798
Supervisor: Christiaan van der Kwaak
July 2015
2
Statement of Originality
This document is written by Vincent Waterman who declares to take full responsibility for the contents of this document.
I declare to that the text and 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.
Table of Content
Introduction ... 3
Literature review ... 4
How does quantitative easing work? ... 6
Empirical model and data ... 8
Prediction of regression coefficient signs ... 8
Description of the data ... 9
Hypothesis ... 11 Results ... 11 Endogenous variables ... 12 2SLS-regression ... 13 Explanation of results ... 14 Discussion ... 15 Conclusion ... 15 References ... 16 Appendix ... 17
3
Introduction
The worsening of the global financial crisis in 2008 led to the increased risk of a severe downturn that the world had not experienced since the Great Depression (Kapetanios, Mumtaz, Stevens, & Theodoridis, 2012). Before the crisis most central banks aimed for low and stable inflation, which is called inflation targeting. Their main instrument to reach their goal was setting the short term interest rate for the interbank market. The effect and impact of this official bank rate was reliable. This was the main conventional monetary policy in mature economies (Joyce, Miles, Scott, & Vayanos, 2012).
Major central banks started cutting their short term interest rates in October 2008 and in April 2009 the zero lower bound was hit. The search for unconventional measures to further loosen monetary conditions started because the usually reliable correlation between the official bank rate and market interest rates became weaker. For the Federal Reserve and the Bank of England (BoE) a big part of these
unconventional measures has been the policy of large-scale asset purchases financed by central bank money (Joyce & Spaltro, 2014).
In the United Kingdom conventional and unconventional measures were taken for a monetary easing policy. First the BoE’s Monetary Policy Committee (MPC) cut the official bank rate in a few steps from 3% in December 2008 to 0.5% in March 2009. But despite the conventional measures the BoE took, the MPC judged that without extra measures the nominal output would be too low to meet the 2% inflation target in the medium term (Joyce, Tong, & Woods, 2011). Simultaneously with lowering the official bank rate to 0.5%, the BoE announced an extensive asset purchase program.
This unconventional policy of extensive asset purchases aimed on lowering the long term interest rate is nowadays called quantitative easing (QE). QE is defined as central banks purchasing public and private sector assets using central bank money. In this way the bank is injecting money into the economy to provide an additional stimulus to nominal output in order to meet the inflation target (Benford, Berry, Nikolov, Young, & Robson, 2009). Although conventional measures of central banks influencing the interest rate often contained a form of asset purchases, the quantities were very small compared to the quantities during the QE program. After the first round of the QE program in January 2010 the Bank of England purchased £200 billion in government bonds (gilts), an amount that was equal to 14% of the annual nominal GDP. At the end of 2012 the total amount of purchased assets was close to £375 billion, almost 23% of the annual nominal GDP of the UK (Joyce & Spaltro, 2014).
The literature concerning quantitative easing and the effect this policy used by the major central banks is growing. The lower zero bound is researched by Krugman, Dominquez, & Rogoff (1998) and Eggertsson & Woodford (2003). Most of the
previous research like Joyce, Lasaosa, Stevens, & Tong (2011) focus on the effect of QE in the UK on asset prices. The last couple of years more and more research is focusing on wider economic effects. The macroeconomic effects in the UK have been researched by Kapetanios et al. (2012), Bridges & Thomas (2012) and Pesaran &
4 Smith (2012).
This research is focusing on the effect of quantitative easing in the United Kingdom on the expected inflation. Since the European Central Bank has decided to expand its asset purchase program with the goal to raise inflation, this paper
investigates if there is statistical proof that the extensive QE policy of the BoE had an effect on the expected inflation. The expected inflation is regressed on a QE variable and some control variables.
This paper is structured as follows. First a literature review is given on research regarding the asset purchase programs of central banks. Secondly, an explanation of quantitative easing and its theoretical effects are given. The
subsequent section examines the empirical model and the data. After that the results are presented and evaluated. This paper concludes with the main findings and ideas for future research.
Literature review
The global financial crisis intensified after the collapse of Lehman Brothers in September 2008. This led to a deep recession (Kapetanios et al., 2012).
Governments and central banks from all over the world started introducing measures aimed on stabilizing financial conditions and supporting consumption. Klyuev, De Imus, & Srinivasan (2009) give a review of the measures taken by the G-7 central banks.
The first research on QE mainly relied on event study methods. Bernanke, Reinhart, & Sack (2004) use methods of empirical finance to research the potential effectiveness of nonstandard monetary policies at the zero bound. They focus on Japan, where QE was used for the first time. Bernanke et al. (2004) advise that central banks should maintain a sufficient inflation buffer and easing preemptively as necessary to minimize the risk of hitting the zero lower bound. An inflation buffer is a higher than zero long run average rate of inflation, that protects the economy from downward shocks against deflation. A way central banks can set up an inflation buffer is with inflation targeting, i.e. the 2% inflation target of the BoE.
5 In graph 1 Krugman (2011) shows the state of the economy just after the 2008 crisis in a IS-LM model. In the investment-savings curve the independent variable is given by the interest rate and the dependent variable is given by the output. The IS-curve is downward sloping because lower interest rates encourage higher investment so output rises. For the liquidity-money curve the independent variable is output and the dependent one is the interest rate. The LM-curve gives us all combinations of output and the interest rate for which the money market equilibrium condition is satisfied. The condition holds if real money demand equals the given real money supply. When assumed that real money demand depends positively on GDP and negatively on the cost of holding money r the LM-curve is upward sloping.
Before the 2008 financial crisis the general consensus had been that
conventional monetary policies are able to stimulate economies and real economic growth after a financial crisis. But in the recent crisis the zero lower bound on nominal interest rates has been met. There is a zero lower bound on nominal interest rates because agents can always hold non-interest bearing cash (Joyce et al., 2012). This is why the LM curve is graph 1 is flat at an interest rate of zero.
Krugman et al. (1998) state that Japan already had this problem in the 1990s. They define the liquidity trap as a situation in which conventional monetary policies have become impotent, because nominal interest rates are at or near zero. Also injecting monetary base in the economy has no effect, because bonds and monetary base are considered perfect substitutes by the private sector. The authors also state that theoretically, the only way an economy can get out of this trap is if the central bank can credibly promise to be irresponsible with unconventional measures. If a monetary expansion is considered permanent it will raise prices (Eggertsson & Woodford , 2003).
The use of QE by the Bank of England during the 2008 financial crisis has been documented in a number of studies. Meier (2009) uses an event study
approach to assess the impact of the QE announcements on inflation. He finds that during the first year of QE, the 5-year spot breakeven inflation rate rose by a
significant 45 basis points in the months after the announcements. The 5-year spot breakeven inflation rate, a measure for the expected inflation, is measured by the average difference in yields of nominal and inflation-indexed gilts. When Meier (2009) researches another measure for expected inflation using the 5-year forward rates, the implied rate of inflation five years from now, he surprisingly finds that it fell by 20 basis points. Joyce et al. (2011) investigate the impact of the BoE asset purchase program on UK asset prices using an event studies method and a portfolio balance model. They find that long term gilt yields lowered 100 basis points compared to what it would be without QE. Joyce et al. (2011) also state that compared to the estimates of Gagnon, Raskin, Remache, & Sack (2011) the effects of the Federal Reserve’s purchases have a similar order of magnitude to the BoE.
Recent studies focus more on the broader effects of QE. Kapetanios et al. (2012) focus on the effects of the asset purchases on output and inflation. To conduct a counterfactual analysis to estimate what would have happened without QE, they use three different time series VAR models. The comparison is made by
6 using the estimates found by Joyce et al. (2011), who found that long-term gilt yields were lowered 100 basis points by QE. Although there is considerable uncertainty in counterfactual estimations, Kapetanios et al. (2012) estimate that without QE the real GDP would have been between 1.4% and 3.6% lower and annual CPI inflation
between 1.2% and 2.6% lower. Bridges & Thomas (2012) almost produce the same estimates. They use money accounting techniques to estimate the size of the money supply shock caused by QE and use the estimates in two econometric models.
Pesaran & Smith (2012) emphasize the long run effect of the change in output growth using a counterfactual analysis. Using VAR regressions they conclude that there is a significant impact effect of QE but this effect tends to disappear within a year. In all their cases the long-run effect is not significantly different from zero.
In 2009 the Federal Reserve started an asset purchase program. Chung, Laforte, Reifschneider, & Williams (2012) studied the impact of QE on the US macro economy. With various models combined with the portfolio balance method they find that the combination of the first and second round of asset purchases raised the level of real GDP by 3% and that the inflation was 1% higher. Although QE raised the level of these economic factors the authors suggest that the zero lower bound had a
negative effect on real output and inflation. The authors also state that the
unemployment rate was reduced with 1,5% compared to what it would have been without an asset purchase program. A summary of the research until 2012 is given by Joyce et al. (2012). They conclude that the consensus of the literature is that QE works, but they point out several areas of concern. Because there is not enough historical evidence policymakers have to behave careful.
How does quantitative easing work?
The aim of extensive asset purchases is to reduce the long term interest rate. This will boost investment and stimulate nominal spending, which will generate inflation. The BoE does this to meet their 2% inflation target in the medium term (Joyce et al., 2011). The search for unconventional measures to further loosen monetary
conditions started because the usually reliable relationship between the official bank rate and market rates failed (Joyce & Spaltro, 2014).
With quantitative easing, a central bank purchases financial assets from the private sector such as banks. By buying mainly UK gilts (government bonds) the BoE hugely increases its balance sheet. QE increases the monetary base (‘narrow
money’) when the private sector sells their assets to the central bank. The monetary base is defined as the portion of the commercial banks’ reserves that are maintained at the central bank plus the total currency circulating in public. The central bank does this by crediting the commercial bank’s reserve account with additional funds.
At commercial banks households and companies hold deposits, which they can spend to buy goods. These deposits form the biggest part of what is called ‘broad money’ (Benford et al., 2009). When the BoE buys assets from a non-bank company, it pays for the asset via the seller’s bank. Not only the reserve account of
7 the seller’s bank is credited with extra fund, but the bank credits the account of the seller with a deposit. This means a purchase from a non-bank party does not only raise the ‘narrow money’ but also the ‘broad money’ (Benford et al., 2009).
Purchasing of public and private assets provides different channels through which it can boost spending. There are five different transmission channels through which asset purchases affect the economy (Joyce et al., 2011). First there are three channels working through asset prices.
The first one is the policy signaling effect, which includes all information agents can learn about future monetary policy. QE could for example be a commitment signal that the official bank rate remains low for some time. As
discussed, central banks commit themselves to a QE program because the official bank rate is already at the zero lower bound. This means that the short term interest rate can’t be lowered to decrease the long term interest rate. Nominal interest rates can´t generally be negative because agents can always hold non-interest bearing cash. So if a central bank starts an extensive asset purchase program, a central bank commits itself to a low official bank rate. QE can also be seen as a sign that a central bank does everything to meet the inflation target, which raises the expected inflation (Joyce et al., 2011) .
The second channel is the portfolio balance effect. When a central bank purchases relatively low risk assets, this raises the price of assets. Higher bond prices lower the yields of those bonds. This lowers the long term interest rates which reduces the cost of borrowing for companies and households. Low interest rates raise investments and raises consumption. The higher asset prices will also increase the total wealth of asset holders, which will increase their spending in other assets. With their increased wealth and available capital asset holders buy riskier assets than the ones they sold to the central banks. This increases the yield of the riskier assets. (Benford et al., 2009). Another effect is the liquidity premium effect. At times of distress when markets are dysfunctional, central banks can improve the market by increasing liquidity. By injecting money in the economy, the central bank is making liquidity cheaper and easier to obtain.
Another channel is the confidence effect. An asset purchase program can lead to an improved economic outlook. By demonstrating the central bank’s commitment do whatever it takes to meet the inflation target, it raises the expected inflation. Even with the nominal interest rate fixed at the zero lower bound, this implies that the real interest rates are kept low. Low real interest rates raise investments and even directly boost consumer confidence and their willingness to spend. This confidence can be found back in higher asset prices by a reduced risk premium. At last there are the bank lending effects. When the BoE purchases assets, banks end up with higher reserves. Benford et al. (2009) argue theoretically that a higher level of liquid assets could encourage the banks to extend more loans then they did before. More lending to consumers and businesses leads to higher consumption and investment.
8
Empirical model and data
What is the effect of quantitative easing on inflation expectation in the United Kingdom?
To answer this question an ordinary least squares (OLS) and two two-stage-least-square-regressions (2SLS) are performed with the expected inflation as dependent variable in period t. The expected inflation is regressed on a QE variable and control variables. The following equation is used.
e
= 0 + 1 + 2OR + 3RGDP + 4UR + 5 e -1 + 6 +
t= Time measured in months
e
= The expected annual inflation, in period t
0= The constant in the regression
= The annual inflation measured by the consumer price index annual percentage change, in period t
OR= The monthly average Official Bank Rate set by the Bank of England, in period t
UR= The unemployment rate in percentage rate of the workforce age 16-64, in period t
RGDP= Percentage change in real GDP on same quarter a year ago, per quarter of a year. Three months of the same quarter have the same value.
e
-1= The annual expected inflation, in period t-1
QE= The percentage rate of the nominal accumulated quantity of purchased asset stock under the QE program by the Bank of England by the end of each month of the yearly nominal GDP, in period t
t= The error term, in period t
Prediction of regression coefficient signs
This part describes the expected relationships between the dependent variable and the independent variables. The current inflation is expected to have a positive relationship with the expected inflation. When the inflation at time t is high people expect the inflation to be high at time t+1 because when the inflation is rising people expect it to stay high. The inflation expectations react slowly to the changes in the inflation, so the current inflation takes time to have effect. When the inflation at time t is high it will raise the expected inflation in time t+1. All together the inflation is
expected to have a positive effect on the dependent variable. The ceteris paribus assumption is made for all predictions.
The lagged expected inflation is expected to have a positive effect on the expected inflation. There is lagged correlation because the two time series shifted in time relative to one another. The lagged expected inflation has a delayed response to
9 the expected inflation. This means that when the expected inflation is high at period t-1, the expected inflation will probably be high at time t. The expected inflationt-1 has
a positive relationship with dependent variable.
The Official Bank Rate is expected to have a negative relationship with the dependent variable. In general, when the short term interest rates are lowered this also lowers the long term interest rates. Low long term interest rates make borrowing cheaper which raises investments. When the interest rates go down consumers rather spend their money now than saving it. The increased spending raises the national output. A growing economy raises the inflation which as discussed raises expected inflation.
The expected negative relationship between the expected inflation and the unemployment rate can be derived from the Phillips curve. When the unemployment rate decreases workers are empowered to push for higher wages. With higher wages consumers have more income resulting in consumers demanding more goods and services which causes a higher national output. A higher demand will raise prices which means the inflation rises. As discussed before, a rise in inflation in period t will raise the expected inflation.
The change in real GDP is expected to have a positive effect on the expected inflation. In general, when real GDP is growing people have more to spend which raises the demand for goods and services. A higher demand will raise prices which means the inflation rises. As discussed before, a rise in inflation in period t will raise the expected inflation.
The relationship between the quantitative easing variable and the expected inflation is expected to be positive based on the literature review and economic theory. With QE central banks purchase vast amounts of bonds which raises bond prices and lowers their yields. These lower yields also lower the long term interest rates. Lower long term interest rates make borrowing cheaper which raises
investments. When the interest rates go down consumers rather spend their money now than saving it. The increased spending raises the national output. A growing economy raises the inflation which as discussed raises expected inflation.
Description of the data
The data covers the period from January 1999 up to December 2014. This period is chosen because all variables are available since January 1999. The data is covered up to December 2014 because at time of writing the real GDP change was available up to Q42014. All data, except for the real GDP change, are monthly because GDP data is only published quarterly. The real GDP change data is transformed to monthly data by assigning each month in a quarter the same value. Theoretically it is possible to convert the quarterly data to separate monthly data but the level of tools do so are above this thesis’ academic level. Not the nominal value of the real GDP is taken but the GDP change on the same quarter a year ago. This creates a time series and a static variable.
10 The dependent variable, the expected inflation, is calculated by subtracting the real interest rate from the nominal interest rate. For the real interest rate the monthly average yields are used of 5 year British government index-linked securities. An index-linked-gilt links the income of payments to the consumer price index (CPI). This offers protection for investors against changes in the CPI. The cash flows are
adjusted so that the holder of the bond receives a constant real rate of return. The nominal interest rate is the monthly average yields of 5 year British government securities.
The expected inflation data and the official bank rate are made available by the BoE. The nominal accumulated quantity of purchased asset stock in the QE-program at the end of each month is also published by the Bank of England. To calculate the QE variable the total accumulated nominal value is divided by the annual nominal GDP. The data for inflation, GDP, and unemployment rate are taken from the UK Office for National Statistics. The total number of observations is 192.
Table 1 provides some more descriptive statistics. Under Variable the different variables are listed starting with the dependent variable. Obs gives the amount of observations per variable, mean the average and Std. Dev. gives the standard deviation. Min and Max show the lowest and highest measured value of the variable. The mean, lowest and highest value are higher for the lagged official bank rate than for the official bank rate, because the first measurement in December 1998 was much higher than the average rate.
Variable Obs Mean Std. Dev. Min Max
Expected Inflation 192 2.63110 0.50250 -0.14670 3.82530 QE 192 5.97661 8.55398 0 22.65209 Inflation 192 2.15417 1.06642 0.50000 5.20000 Real GDP Change 192 1.94063 2.14516 -5.80000 4.80000 Unemployment Rate 192 6.13385 1.23435 4.70000 8.50000 Official Bank Rate
192 3.19517 2.16462 0.50000 6.03750 Lagged Inflation 192 2.15990 1.06043 0.50000 5.20000 Lagged Official Bank
Rate 192 3.22599 2.16816 0.50000 6.41670 Lagged Expected Inflation 192 2.62923 0.50332 -0.14670 3.82530 Table 1
11 In the appendix plots of the variables can be found in graph 2 up to graph 7. At the start of the 2008 crisis in September, after the fall of Lehman Brothers, the graphs show the collapse of the British economy. The expected inflation in graph 2 shows a great decrease around November 2008 and even becomes negative. The inflation rate and the real GDP change also show a big decrease around that month in graph 3 and 6. Graph 4 shows us the official bank rate of the Bank of England. After the start of the crisis the BoE quickly decreased the short term interest rate to a level close to the zero lower bound at 0.5%, after which it stayed constant up to now. Graph 5 shows the development of the unemployment rate. It rises after the start of the crisis and slowly returns to its pre-crisis level. At last graph 7 shows us the development of the QE program. The two rounds of asset purchases are visualized by the stair shaped graph.
Hypothesis
Given the literature review and economic theory quantitative easing is expected to have a positive effect on the expected inflation. This is explained in the prediction of regression coefficient signs part. This means a positive 6 is expected. To test this a
significance level of 5% is chosen.
Results
The results of model 1, the results of the OLS-regression of the empirical model, are given in table 2. Because of possible autocorrelation the heteroskedasticity-robust standard errors are used for the tests. The numbers without brackets are the betas, the numbers between brackets are the standard errors.
12 Endogenous variables
An endogenous variable is one that correlates with the error term. Endogeneity can arise as a result of simultaneity. To test for variable endogeneity the Durbin Wu Hausman can be used (Nakamura & Nakamura, 1981). A Durbin Wu Hausman test is done by doing two OLS-regressions. First the expected endogenous variable is regressed on all other independent variables and the instruments. The second step is to calculate the residual from step one. Thirdly the original model is regressed but the residual is added. The fourth step is to test for the significance of the residual
variable. When the coefficient of the residual variable is significantly different from zero in the last regression, the variable is endogenous (Van Ophem, 2014).
According to the Durbin Wu Hausman test the QE variable and the official bank rate variable are endogenous. As discussed the total amount of purchased assets during the QE program has a positive effect on the expected inflation. It could be possible that the expected inflation also has an effect on the QE program. When the expected inflation is falling, the opposite of what central banks want, this could be a trigger for central banks to increase the total amount of assets purchased. Because of the simultaneity the QE variable could be correlated with the error term.
VARIABLE Model 1
Robust standard errors in parentheses
Dependent variable Exp. Inf.
*** p<0.01, ** p<0.05, * p<0.1 QE -0.00593 (0.0042) Inflation -0.045 (0.0292) Real GDP Change 0.0281* (0.0166) Unemployment Rate 0.0482 (0.0301) Official Bank Rate -0.0195
(0.0199) Lagged Expected Inflation 0.868*** (0.0772) Constant 0.192 (0.261) Observations 192 R-squared 0.826 Table 2
13 As discussed before the official bank rate has a negative effect on the expected inflation. It could be possible that the expected inflation also has an effect on the official bank rate. When the expected inflation rises to a high level, this could
overheat the economy. This high expected inflation could be a trigger for the central bank to raise the official bank rate to lower the expected inflation again. This could be the reason for simultaneity.
Table 4 in the appendix shows the results of the last regressions of the Durbin Wu Hausman tests. Durbin Wu Hausman QE shows the endogeneity test for the QE variable. Durbin Wu Hausman OR shows the endogeneity test for the official bank rate variable. The two coefficients of the residual variable are printed bold. Both coefficients are significantly different from zero at a 5% level so both the QE variable and the official bank rate are tested endogenous. To test for both variables to be endogenous at the same time is above the academic level of this paper (Van Ophem, 2014).
2SLS-regression
Because two variables are tested to be endogenous, so correlated with the error term, the results of the OLS-model are biased. To get unbiased results it is possible to use a 2SLS-regression. In Model 2 the Inflationt-1 is used as an instrument for the
QE variable. The lagged inflation has an effect on the QE variable. The inflation rate is one of the economical indicators central banks use to base their decision on to continue with the QE asset purchase program. The effect is lagged because it takes time to implement the QE program after the inflation rate is published. As discussed before the inflation rate has a positive effect on the expected inflation. It is
reasonable to assume that the lagged inflation has no relationship to with expected inflation. This means the lagged inflation can be used as an instrument for the expected inflation.
In Model 3 the official bank rate t-1 is used as an instrument for the official bank
rate. The lagged short term interest rate has a positive effect on the official bank rate. There is lagged correlation because the two time series shifted in time relative to one another. The lagged official bank rate has a delayed response to the official bank rate. This means that when the official bank rate is high at period t-1, the official bank rate will probably be high at time t. As discussed earlier the official bank rate has a negative effect on the expected inflation. It is reasonable to assume that there is no lagged effect so the lagged official bank rate can be used as an instrument.
It is possible to test for weakness of an instrument with a F-test. In the first stage of the 2SLS-regression the endogenous variable is regressed on the
exogenous variables and the instrument. When the F-value of the test is smaller than 10 the instrument is considered weak. In table 5 and 6 the instruments are tested for respectively the QE variable and the official bank rate. The test for the QE variable gives a F-value of 73,57. This means the lagged inflation is a strong instrument for the QE variable. The test for the official bank rate gives a F-value of 10871,21. This
14 means the lagged official bank rate is a strong instrument for the official bank rate. Although both regressions give a F-value above 10, it is clear that compared to one another the instrument for the official bank rate is much stronger.
Ideally, a 2SLS-regression with two instruments for two endogenous variables would be a regression that could be performed. This test is also above this thesis’ academic level so could not be performed in this thesis (Van Ophem, 2014). Because of possible autocorrelation the robust versions of the tests are used. The regressions are performed in STATA. The results are given in table 3. The numbers without brackets are the betas, the numbers between brackets are the standard errors.
VARIABLES Model 2 Model 3
Robust standard errors in parentheses
Dependent variable Exp. Inf. Exp. Inf.
*** p<0.01, ** p<0.05, * p<0.1 QE -0.801 -0.00947* (2.601) (0.0057) Inflation -0.904 -0.0474 (2.801) (0.0302) Real GDP Change 0.405 0.0291* (1.256) (0.0169) Unemployment Rate 1.39 0.0392 (4.455) (0.0273) Official Bank Rate -2.684 -0.0409
(8.696) (0.0297) Lagged Expected Inflation 3.58 0.887*** (8.815) (0.0814) Constant -0.788 0.291 (4.285) (0.246) Observations 192 192 R-squared 0.825 Table 3 Explanation of results
In model 1 there is only one significant variable, the lagged expected inflation. The lagged expected inflation is significant at a 5% level. The coefficient is 0.868 which means that in model 1 a raise of one percent in expected inflation in period t-1 will cause the expected inflation at time t to rise with 0.868%. The sign is positive as expected. All the other variables are insignificant.
The results of the model 2 regression show that none of the results are
significant. A probable reason for this is that lagged inflation is not a good instrument for the QE variable. A possible explanation is that the instrument is not strong
15 enough.
The results of model 3 show one significant variable at a 5% level, the lagged expected inflation. The coefficient is 0.877. This means that in model 3 a raise of one percent in expected inflation in period t-1 will cause the expected inflation at time t to rise with 0.877%. The sign is positive as expected. In all three models the QE
variable is not significant at a 5% level. According to the models used, the QE program did not significantly change the expected inflation.
Discussion
The result of this paper is not in line with other literature. As explained in the literature review of this paper, both research from the US and the UK show that QE did in fact raise the inflation and expected inflation. Although they did not use OLS and 2SLS regressions but generally used time series VAR models, they did find significant results.
One of the possible weaknesses of this research is that the instrument used in model 2 in the 2SLS-regressions could be not strong enough. Although the F-value to test the instrument in model was above 10, it was still far below the F-value for the instrument test for model 3. This could explain why none of the coefficients were significant in model 2. As discussed it is not within this paper’s academic reach to test for this weakness. There could be variables for expected inflation that are omitted from the model, but are correlated with the independent variables.
There are many options for further research. First of all what would be relevant to this paper is to test the model with a 2SLS-regression with two endogenous
variables. For a bachelor student this is above the academic level, but it could be interesting for a master student. It could also be possible to do a endogeneity test for two variables at the same time. It could be especially interesting to include the period where the QE program stops and where the assets are being sold back to the market by the BoE. According to the economic theory the expected inflation should fall
during the period the assets are sold back by the central bank. It could also be interesting to use the models on the Euro zone datasets after the asset purchase program ends in Europe. This can be possible in a couple of years.
Conclusion
The Bank of England started an extensive asset purchase program after the global financial crisis in 2008. One of the most important objectives of the QE program was to raise inflation. Since 2008 the total value of the asset purchase program has risen to almost 23% of the yearly nominal GDP. The ECB just started a QE program but its effects are still unknown.
In this paper an OLS-regression and two 2SLS- regressions are performed. The expected inflation is regressed on a QE variable and several control variables.
16 The main finding of this paper is that quantitative easing did not significantly change the inflation expectations at a 5% level. The implication of this result is that according to the used model, the BoE was not able to significantly alter the expected inflation with a QE program. This result can be important to the ECB to revise their own asset purchase program.
References
Benford, J., Berry, S., Nikolov, K., Young, C., & Robson, M. (2009). Quantitative easing. Bank of England.Quarterly Bulletin, 49(2), 90-100.
Bernanke, B., Reinhart, V., & Sack, B. (2004). Monetary policy alternatives at the zero bound: An empirical assessment. Brookings Papers on Economic Activity, 2004(2), 1-100.
Bridges, J., & Thomas, R. (2012). The Impact of QE on the UK Economy: Some Supportive Monetarist Arithmetic,
Chung, H., Laforte, J., Reifschneider, D., & Williams, J. C. (2012). Have we
underestimated the likelihood and severity of zero lower bound events? Journal of Money, Credit and Banking, 44(s1), 47-82.
Eggertsson, G. B., & Woodford, M. (2003). Zero bound on interest rates and optimal monetary policy. Brookings Papers on Economic Activity, 2003(1), 139-233. Gagnon, J., Raskin, M., Remache, J., & Sack, B. (2011). The financial market effects
of the federal reserve’s large-scale asset purchases. International Journal of Central Banking, 7(1), 3-43.
Joyce, M., Lasaosa, A., Stevens, I., & Tong, M. (2011). The financial market impact of quantitative easing in the united kingdom. International Journal of Central Banking, 7(3), 113-161.
Joyce, M., Miles, D., Scott, A., & Vayanos, D. (2012). Quantitative easing and unconventional monetary policy–an introduction*. The Economic Journal, 122(564), F271-F288.
Joyce, M., & Spaltro, M. (2014). Quantitative easing and bank lending: A panel data approach.
Joyce, M., Tong, M., & Woods, R. (2011). The united kingdom’s quantitative easing policy: Design, operation and impact. Bank of England Quarterly Bulletin, 51(3), 200-212.
Kapetanios, G., Mumtaz, H., Stevens, I., & Theodoridis, K. (2012). Assessing the economy‐ wide effects of quantitative easing*. The Economic Journal, 122(564), F316-F347.
17 Klyuev, M. V., De Imus, P., & Srinivasan, M. K. (2009). Unconventional choices for
unconventional times credit and quantitative easing in advanced economies International Monetary Fund.
Krugman, P. R. (2011). The conscience of a liberal. Retrieved from
http://krugman.blogs.nytimes.com/2011/10/09/is-lmentary/?_r=0
Krugman, P. R. (1998). It's baaack: Japan's slump and the return of the liquidity trap. Brookings Papers on Economic Activity, , 137-205.
Meier, A. (2009). Panacea, curse, or nonevent? unconventional monetary policy in the united kingdom International Monetary Fund.
Nakamura, A., & Nakamura, M. (1981). On the relationships among several
specification error tests presented by durbin, wu, and hausman. Econometrica: Journal of the Econometric Society, , 1583-1588.
Pesaran, M. H., & Smith, R. P. (2012). Counterfactual Analysis in
Macroeconometrics: An Empirical Investigation into the Effects of Quantitative Easing,
Van Ophem, H. (2014). Econometrics slides week 5. Retrieved from
http://blackboard.uva.nl/bbcswebdav/pid-5186283-dt-content-rid-6809587_1/courses/2410E001.6012B0212Y.S11.1.2014/Lecture_5%282%29.pd f
Appendix
Graph 2 -1 0 1 2 3 4 5 1999 Jan 1999 Sep 20 00 May 2001 Jan 2001 Sep 2002 M ay 2003 Jan 2003 Sep 2004 M ay 2005 Jan 2005 Sep 2006 M ay 2007 Jan 2007 Sep 2008 M ay 2009 Jan 2009 Sep 2010 M ay 20 11 J an 2011 Sep 2012 M ay 2013 Jan 2013 Sep 2014 M ayExpected Inflation
Expected Inflation18 Graph 3 Graph 4 Graph 5 0 1 2 3 4 5 6 1999 Jan 1999 Au g 2000 M ar 2000 O ct 2001 M ay 20 01 De c 2002 Jul 2003 Feb 2003 Sep 20 04 Ap r 200 4 N o v 2005 Jun 2006 Jan 2006 Au g 2007 M ar 2007 O ct 2008 M ay 2008 Dec 2009 Jul 2010 Feb 2010 Sep 2011 Ap r 201 1 N o v 2012 Jun 2013 Jan 2013 Au g 2014 M ar 2014 O ct
Inflation
Inflation 0 1 2 3 4 5 6 7 1999 Jan 1999 Sep 2000 M ay 2001 Jan 2001 Sep 2002 M ay 2003 Jan 2003 Sep 2004 M ay 2005 Jan 2005 Sep 2006 M ay 2007 Jan 2007 Sep 2008 M ay 2009 Jan 2009 Sep 2010 M ay 2011 Jan 2011 Sep 2012 M ay 2013 Jan 2013 Sep 20 14 MayOfficial Bank Rate
Official Bank Rate
0 2 4 6 8 10 1999 Jan 1999 Sep 20 00 May 2001 Jan 2001 Sep 2002 M ay 2003 Jan 2003 Sep 2004 M ay 2005 Jan 2005 Sep 2006 M ay 2007 Jan 2007 Sep 2008 M ay 2009 Jan 2009 Sep 2010 M ay 2011 Jan 20 11 Se p 2012 M ay 2013 Jan 2013 Sep 2014 M ay
Unemployment Rate
Unemployment Rate19 Graph 6 Graph 7 -8.00 -6.00 -4.00 -2.00 0.00 2.00 4.00 6.00 1999 Jan 1999 Sep 2000 M ay 2001 Jan 2001 Sep 2002 M ay 2003 Jan 20 03 Se p 2004 M ay 2005 Jan 2005 Sep 2006 M ay 2007 Jan 2007 Sep 2008 M ay 2009 Jan 2009 Sep 2010 M ay 2011 Jan 2011 Sep 2012 M ay 2013 Jan 2013 Sep 2014 M ay
Real GDP Change
Real GDP Change 0 5 10 15 20 25 199 9 Ja n 1999 Au g 2000 M ar 2000 O ct 2001 M ay 2001 Dec 2002 Jul 2003 Feb 2003 Sep 2004 Ap r 2004 N o v 2005 Jun 200 6 Ja n 2006 Au g 2007 M ar 2007 O ct 2008 M ay 2008 Dec 2009 Jul 2010 Feb 2010 Sep 2011 Ap r 2011 N o v 2012 Jun 201 3 Ja n 2013 Au g 2014 M ar 2014 O ctQE
QE20 Table 4 VARIABLES Durbin Wu Hausman QE Durbin Wu Hausman OR Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 QE -0.801*** -0.00947** (0.216) (0.0042) Inflation -0.904*** -0.0474*** (0.234) (0.0176) RealGDPChange 0.405*** 0.0291*** (0.103) (0.0095) UnemploymentRate 1.390*** 0.0392 (0.365) (0.0258) LaggedExpectedInflation 3.580*** 0.887*** (0.737) (0.0389) OfficialBankRate -2.684*** -0.0409** (0.723) (0.0191) QE_res 0.795*** (0.216) OR_res 0.778*** (0.115) Constant -0.788** 0.291 (0.329) (0.18) Observations 192 192 R-squared 0.838 0.861
Durbin Wu Hausman QE shows the endogeneity test for the QE variable. Durbin Wu Hausman OR shows the endogeneity test for the official bank rate variable. The numbers without brackets are the betas, the numbers between brackets are the standard errors.
21 Table 5 Table 6 _cons -37.36942 3.101809 -12.05 0.000 -43.48866 -31.25017 LaggedExpectedInflation 1.108322 .9255915 1.20 0.233 -.7176851 2.934329 LaggedInflation .3177642 1.30751 0.24 0.808 -2.261691 2.89722 UnemploymentRate 6.906882 .4071317 16.96 0.000 6.103692 7.710071 RealGDPChange .7232209 .2433868 2.97 0.003 .2430674 1.203375 Inflation -1.867844 1.266384 -1.47 0.142 -4.366166 .630479 QE Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 13975.5673 191 73.1705093 Root MSE = 5.0234
Adj R-squared = 0.6551 Residual 4693.54275 186 25.2341008 R-squared = 0.6642 Model 9282.02453 5 1856.40491 Prob > F = 0.0000 F( 5, 186) = 73.57 Source SS df MS Number of obs = 192
_cons -.3534938 .1230606 -2.87 0.005 -.5962677 -.1107199 LaggedExpectedInflation .2175656 .0238227 9.13 0.000 .1705682 .264563 LaggedOfficialBankRate .9652586 .0081823 117.97 0.000 .9491164 .9814007 UnemploymentRate -.01464 .0166255 -0.88 0.380 -.0474387 .0181587 RealGDPChange .0065334 .0062377 1.05 0.296 -.0057724 .0188391 Inflation -.0279253 .0114592 -2.44 0.016 -.050532 -.0053187 OfficialBankRate Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 894.94516 191 4.68557675 Root MSE = .1281
Adj R-squared = 0.9965 Residual 3.05195333 186 .016408351 R-squared = 0.9966 Model 891.893207 5 178.378641 Prob > F = 0.0000 F( 5, 186) =10871.21 Source SS df MS Number of obs = 192
